doi
stringlengths 28
28
| title
stringlengths 19
311
| abstract
stringlengths 217
5.08k
| plain language summary
stringlengths 115
4.83k
| article
stringlengths 3.87k
161k
|
|---|---|---|---|---|
10.1371/journal.ppat.1000282 | Local Inflammation Induces Complement Crosstalk Which Amplifies the Antimicrobial Response | By eliciting inflammatory responses, the human immunosurveillance system notably combats invading pathogens, during which acute phase proteins (CRP and cytokines) are elevated markedly. However, the Pseudomonas aeruginosa is a persistent opportunistic pathogen prevalent at the site of local inflammation, and its acquisition of multiple antibiotic-resistance factors poses grave challenges to patient healthcare management. Using blood samples from infected patients, we demonstrate that P. aeruginosa is effectively killed in the plasma under defined local infection-inflammation condition, where slight acidosis and reduced calcium levels (pH 6.5, 2 mM calcium) typically prevail. We showed that this powerful antimicrobial activity is provoked by crosstalk between two plasma proteins; CRP∶L-ficolin interaction led to communication between the complement classical and lectin pathways from which two amplification events emerged. Assays for C4 deposition, phagocytosis, and protein competition consistently proved the functional significance of the amplification pathways in boosting complement-mediated antimicrobial activity. The infection-inflammation condition induced a 100-fold increase in CRP∶L-ficolin interaction in a pH- and calcium-sensitive manner. We conclude that the infection-induced local inflammatory conditions trigger a strong interaction between CRP∶L-ficolin, eliciting complement-amplification pathways which are autonomous and which co-exist with and reinforce the classical and lectin pathways. Our findings provide new insights into the host immune response to P. aeruginosa infection under pathological conditions and the potential development of new therapeutic strategies against bacterial infection.
| An opportunistic pathogen, Pseudomonas aeruginosa causes local inflammation in various tissues and is an important cause of hospital-acquired infections. While this may manifest as a hot-tub rash, failure to limit such localized inflammation may cause life-threatening septic shock to ensue, leading to multiple organ failure and possibly death. Effective treatment with antibiotics is complicated by the increasing prevalence of multidrug-resistant P. aeruginosa. In this study, we demonstrated that under typical local infection-inflammation conditions of slight acidosis and reduced calcium levels in the blood, P. aeruginosa is killed effectively. This is because such a milieu provokes crosstalk between two proteins involved in host defense, CRP and L-ficolin, which triggers communication between antimicrobial complement pathways that were previously presumed independent of each other. The CRP∶L-ficolin interaction amplifies their antimicrobial activity and enhances bacterial clearance. Insights gained from this host–pathogen interaction mechanism offer fresh perspectives on the potential development of complement immune therapies to enhance eradication of the invading pathogen.
| The human immune system is notable for its ability to combat infectious microorganism by eliciting inflammatory responses [1]. Acute phase proteins such as CRP [2],[3] and cytokines [4] are elevated markedly in association with infection and inflammation. During this process, local acidosis occurs due to massive infiltration of neutrophils and macrophages [5] to the site of infection, which subsequently, activates the respiratory burst [6],[7] in many infection-inflammation related diseases such as trauma-induced infection [8], acute renal failure [9], and intra-abdominal infection [10]. These pathological conditions can decrease the pH to as low as 5.5–7.0 [11]. Simultaneously, mild hypocalcaemia is closely associated with bacterial infection [12]. One possibility is that intracellular NF-κB activation requires the influx of calcium into the immune cells [13], which is strengthened under acidic condition [14],[15], causing a transitory drop in the extracellular calcium at the infection site. The transient changes in the (i) levels of acute phase proteins, (ii) local pH and (iii) calcium concentration, all contribute to a typical infection-inflammation environment generally proposed to result from pathogenic metabolic disorder [16]. However, the concurrence of the infection-inflammation and the changes in pH, calcium and acute phase protein concentrations indicate that these pathophysiological conditions might also be essential for effective host defense.
Emerging evidence has demonstrated that a transient drop in pH and calcium level is crucial for triggering many immune processes, for example, TLRs 3, 7, 8 and 9 all require an acidic environment for their activation of the endosomes [17]. Extracellular acidosis was also found to boost adaptive immunity by activating dendritic cells [18], and CD8+ T cells in the peripheral tissues to improve MHC class I-restricted Ag presentation by neutrophils [19],[20]. However, complement activity, which is the major frontline host defense expected to occur under such typical infection-inflammation condition, is hardly explored even though it elicits a more rapid and direct antimicrobial action against the invading bacteria.
CRP and ficolins are known initiators of the complement classical pathway [3] and lectin pathway [21], respectively, and they are the key molecules that boost the immune response [22]. As an acute phase inflammation marker [23], CRP is also a multifunctional protein [3] upregulated in many diseases such as acute pneumonia [24], myocardial infarction [25] and atherothrombosis [26]. Besides binding to a wide range of ligands including phosphorylcholine (PC), polycations and polysaccharides displayed on the surface of the invading bacteria [27],[28], the CRP was also found to be deposited at the site of injury [2] indicating its crucial role in local inflammation. Similarly, as a pattern recognition receptor (PRR), ficolin which is composed of collagen-like domain and fibrinogen-like domain (FBG), recognizes lipoteichoic acid [29] of Gram positive bacteria; lipopolysaccharide of Gram negative bacteria; and 1,3-β-D-glucan of fungi; via the acetyl group on the N-acetylglucosamine (GlcNAc) moiety of these pathogen-associated molecular patterns [22],[30],[31]. There are three ficolin isoforms; L- and H- ficolins are soluble serum proteins whereas M-ficolin is mainly associated on the monocyte membrane, with very low concentration (∼60 ng/ml) in the serum [32]. Patients with ficolin disorder are susceptible to inflammation brought about by respiratory infections [33] and Behçet's disease [34]. Importantly, both CRP and ficolin undergo calcium- and pH- regulated conformational change when binding to their respective ligands [35],[36], indicating that their functions might be modulated by inflammation. However, no pathophysiological significance was proposed for CRP and ficolins although Ng et al. (2007) reported the functional relationship of their respective homologues, CrCRP and CL5, in the horseshoe crab.
As the most common cause of pneumonia in intensive care unit and the second most common cause of nosocomial pneumonia, the Pseudomonas aeruginosa is a ubiquitous opportunistic pathogen which easily overcomes immunocompromised patients. Its acquisition of multiple antibiobitic-resistance factors [37],[38] poses a grave challenge to drug manufacturers and patient healthcare management. Here, we found that P. aeruginosa, which thrives in local infections of the urinary tract [39], soft tissue [40], bone and joint [41], was effectively killed within 30 min in a typical infection-inflammation condition in contrast to normal condition. To determine the underlying mechanisms of the antimicrobial action, the interaction between ficolins and CRP was extensively characterized. Importantly, we demonstrated that L-ficolin and CRP collaborate through protein-protein crosstalk resulting in two amplification pathways which reinforce the classical and lectin complement pathways, and ultimately control downstream complement events like C4 deposition and opsonization of the microbe. These results emphasize the critical role of the typical “infection-inflammation condition” in provoking the complement amplification pathways, which the host uses to effectively fight against invading pathogens such as P. aeruginosa.
Calcium concentrations of 2 and 2.5 mM were employed to represent mild hypocalcaemia and normal conditions, respectively, since calcium concentration in healthy serum ranges from 2.2 to 2.6 mM [12] as compared to <2.12 mM [42] in the serum of patients with infection. pH 6.5 was reportedly typical of local acidosis [11]. CRP levels in normal (n = 5) and patient sera (n = 9) were measured and patients with acute phase infection recruited in this study typically showed serum CRP level of 10 µg/ml (Figure S1), which was used in subsequent experiments. Hence, two typical conditions were defined: (i) the infection-inflammation induced local acidosis (pH 6.5, 2 mM calcium in infected serum with CRP of 10 µg/ml) and (ii) normal condition (pH 7.4, 2.5 mM calcium in normal serum with CRP<0.5 µg/ml). Henceforth, these two conditions are referred as “normal” and “infection-inflammation” condition (mimicking local acidosis). Unless otherwise stated, the two buffers used to dilute the serum to mimic the normal and infection-inflammation conditions are: TBS buffer containing 25 mM Tris, 145 mM NaCl, pH 7.4 and 2.5 mM CaCl2 and MBS buffer containing 25 mM MES, 145 mM NaCl, pH 6.5 and 2 mM CaCl2 [43],[44].
As a proof-of-concept and to demonstrate the prowess of the two typical conditions ascribed above, we compared the bactericidal activity against P. aeruginosa, a clinically challenging pathogen commonly found at the site of local inflammation. Figure 1A and Figure S2A show that within 30 min, under the infection-inflammation condition, 97% of the bacteria was killed in the patient serum (Video S1) whereas under the normal condition, <10% succumbed in the healthy serum (Video S2) indicating that the infection-inflammation condition elicits a highly robust antibacterial action. P. aeruginosa incubated with two buffers (TBS, pH 7.4, 2.5 mM CaCl2 or MBS, pH 6.5, 2 mM CaCl2) without serum remained viable (Video S3 and Video S4), suggesting that the enhanced bactericidal activity shown in Figure 1A was attributed to the serum components. Heat-inactivated patient serum restored the bacterial survival to 90% (Video S5), suggesting the involvement of the complement system, which is highly heat-sensitive. However, the patient serum failed to kill Staphylococcus aureus (Video S6), known to astutely evade the complement system [45], indicating that the local acidosis-mediated killing effect targets complement-susceptible pathogens.
To delineate the mechanisms of complement enhancement by infection-inflammation, we focused on the CRP and ficolins, two pH- and calcium- sensitive components of the complement system. Here, the antibacterial activity resulting from bacteria incubated with heat-inactivated serum was used as a background control to normalize off any potential effect of other heat-resistant serum factors. Figure 1B shows that within 1 h, the patient serum elicited >95% antibacterial activity compared to 30% by healthy serum whereas the two buffers (TBS or MBS) did not kill the bacteria (Figure S2B). However, depletion of either CRP or ficolin from the sera drastically decreased the antimicrobial activity to 20–30% for both of the healthy and patient sera, indicating that in the infection-inflammation condition, CRP and ficolin are necessary to trigger efficient bactericidal activity. Interestingly, we observed synergistic action of CRP and L-ficolin when both proteins were present in the patient serum under infection-inflammation condition, which accounted for ∼50% of the enhanced killing effect. To verify this synergistic effect, serum depleted of both CRP and ficolin (by PC- or GlcNAc- beads) was used for incubation with the bacteria during which increasing concentrations of CRP or L-ficolin or both of the proteins were added. The results confirmed that addition of the two proteins exhibited a significant amplification of bacterial killing compared to adding either of the two proteins, although this process did not restore the antimicrobial effect to the same level as that of the original undepleted serum (Figure 1C). This implies that some other serum factors might have been lost through their association with the GlcNAc or PC beads together with L-ficolin or CRP, and that they (for example mannose binding lectin [46]) might also contribute to the antimicrobial activity via complement pathways. To further confirm the significance of the infection-inflammation condition, normal healthy serum was supplemented with 10 µg CRP and incubated with P. aeruginosa to simulate a condition where higher CRP level prevails (but without infection) such as in cardiovascular disease. Our results indicated that without infection-inflammation condition, a high CRP level did not bring about antimicrobial activity (Figure 1D). Therefore, we envisaged potential interaction and/or co-operation between CRP and L-ficolin which might be strengthened under infection-inflammation condition, causing amplification of the complement activity and higher bactericidal activity. To ascertain this possibility, P. aeruginosa was incubated with either of the purified CRP and L-/H- ficolins (major serum-type ficolins) or with these purified protein plus serum components. Results showed that although neither of the purified CRP nor L-ficolin was bound directly to P. aeruginosa (in the absence of serum), the CRP was bound indirectly but equally well to P. aeruginosa in the patient or healthy serum independent of depletion of ficolins (Figure 1E). Furthermore, in patient serum under infection-inflammation condition, CRP enabled the indirect association of L-ficolin to the invading bacteria. This was confirmed by a drop in the level of bound L-ficolin in CRP-depleted serum. However, the purified H-ficolin on its own (used here for comparison with L-ficolin), was bound directly to the bacteria under infection-inflammation condition, independent of CRP (Figure 1E). The low homology (42%) between H-ficolin and L- ficolin (Figure S3), may explain the lack of interaction between H-ficolin and CRP. Thus, our data suggest: (i) the possibility of the PRR∶PRR interactions between L-ficolin and CRP, (ii) that in the healthy serum, only CRP binds to bacteria resulting in minimal complement activity, (iii) in infection-inflammation condition, CRP enables L-ficolin to bind bacteria, which upregulates the complement pathway.
Based on the in vitro bactericidal results we hypothesized that CRP and L-ficolin might interact in the patient's serum under infection-inflammation condition with mild acidosis. Thus, it was imperative for us to investigate whether CRP has a propensity to complex with L-ficolin in the patient serum. To delineate the effect of different serum factors, we tested the potential interaction between CRP and L-ficolin by varying the pH, calcium and CRP levels. Consistent with the antibacterial results, co-immunoprecipitation (Co-IP) of the patient sera under pH 6.5 showed strongest complexes of CRP∶L-ficolin (Figure 2A) independent of the calcium level. However, at pH 7.4, increasing calcium was found to dramatically inhibit the CRP∶L-ficolin interaction, indicating that under normal condition, calcium prevents CRP∶L-ficolin interaction. Furthermore, the protein complex could not be isolated from the healthy serum regardless of pH and calcium status (Figure 2B, lanes 2–17). This might be because healthy individuals have a dramatically lower CRP level (100–1000× less than acute phase patients; see Figure S1). To confirm this, we reconstituted the healthy serum with 10 µg/ml CRP (acute phase level) and adjusted the pH and calcium to infection-inflammation condition. Indeed, we were able to isolate CRP∶L-ficolin complex from the simulated patient serum (Figure 2B, lanes 18–21). In contrast, H-ficolin was shown not to complex with CRP in vivo, regardless of conditions (Figure 2A and 2B), indicating the specificity of CRP for its cognate ficolin partners.
To substantiate that CRP and ficolins interact directly, the three ficolin isoforms, which share some common functions such as complement activation [47]–[49], were purified (Figure S4) and ELISA was used to test their interaction status with CRP. Figure 2C shows that only L- and M- ficolins exhibited dose-dependent binding to immobilized CRP whereas H-ficolin, which is independent of calcium and orders of immobilization, did not associate with CRP (Figure S5A). On the premise that both ficolins and CRP can anchor to the invading bacteria directly or indirectly [22],[29], we also analyzed the reverse orders of binding by immobilizing the ficolins first to a solid phase followed by addition of CRP. Figure 2D shows that the interactions between CRP and L-/M-ficolins were weaker when ficolins were immobilized first indicating that prior immobilization of CRP appeared to be a preferred position/orientation for subsequent binding of ficolins. This is consistent with our observation that after anchorage, CRP might recruit ficolin to the bacterial surface (Figure 1D). Our observation that M-ficolin also bound to immobilised CRP (Figure 2C), prompted us to perform yeast two-hybrid analysis (Figure S5B) to confirm their interaction since M-ficolin, the major membrane-associated form of ficolin, has also been implicated in complement activation [47] and phagocytosis of pathogens [50]. Our results show that besides serum L-ficolin∶CRP interaction, the M-ficolin∶CRP interaction also occurred (Figure S5B), and that this partnership is probably significant to intracellular/downstream activities.
As both CRP and ficolins are hitherto well known to separately trigger the classical and lectin complement pathways, we hypothesize that the interaction between CRP and L-ficolin under infection-inflammation condition might connect these two pathways and potentially ramify new conduits to potentiate the bactericidal activity in patients with infection-induced mild local acidosis. Thus, C4 cleavage assay was performed to investigate the complement activity via sequential incubation of PC-beads with CRP, L-ficolin, MASP-2 and C4 (Figure 3A, left panel) or GlcNAc-beads with L-ficolin, CRP, C1 complex and C4 (Figure 3A, right panel). The C4 cleavage product, C4b (α,β,γ, chains), resulting from both of the reactions only emerged under infection-inflammation condition. These results suggest that both orders of interactions: L-ficolin∶CRP and CRP∶L-ficolin produced functional complement components under pathophysiological conditions although prior anchorage of CRP was apparently the preferred position for interaction with L-ficolin (Figure 1E and Figure 2D). To confirm that the opsonized particle generated by interaction between CRP and L-ficolin can be phagocytosed, and to avoid background interaction between other receptors directly/independently with pathogen-associated molecular patterns on the bacterial membrane, we incubated the above C4-deposited beads rather than opsonized bacteria with phagocytes. Consistent with the C4 deposition results, only beads with all the components of the amplification pathway added in the infection-inflammation condition underwent significant opsonization and phagocytosis within 15 min (Figure 3B and Figure S6). Under infection-inflammation condition, the amplification pathways 1 and 2 induced phagocytic efficiencies up to 70% and 54.3%, respectively, with <5% in the other controls (Figure 3C). Taken together, we propose that inflammation drives crosstalk between CRP and L-ficolin from which two new autonomous complement amplification pathways emerge leading to membrane attack complex (MAC) formation and antimicrobial activity:
To determine whether C1q and L-ficolin can bind to CRP simultaneously and which of the two amplification pathways is dominant, competition assay was performed. Under either infection-inflammation or normal condition, the addition of increasing amounts of L-ficolin to a fixed concentration of C1q did not dissociate C1q from CRP and vice versa (Figure 4A and 4B) indicating that C1q and L-ficolin might bind to different domains of the CRP molecule. Similarly, under either normal or infection-inflammation condition, CRP and MASP-2 did not compete with each other for L-ficolin (Figure 4C and 4D). Interestingly, under infection-inflammation condition, we detected a 5-fold increase in the CRP∶L-ficolin interaction but the CRP∶C1q interaction was reduced to half compared to physiological condition. In comparison, both the binding between L-ficolin∶MASP-2 and L-ficolin∶CRP became significantly increased in infection-inflammation condition indicating that the activities of the four pathways (classical, lectin, amplication 1 and amplification 2) might be tightly regulated to kill the pathogens effectively. Overall, the two amplification pathways do not compete against the classical and lectin-mediated pathways, rather, they integrate and boost the classical and lectin-mediated complement pathways towards a common goal of overcoming the pathogen more effectively while avoiding complement over-reaction.
To delineate the mechanism underlying the CRP∶L-ficolin interaction induced by the infection-inflammation condition, the molecular interaction between these two proteins were analyzed. Given that ficolins are composed of FBG and collagen-like domains, and that the FBG domain harbors the ligand binding sites [51],[52], we hypothesize that CRP binds to the FBG domain of L-ficolin. To prove this, the recombinant FBG domain of L-ficolin (henceforth referred as L-rFBG) was expressed and purified (Figure S4). ELISA showed that L-rFBG displayed dose-dependent binding to CRP immobilized on Maxisorp™ plates (Figure 5A), indicating that CRP binds to the L-rFBG. However, when L- rFBG was immobilized first, its binding of CRP was weaker (Figure 5B), which is consistent with our observation that prior anchorage of CRP to the bacterial surface is the preferred position for the CRP∶ficolin interaction (Figure 2D). Real-time biointeraction analysis showed binding affinity between CRP∶L-rFBG to be 1.11×10−6 M under normal condition whereas infection-inflammation condition induced a 100-fold increase in their affinity (KD = 1.26×10−8 M) (Figure 5C and 5D), which triggers two amplification pathways. The lower binding affinity between CRP and L-ficolin under normal condition suggests that under physiological condition, the two proteins only co-exist and not interact so as to avoid random complement activation.
To demonstrate how infection-inflammation condition triggers the interaction between CRP and ficolins, we examined their binding characterisitics over a range of pH 5.5 to 7.4 under either 2 mM or 2.5 mM calcium. Considering that prior anchorage of CRP is the preferred position for the recruitment of L-rFBG (Figure 5B), we immobilized CRP first. Figure 5E shows that the interaction between CRP and L-rFBG was influenced dramatically by pH, with stronger binding under acidic condition, implying that CRP∶L-ficolin interaction is triggered by the infection-inflammation condition whereas under normal condition, the CRP and L-ficolin co-exist with very weak or no complexation, possibly to avoid undesirable complement-enhancement. Furthermore, it was observed that interaction under 2 mM calcium showed stronger binding compared to that under 2.5 mM calcium. This is consistent with the previous observations that calcium influences the conformations of CRP and ficolins [51],[53] and suggests that blood calcium concentration may regulate the recruitment of ficolin to the CRP when the later is anchored on the bacterial surface. Overall, our data indicate that CRP∶L-ficolin interaction is enhanced by low pH and low calcium, which occur under infection-inflammation condition. Thus it is possible that CRP anchored to the invading bacteria is recognized by L-ficolin, and their interaction potentially activates the amplification pathways.
At this juncture, it was imperative for us to show whether the two amplification pathways bridged by CRP∶ficolin crosstalk were responsible for the enhanced bactericidal activity of P. aeruginosa under infection-inflammation condition (see Figure 1). Since C3-deposition is the pivotal step towards the formation of the MAC, we sought to detect C3 deposition on the bacteria by incubating P. aeruginosa with whole serum, or serum depleted of CRP or ficolin in the healthy or infection-inflammation conditions. Figure 6A shows that bacteria incubated with patient serum in infection-inflammation condition had higher C3 deposition compared to normal serum. However, when the serum was depleted of CRP or ficolin, C3 deposition was reduced dramatically even in infection-inflammation condition, resulting in very little difference in the level of C3-deposition under the two conditions. This indicates that: (i) the presence of CRP and ficolin, and (ii) their interaction, were crucial to C3-deposition particularly in the infection-inflammation condition. Furthermore, after subtracting readings from the both the CRP- and ficolin- depleted serum (to exclude the effects of other serum factors), flow cytometry quantified a 7-fold greater C3-deposition on the P. aeruginosa induced by CRP∶L-ficolin interaction in the whole patient serum (Figure 6B) compared to that in the whole normal serum (Figure 6C). Furthermore, stepwise depletion of ficolin and CRP from the patient serum led to progressive loss in C3-deposition. Interestingly, in infection-inflammation condition, the combined presence of CRP and ficolin in the patient serum induced 58.2% C3-deposition compared to 24.3% with serum depleted of both CRP and ficolin (Figure 6B), a difference of 33.9% (Figure 6C). In contrast, CRP- or L-ficolin- depleted serum only induced 2.1% or 14.2% more deposition of C3, respectively (Figure 6C). Remarkably, the increase in C3-deposition in the presence of CRP and L-ficolin under infection-inflammation condition (pH 6.5, 2 mM calcium) was higher than the sum of the increased effect from either of the two individual proteins in the serum, demonstrating synergistic effect of the CRP∶ficolin interaction, and the enhancement of the complement activity against the invading bacteria. This is the underlying reason for the more efficient killing of the bacteria in such local inflammation condition. In healthy individuals, we envisage that these two molecules only co-exist and do not interact with each other, thus preventing random amplification of the complement activity, which would otherwise be detrimental to the host.
Local acidosis, mild hypocalcaemia and high CRP level are characteristic of local inflammation. By defining a typical infection-inflammation condition (pH 6.5 and 2 mM calcium), under which the serum kills the P. aeruginosa more effectively compared to normal physiological condition (pH 7.4 and 2.5 mM calcium), we demonstrated crosstalk between CRP and ficolin, which resulted in two new autonomous amplification pathways leading to a synergistic level of C3-deposition on the bacteria. This serves to boost the bactericidal activity in the serum under local infection-inflammation.
The complement system is a major microbicidal force in the infection-inflammation condition (Figure 1A) mediated by the synergistic effect of CRP and ficolins (Figure 1B–1D). We propose that CRP interacts with L-ficolin under infection-inflammation condition, enhancing the complement-activated killing of the P. aeruginosa. Although it was shown that purified CRP and L-ficolin did not bind to P. aeruginosa (Figure 1E), the presence of serum factors enabled CRP to bind to bacteria indirectly and the bound CRP recruited L-ficolin (Figure 1E). Co-IP confirmed that the CRP∶L-ficolin complex formation was induced in the serum by infection-inflammation condition (low pH, low calcium level and high CRP level) (Figure 2A and 2B) and ELISA demonstrated that L- and M- ficolins displayed stronger binding to CRP when the latter was anchored first to bacteria (Figure 2C and 2D).
Hitherto, two complement pathways involving CRP and L-ficolin are well-established [21],[28]. Here, we provide evidence for the inflammation-induced crosstalk between CRP and L-ficolin which impels two new amplification pathways:
The cleavage of C4 (Figure 3A) demonstrated that these two amplification pathways were functional only under infection-inflammation condition but are negligible in healthy condition. The opsonized particles generated by these amplification pathways were recognized and removed by the immune cells indicating the autonomy and effectiveness of the two new pathways (Figure 3B) which were further confirmed by quantification of the phagocytic efficiency (Figure 3C). Competition assays showed that the amplification pathways do not compete against the classical and lectin pathways (Figure 4), but rather, they boost these two well-known complement pathways.
To unequivocally demonstrate that the CRP∶L-ficolin interaction is triggered by infection-inflammation condition, we showed that the L-rFBG interacts avidly with CRP (Figure 5A) at KD of 1.26×10−8 M under infection-inflammation condition compared to 1.11×10−6 M under normal condition (Figure 5C and 5D). This represents a 100-fold increase in affinity between the two proteins. A positional preference was evident, where prior anchorage of CRP resulted in a more efficient recruitment of the L-ficolin (Figure 5B). Low pH dramatically enhanced CRP∶L-ficolin interaction (Figure 5E), showing that inflammation-induced acidosis promoted stronger crosstalk between CRP and L-ficolin. Furthermore, lower level of calcium enhanced the CRP∶L-ficolin interaction (Figure 5E), thus indicating that calcium is a main regulator of the molecular crosstalk between L-ficolin and CRP. Taken together, our findings explain why and how a typical infection-inflammation condition provokes crosstalk between CRP and the L-ficolin, which boosts the complement system. This key mechanism is further confirmed by the fact that in infection-inflammation condition, the CRP∶L-ficolin amplification pathway caused more C3-deposition (Figure 6).
In conclusion, we provide evidence that local infection-inflammation elicits new complement amplification mechanisms to fight the invading pathogen (Figure 7). This novel antimicrobial mechanism is particularly effective against the P. aeruginosa, an opportunistic pathogen which causes mortality in critically ill and immuno-compromised patients [39]. Both the immunoevasive nature [54] of P. aeruginosa as well as its acquisition of multi-drug resistance [55] makes elimination of this microorganism a particular challenge. To date, the effective therapy against P. aeruginosa infection remains indistinctive. However, insights gained into the mechanisms of complement amplification shown in the present work are crucial in understanding the host defense to counter pathogen immune evasion, which will contribute to the development of complement-immune therapies. Furthermore, the arena of complements is far from saturation and there are still intriguing forward and reverse defense mechanisms awaiting elucidation and understanding.
All experiments were performed according to national and institutional guidelines on ethics and biosafety (Institutional Review Board, Reference Code: NUS-IRB 06-149).
Human C1 complex, C4, CRP and polyclonal goat anti-human CRP antibody were purchased from Sigma. The purity of CRP was verified by SDS PAGE and Mass Spectrometry (Figure S7A and S7B). C1q was from Quidel (San Diego, CA). L- and H-ficolins were purified from human serum. Recombinant MASP-2 was expressed and purified according to Vorup-Jensen et al. [56]. M-ficolin and L-rFBG were recombinantly expressed and purified (in Protocol S1). Mouse anti-L- and H- ficolin monoclonal antibodies and rabbit anti-M-ficolin polyclonal antibody were from Hycult (Uden, The Netherlands). Goat anti-rabbit, rabbit anti-goat, goat anti-mouse secondary antibodies with HRP-conjugation, and polyclonal rabbit anti-C3d, anti-C4c and anti-C1q antibodies were from Dako A/S (Glostrup, Denmark). Goat anti-rabbit antibody with R-phycoerythrin conjugation and mouse anti-myc antibody were purchased from Invitrogen (Carlsbad, CA). Polyclonal goat anti-MASP-2 antibody was purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Serum samples were obtained from adult male healthy and infected patient volunteers with informed consent. Clinical data indicated infection-inflammation in these patients. CRP- or ficolin- depleted serum was generated by incubating the serum with p-aminophenyl phosphorylcholine (PC)- or GlcNAc- Sepharose beads for 2 h at room temperature. The concentration of the sera was measured by NanoDrop™ ND-1000 Scientific, Wilmington spectrophotometer (Thermo Fisher Scientific, DE) to ensure that equal amounts of proteins were added. 10% (v/v) healthy serum was prepared by diluting the normal healthy serum in TBS buffer (25 mM Tris, 145 mM NaCl, pH 7.4, 2.5 mM CaCl2) buffer and 10% (v/v) patient serum was prepared by diluting the patient serum in MBS buffer (25 mM MES, 145 mM NaCl, pH 6.5, 2 mM CaCl2) unless otherwise stated. The CRP level was determined (Figure S1) using CRP Bioassay ELISA kit (BD Biosciences, San Jose, CA).
P. aeruginosa strain PAO1 is commonly used in many studies including trials on multi-drug resistance [37],[38], simulation of P. aeruginosa infection [57],[58] and biofilm formation [59]. The lab-adapted PAO1 [60],[61] used in this study was kindly provided by Professor B. H. Iglewski (University of Rochester, Rochester, USA). The GFP functional fragment was cloned into the Hind III and Sma I sites in the pDSK 519 vector to form the pDSK-GFP plasmid which was subsequently transformed into PAO1 (PAO1-GFP). S. aureus strain PC1839 provided by Professor S. Arvidson (Karolinska Institutet, Stockholm, Sweden) was transformed with plasmid pALC 1420 (PC1839-GFP) [62]. The PAO1-GFP and PC1839-GFP which express enhanced GFP were used in the in vitro bacterial killing assay. The PAO1 without pDSK-GFP was used in flow cytometry experiment to avoid fluorescence quenching between different dyes.
Single colonies of P. aeruginosa PAO1 and S. aureus PC1839 were inoculated into 10 ml of Luria-Bertani broth and Tryptone Soy Broth, respectively, and shaken at 220 rpm for 16 h at 37°C. An aliquot of 100 µl overnight culture was inoculated into 10 ml fresh medium and shaken at 220 rpm, 37°C for 3.5 h until OD600 nm reached 1.0. Bacteria cultures were collected by pelleting at 5000 g for 10 min, followed by washing with 145 mM saline, and the bacteria were resuspended to OD600 nm = 10. Aliquots of 10 µl bacteria were collected by centrifugation. The P. aeruginosa was incubated with 10% patient or normal serum at 37°C for 60 min. The reaction was ceased on ice. The remnant bacterial population in the reaction mixture was enumerated by plating 100 µl of serially diluted samples on agar plates and incubating at 37°C overnight. To standardize the effect of CRP∶L-ficolin interaction in complement activation, and to exclude the effects of other mechanisms, a parallel experiment was set up with bacteria incubated in heat-inactivated serum, which served as a control to account for 0% killing. The % bacterial killing was calculated as follows:
The real-time imaging of the bacterial clearance elicited by the serum (with or without CRP∶L-ficolin interaction) was conducted by mixing 10 µl aliquot of bacteria (P. aeruginosa PAO1-GFP or S. aureus PC1839-GFP) with 100 µl 10% serum. Three µl of the mixture was examined by real-time fluorescence microscopy (Axio observer Z1, Carl Zeiss Inc, Thornwood, NY). Images were taken at intervals of 15 s for 30 min, at magnification of 63×1.6 and a movie was made. Bacteria samples incubated with diluting buffers (TBS or MBS) served as the negative controls. To monitor the antibacterial effect over 30 min, the % bacteria killed (shown in the fluorescence image) was calculated as follows:
P. aeruginosa PAO1 (OD600 nm of 10 Units) was prepared as mentioned above. An aliquot of 100 µl of the bacteria was fixed with 5% (v/v) acetic acid for 5 min followed by washing thrice with TBS or MBS containing 0.05% v/v Tween-20 until the pH was stabilised to 7.4 or 6.5 in the respective buffers. The bacteria were then blocked with 1% (w/v) HSA in wash buffers (blocking buffers) under normal or infection-inflammation condition, for 2 h. To assess the binding of the purified proteins to the bacteria, 5 µg of purified CRP or 1.5 µg of purified L-/H- ficolins (diluted in 500 µl of blocking buffers under the two conditions) were incubated with the three-time washed bacteria at room temperature for 2 h. To study the effects of serum on the binding of CRP and/or L-/H- ficolins to the bacteria, after three washes, 5 µg of purified CRP or 1.5 µg of purified L-/H- ficolins (diluted in 500 µl of 10% (v/v) patient or healthy serum in the corresponding MBS or TBS buffers with 0.05% (v/v) Tween-20) were incubated with the bacteria at room temperature for 2 h. After 4 washes, proteins bound to the bacteria were eluted with 0.5 M ammonium formate (pH 6.4) and visualized by SDS PAGE and Western blot analysis.
Sera were diluted to 10% (v/v) in buffers containing 50 mM MES, pH 6.5 or 50 mM Tris, pH 7.4, 145 mM NaCl, 1% (v/v) NP-40 with 0, 2 or 2.5 mM CaCl2. Then 500 µl of the diluted sera were pre-cleared by incubating with 20 µl Protein A Sepharose (GE healthcare, Uppsala, Sweden) at 4°C for 1 h. Polyclonal antibody against human CRP was added to the supernatant and incubated for 1 h at 4°C with shaking. As a negative control, antibody against the horseshoe crab Carcinoscorpius rotundicauda CRP, CrCRP, which does not cross-react with human CRP [22], was added to a replicate supernatant. The inhibition effect of calcium on CRP∶L-ficolin interaction was analyzed by varying the calcium concentration at 0, 2 and 2.5 mM. Then, 20 µl protein A Sepharose was added to the supernatant and incubated for 1 h at 4°C with shaking. The beads were boiled in loading buffer and electrophoresed on 12% SDS-PAGE. The CRP∶L-ficolin complex on the beads was detected by Western blot.
BIAcore 2000 was used to demonstrate real-time biointeraction. PC-HPA chip was prepared by immobilizing 1-palmitoyl-2-oleoyl-phosphatidylcholine (Avanti Polar Lipids, Birmingham, AL) on HPA chip [63] (GE Healthcare). Then 200 nM CRP in running buffer was injected over the surface of PC-HPA followed by separate injections of 50, 100, and 200 µM L-rFBG under normal or infection-inflammation conditions at a flow rate of 30 µl/min. The running buffers for normal condition was 50 mM Tris, 145 mM NaCl and 2.5 mM calcium, pH 7.4, and for infection-inflammation condition, was 50 mM MES, 145 mM NaCl and 2 mM calcium, pH 6.5. The dissociation was for 180 s at the same flow rate. The PC-HPA chip was regenerated by injection of 15 µl of 0.1 M NaOH at 30 µl/min. Injection of binding buffer without CRP was used as the negative control. BIAevaluation 3.2 software was used to calculate the KD. All the surface plasmon resonance curves used in KD calculation were normalized against negative controls.
To achieve CRP-initiated complement activation, 20 µl immobilized PC beads (Pierce, Rockford, IL) was incubated at room temperature for 2 h with 500 µl protein solution in MBS, and TBS containing 10 µg/ml CRP, 3 µg/ml L-ficolin [64], 1 µg/ml MASP-2 [65], following which 1 µg/ml C4 [22] was added and incubated at room temperature for 1 h. For L-ficolin-initiated complement activation, a similar protocol was followed except that GlcNAc-Sepharose and 1 µg/ml C1 complex were used in place of PC-Sepharose and 1 µg/ml MASP-2. The negative controls were beads incubated with incomplete complement components. The beads were boiled in loading buffer and electrophoresed on 12% SDS PAGE. Polyclonal rabbit anti-C4c antibody (1∶1000) and goat anti-rabbit secondary antibody (1∶2000) was used to detect the deposited C4 by immunoblotting.
Red or green fluorosphere beads of 1 µm diameter (Invitrogen) were conjugated with phosphorylcholine-BSA (PC-BSA) (Biosearch Technologies, Novato, CA) or GlcNAc-BSA (Dextra Laboratories, Reading, UK), respectively. The green or red beads conjugated with BSA alone served as reciprocal controls. After a brief sonication, an aliquot of 100 µl fluorosphere beads was incubated with 100 µl of 1 mg/ml PC-BSA, GlcNAc-BSA or BSA alone at room temperature for 15 min. Then, 0.8 mg of 1-ethyl-3-(3-dimethylaminopropyl)-carbodimide hydrochloride, EDC (GE Healthcare) was added to the reaction mixture, vortexed and incubated for 2 h at room temperature. Subsequently, glycine was added to a final concentration of 100 mM and incubated at room temperature for 30 min to stop the reaction. Beads were washed thrice with PBS and resuspended in 100 µl TBS or MBS.
The PC- (red) or GlcNAc- (green) conjugated phagocytic fluorosphere beads were pretreated in the same way as for the PC and GlcNAc beads in the C4 cleavage assay to generate the opsonized particles. U937 cells were treated with 30 ng/ml phorbol myristate acetate for 24 h and the medium was refreshed. The cells were harvested 48 h later and collected by centrifugation, washed twice with PBS and resuspended in TBS and MBS buffers, and enumerated. Then, 1×105 cells were incubated at 37°C for 15 min with 1.5×106 opsonized red PC-beads together with an equal amount of green control beads. Similarly, 1×105 cells were incubated at 37°C for 30 min with 1.5×106 opsonized green GlcNAc-beads together with an equal amount of red control beads. Phagocytosis was stopped by adding 1 ml ice-cold PBS. Cells were collected and washed thrice with PBS and fixed in 4% paraformaldehyde in PBS. Three µl of the cell suspension was examined by fluorescence microscopy (BX 60, Olympus, Tokyo, Japan). For quantification of the phagocytic efficiency, three different fields were chosen. The number of cells and the number of phagocytosed beads were enumerated in each field. As the initial ratio of cells∶beads was 1∶15, the average of 15 beads in a cell was considered 100% phagocytosis. Thus the phagocytic efficiency was calculated as follows:
To test the interaction between L-rFBG and CRP, 0.8 µg of CRP in 100 µl coating buffer (50 mM sodium carbonate/bicarbonate buffer, pH 9.6) was immobilized on 96-well Maxisorp™ plates (NUNC, Roskilde, Denmark) by incubating overnight at 4°C. After three washes with TBST containing 25 mM Tris-HCl, pH 7.4, 145 mM NaCl, 0.05% (v/v) Tween-20, the wells were blocked with 1% (w/v) HSA in TBST (blocking buffer) at 37°C for 2 h. After four washes, 0.8 µg of L-rFBG (with myc fusion tag) in 100 µl blocking buffer was added to the wells and incubated at 37°C for 2 h. Following three washes, the bound L-rFBG was detected with anti-myc antibody (1∶1000) followed by rabbit anti-mouse HRP-conjugated secondary antibody (1∶2000). After adding ABTS substrate (Roche Diagnostics, Mannheim, Germany), the OD405 nm was read. Other ELISA experiments were carried out similarly except for different immobilized proteins and different binding proteins which were detected with the corresponding primary and secondary antibodies. Wells coated with HSA instead of the immobilized proteins served as negative controls unless otherwise stated. It was reported that MBS was commonly used to adjust the acidic condition of the serum [43],[44]. Thus MBS containing 25 mM MES, 145 mM NaCl and 2 or 2.5 mM calcium adjusted to pH: 5.5, 6.0 and 6.5, and TBS containing 25 mM Tris-HCl, 145 mM NaCl and 2 or 2.5 mM calcium, adjusted to pH 7.0 and 7.4 were used as the binding buffers to examine the effect of pH on the protein-protein interaction.
For C1q and L-ficolin competition assay, 0.8 µg CRP in 100 µl coating buffer was immobilized on each well of the Maxisorp™ ELISA plate (NUNC) as mentioned above. After blocking for 2 h in 100 µl blocking buffer (TBS or MBS with 0.05% v/v Tween-20 and 1% w/v HSA), 0.8 µg of C1q with increasing amounts of L-ficolin, or 0.8 µg of L-ficolin with increasing amounts of C1q in 100 µl of blocking buffer was added to each well. Similarly, for MASP-2 and CRP competition assay, 0.8 µg of L-ficolin was immobilized, and 0.8 µg of MASP-2 with increasing amounts of CRP or 0.8 µg of CRP with increasing amounts of MASP-2 in the same binding buffer, were added to each well. The added proteins with constant amount were detected with anti-C1q (1∶1000), anti-L-ficolin (1∶1000), anti-MASP-2 (1∶1000) and anti-CRP (1∶1000) antibodies. Wells coated with HSA instead of CRP were the negative controls.
The aliquoted P. aeruginosa PAO1 without GFP, was prepared similarly to that in the bacterial killing assay. Then the bacteria were fixed in 5% acetic acid for 5 min at room temperature, and washed with either TBS or MBS. The bacteria were blocked with 3% (w/v) HSA in TBS or MBS with 0.05% (v/v) Tween-20 (blocking buffer) at room temperature for 1 h with shaking. Following three washes, the bacteria were incubated at room temperature for 30 min with 10% (v/v) serum, 10% (v/v) ficolin-depleted serum, 10% (v/v) CRP-depleted serum and 10% (v/v) serum depleted of both CRP and ficolin in TBS or MBS buffer to represent normal or infection-inflammation conditions, respectively. After three washes with the corresponding binding buffer, the bacteria were collected and incubated with anti-C3d (1∶100) in blocking buffer for 30 min. After three washes, the bacteria were collected and incubated with R-Phycoerythrin conjugated goat anti-rabbit (1∶200) for 30 min on ice. The bacteria were washed 5 times and fixed with 4% paraformaldehyde for 15 min. After three washes, the bacteria were diluted in PBS (140 mM NaCl, 10 mM phosphate, 2.7 mM KCl, pH of 7.4) which was also the running buffer for the flow cytometry. The bacterial cells with bound C3 signal were collected by Cytomation Cyan LX (Dako A/S) and the counts were analyzed by WinMDI version 2.8. For quantification of the FACS data, the synergistic effect of CRP and ficolin, under infection-inflammation and normal conditions was assessed using (i) the whole serum, (ii) ficolin-depleted serum and (iii) CRP-depleted serum. The relative C3 deposition in each case was calculated by subtracting off the background reading of both CRP- and ficolin- depleted serum.
Data represent means±s.e.m. of three independent experiments with triplicates each. P values of less than 0.05 and 0.01 were respectively considered significant and very significant by Student's t test.
|
10.1371/journal.pgen.1002146 | Pathologic and Phenotypic Alterations in a Mouse Expressing a Connexin47 Missense Mutation That Causes Pelizaeus-Merzbacher–Like Disease in Humans | Gap junction channels are intercellular conduits that allow diffusional exchange of ions, second messengers, and metabolites. Human oligodendrocytes express the gap junction protein connexin47 (Cx47), which is encoded by the GJC2 gene. The autosomal recessive mutation hCx47M283T causes Pelizaeus-Merzbacher–like disease 1 (PMLD1), a progressive leukodystrophy characterized by hypomyelination, retarded motor development, nystagmus, and spasticity. We introduced the human missense mutation into the orthologous position of the mouse Gjc2 gene and inserted the mCx47M282T coding sequence into the mouse genome via homologous recombination in embryonic stem cells. Three-week-old homozygous Cx47M282T mice displayed impaired rotarod performance but unchanged open-field behavior. 10-15-day-old homozygous Cx47M282T and Cx47 null mice revealed a more than 80% reduction in the number of cells participating in glial networks after biocytin injections into oligodendrocytes in sections of corpus callosum. Homozygous expression of mCx47M282T resulted in reduced MBP expression and astrogliosis in the cerebellum of ten-day-old mice which could also be detected in Cx47 null mice of the same age. Three-month-old homozygous Cx47M282T mice exhibited neither altered open-field behavior nor impaired rotarod performance anymore. Adult mCx47M282T expressing mice did not show substantial myelin alterations, but homozygous Cx47M282T mice, additionally deprived of connexin32, which is also expressed in oligodendrocytes, died within six weeks after birth and displayed severe myelin defects accompanied by astrogliosis and activated microglia. These results strongly suggest that PMLD1 is caused by the loss of Cx47 channel function that results in impaired panglial coupling in white matter tissue.
| Oligodendrocytes are the myelinating cells of the central nervous system. Together with astrocytes, oligodendrocytes form networks of coupled glial cells—so-called panglial networks—which are built by gap junctions, i.e. intercellular channels composed of connexin proteins. Pelizaeus-Merzbacher–like disease is an inherited early onset myelin disorder of the central nervous system. Certain mutations of the connexin47 gene, which is expressed by oligodendrocytes, cause this disease. Although the course of the human disease is conspicuous and progressive, connexin47 null mice do not show obvious phenotypic alterations suggesting that the disease may be caused by gain of detrimental function due to the connexin47 mutations. Here we introduced a missense mutation that was found in Pelizaeus-Merzbacher–like disease patients into the connexin47 mouse gene. Expression of the mutant connexin47 gene in oligodendrocytes resulted in myelin malformations in young mice but to a relatively mild extent. From a comparison of connexin47 null and connexin47 mutant mice, we conclude that the human Pelizaeus-Merzbacher–like disease is caused by loss of gap junctional coupling, which results in a decreased number of cells coupled within glial networks, and not by a gain of detrimental function of the mutated protein.
| The autosomal, recessively inherited Pelizaeus-Merzbacher-like disease 1 (PMLD1; MIM: 608804) is an early onset hypomyelinating leukodystrophy caused by mutations in the human connexin47 (Cx47) gene GJC2 (previously called GJA12; MIM: 608803) [1]. Like X-linked Pelizaeus-Merzbacher disease (PMD, MIM: 312080), which is caused by mutations in the gene encoding proteolipid protein 1 (PLP1, MIM: 300401), one of the major proteins in the central nervous system (CNS) myelin, PMLD is characterized by impaired motor development resulting in nystagmus, dysarthria, progressive spasticity and ataxia. First symptoms, nystagmus and poor control of head and trunk movements, occur during early infancy. Twenty-four different mutations including missense, nonsense, partial deletion and frameshift mutations of the GJC2 gene have been reported for PMLD-affected patients to date [1]–[8]. The milder late onset hereditary spastic paraplegia (SPG44, MIM: 613206) is associated with another recessive missense mutation I33M in the GJC2 gene [9].
All PMLD1 patients are homozygous or compound heterozygous for mutations of the GJC2 gene. Neurological symptoms or MRI abnormalities were not detected in heterozygous individuals [1]–[3] but dominantly inherited lymphedemas (MIM: 613480) were recently described to be associated with GJC2 mutations [10].
The GJC2 gene encodes the gap junction protein Cx47. Gap junction channels (GJCs) are intercellular conduits for diffusional exchange of ions and small molecules like metabolites and second messengers. Each of the apposed cells contribute per GJC one connexon (hemichannel) which consists of six connexin proteins. Twenty-one human connexins and 20 rodent connexins were described so far which adds to the great theoretical diversity of gap junction channels, since connexons may be composed of one (homomeric) or more than one (heteromeric) connexin isoform. Coupling of connexons consisting of different connexin isoforms is referred to as heterotypic coupling, in contrast to homotypic coupling resulting from GJCs composed of the same connexin isoform. However, the diversity for GJCs is limited because not all heterotypic channels appear to be functional in cultured cells [11], and different cell types express only few connexin isoforms [12].
In humans, Cx47 expression was detected in CNS and peripheral nervous system (PNS) [1], whereas mice express Cx47 (Gjc2; MGI: 2153060) only in oligodendrocytes contributing to gap junctional communication within panglial networks [13]–[15]. Mature oligodendrocytes in the CNS express Cx32 (Gjb1; MGI: 95719) and Cx29 (Gjc3; MGI: 2153041) in addition to Cx47 [16]. Astrocytes express Cx30 (Gjb6; MGI: 107588) and Cx43 (Gja1; MGI: 95713), while Cx26 (Gjb2; MGI: 95720) is expressed in some gray matter astrocytes [17]–[19]. Besides homotypic Cx47/Cx47 channels, heterotypic Cx47/Cx43 and Cx47/Cx30 channels showed functional coupling in cell culture experiments [11], [20].
Fracture replica immunogold labeling and immunohistological analyses suggested heterotypic channels between oligodendrocytes and astrocytes (O∶A) forming panglial networks [21], [22] and recent dye transfer experiments on oligodendrocytes in the corpus callosum confirmed O∶A intercellular communication. In addition, these experiments revealed functional interoligodendrocytic (O∶O) coupling. Both, O∶O and O∶A coupling were significantly affected in Cx47 deficient mice suggesting a crucial role for Cx47 gap junctions in panglial networks [23]. Accordingly, Cx47 deficient mice showed myelin vacuoles but behavioral or further anatomical abnormalities have not been observed as yet [13]. Mutations in the GJB1 (MIM: 304040) gene coding for hCx32 protein result in the demyelinating peripheral neuropathy X-linked Charcot-Marie-Tooth disease (CMTX, MIM: 302800). Cx32 deficient mice showed only mild, late onset myelination deficits in the CNS [24] but additional deletion of Cx32 in Cx47 deficient mice results in severe myelin abnormalities and premature death. This indicates some functional redundance of these connexins expressed by oligodendrocytes.
Results from transfected cell cultures expressing mutant Cx47 proteins suggested that PMLD in humans is induced by either loss of Cx47 GJC function or hemichannel dysfunction [8], [25].
Since Cx47 null mice show only mild myelin abnormalities, in contrast to human PMLD patients, we generated a mouse model for PMLD by knock-in of the point mutation mCx47M282T into the endogenous Cx47 gene locus. The orthologous human missense mutation hCx47M283T was found homozygously in PMLD patients [1].
The Cx47M282T expressing mice enabled us to investigate whether loss of Cx47 function regarding O∶O and O∶A coupling, a transdominant effect on Cx32 or gain of detrimental function by intracellular effects of the mutant Cx47 contribute to the disease.
Young homozygous Cx47M282T expressing mice showed impaired motor coordination and balancing performance on the accelerating rotarod, decreased punctate Cx47 immunostaining and similar to Cx47 deficient mice, diminished O∶O coupling in the corpus callosum, compared to wildtype mice. Furthermore, like Cx47 deficient mice, homozygous Cx47M282T mice revealed retarded myelin formation during the first weeks after birth.
Similar to Cx47/Cx32 null mice, homozygous Cx47M282T mice deprived of Cx32 showed severe myelin abnormalities and died within the first four months after birth, demonstrating an essential function and partial redundancy of both connexin isoforms for maintenance of mouse myelin.
Our results indicate that loss of Cx47 function in GJC leads to myelin pathology in juvenile mice which can to a large extent be compensated during subsequent development to adulthood.
The mCx47M282T mutation was generated by PCR mutagenesis using purified genomic DNA obtained from a C57BL/6 mouse as template. Codon 282 was mutated to cause a methionine residue to threonine transition by T to C exchange at nucleotide 845 of the mCx47 coding region which in addition resulted in a recognition site (5′-(N)5GAGACG-3′) of the restriction endonuclease BsmBI. We replaced the mouse Cx47 coding region by the mutated mCx47M282T one, an internal ribosome entry site (IRES) followed by the Escherichia coli β-galactosidase (β-gal) coding DNA (LacZ), preceded by a nuclear localization sequence and a frt-site-flanked neomycin selection cassette by homologous recombination in HM-1 embryonic stem cells [26] (Figure 1A). Twenty-four of 377 neomycin resistant clones were positively characterized for homologous recombination by two individual PCRs. Homologous recombination was further verified by Southern blot analysis. Three of the targeted ES cell clones were microinjected into C57BL/6 blastocysts. Two clones yielded germline transmission chimeric mice that produced heterozygous Cx47+/M282Tneo offspring after backcrossing with C57BL/6 mice. The frt-site-flanked neomycin resistance cassette was deleted by mating of Cx47+/M282Tneo mice to hACTB:FLPe mice [27] which yielded the offspring Cx47+/M282T deprived of the neomycin selection cassette (Figure 1B). All mice further investigated harboured at least 96% of C57BL/6 genetic background. Mice carrying the mutant mCx47M282T allele heterozygously (Cx47+/M282T and Cx47+/M282Tneo) were interbred to obtain Cx47+/+, Cx47+/M282T and Cx47M282T/M282T or Cx47+/+, Cx47+/M282Tneo and Cx47M282T/M282Tneo offspring, respectively. Genotypes were demonstrated by PCR (Figure 1C and E) and Southern Blot (Figure 1F) analysis. Genomic presence of the mutated mCx47M282T was analyzed by PCR and subsequent BsmBI digestion (Figure 1D). Female as well as male Cx47+/M282T and Cx47M282T/M282T mice were fertile.
Expression of the mutant mCx47M282T gene in transgenic animals resulted in a bicistronic mRNA coding for mCx47M282T and the reporter LacZ DNA. The β-gal encoded by LacZ DNA featured a nuclear localization signal and was detected by X-Gal and antibody staining. Translation of the LacZ mRNA was mediated by an IRES cassette. To further characterize LacZ expression, immunohistochemical analyses of cell type specific markers were combined with X-Gal staining. The reporter LacZ DNA was expressed in Cx47-positive cells (Figure 2E) and was coexpressed with the oligodendrocytic marker protein 2′,3′-cyclic nucleotide 3′ phosphodiesterase (CNPase) (Figure 2A). X-Gal staining was not colocalized with immunohistochemical signals of the neuronal marker (NeuN) or the astrocytic marker (GFAP) (Figure 2B and 2C). Immunofluorescence analyses showed a marked decrease of Cx47-positive puncta in Cx47M282T/M282T and, although less pronounced, in Cx47+/M282T brain slices compared to wildtype. Furthermore, Cx47 immunostaining revealed that the mCx47M282T protein does not exhibit the typical gap junctional immunosignals at oligodendrocytic somata [28]. β-Gal signals were obvious in Cx47+/M282T and Cx47M282T/M282T but absent in wildtype brain tissue (Figure 2D–2F). Immunoblot analysis of β-gal in brain lysates obtained from P40 mice yielded distinct signals in Cx47+/M282T and Cx47M282T/M282T around 120 kDa but not in corresponding wildtype controls (Figure 2G). Immunoprecipitation (IP) and subsequent immunoblot analysis against Cx47 revealed bands on wildtype and Cx47M282T/M282T brain lysates around 50 kDa. No signals were found in Cx47 deficient (Cx47−/−) control lysates after IP but immunoblot analysis on crude Cx47−/− whole brain lysates yielded false positive signals for all genotypes at approximately 50 kDa. To minimize precipitation and subsequent detection of cross reacting proteins, IPs and immunoblots were conducted using two distinct Cx47 antibodies (Zymed No. 37-4500 and Zymed No. 36-4700). HeLa cell control lysates yielded signals for the transfected Cx47-eGFP fusion protein around 75 kDa with crude lysates and after IP indicating accuracy of the procedure. Signals around 40 kDa in this lane are probably caused by the cleaved Cx47 C-terminus, since antibodies used were directed against peptides derived from the C-terminal region of Cx47.
An increase of β-gal positive cells in CNS white and gray matter of Cx47M282T/M282T mice (Figure 3) and Cx47+/M282T (not shown) was observed during postnatal development (P7–P105). At P7 β-gal positive cells were detected in the cerebellar white matter (Figure 3A), less in corpus callosum and in fimbria of hippocampus and only few isolated cells in the cerebral cortex (Figure 3E). At P10 the number of β-gal positive cells abundantly increased in the corpus callosum and the hippocampal fimbria. Unlike P7 mice, P10 animals displayed scattered β-gal positive cells localized in layer VI of the cerebral cortex (Figure 3F), whereas few β-gal positive cells were found in the cerebellar granular layer (Figure 3B). At P16 β-gal positive nuclei increased more than twofold in the cerebellar white matter and granular layer compared to P10 mice (Figure 3B and C). At this developmental stage, we also observed β-gal positive cells in the hippocampus, predominantly localized at the hippocampal fissure and alveus, and in layers VI-IV of the cerebral cortex, while the number of LacZ expressing cells in white matter tracts was further increased (Figure 3G). After 3.5 months of postnatal development (P105) the number of β-gal positive cells was increased in white matter and adjacent gray matter compared to P16 mice (Figure 3D). X-Gal stained nuclei were furthermore present in all layers of the cerebral cortex, decreasing in number from layer VI to layer I. Isolated β-gal positive cells were found widespread over the hippocampal gray matter (Figure 3H). In cerebellar and cerebral white matter of P105 mice LacZ expressing cells were often organized in a chain-like structure parallel to the direction of the fibers, as typically observed for oligodendrocytes in white matter (Figure 3D and 3H).
Sixteen day-old Cx47M282T/M282T mice featured a neuropathologic phenotype affecting mainly the white matter in some regions of the CNS (Figure 4). Within the cerebellar white matter, cystic spaces of about 100 µm diameter were found, filled with cellular debris, sometimes featuring an onion shell – like lamellar pattern. Within the cerebellar gray matter, scattered groups of mainly Purkinje neurons could be seen at different stages of degeneration.
A similar cystic degeneration of CNS white matter was observed in the prechiasmatic optic fascicle featuring about 2–3 cysts of about 100 µm diameter per cross section.
Interestingly, while the effects in cerebellum and optic fascicle were rather constant between different experimental animals, the effects of mCx47M282T expression in the remainder of the telencephalon were much more variable. Most notably, the main white matter tracts like fimbria hippocampi, corpus callosum or anterior commissure did not show any regularly occurring cystic degeneration. Within the gray matter, we inconstantly observed groups of degenerating cells in the hippocampal field CA1 and in the subgranular zone of the dentate gyrus (Figure S1A and S1B).
Next to the neuronal damage, we inconstantly observed cell debris and degenerating small cells at the outer aspect of numerous brain capillaries, indicative of a disseminated, and again highly variable destruction of perivascular astroglia (Figure S1C–S1E).
In heterozygous mice, the morphology of the cerebellum was essentially normal (Figure 4B, E): likewise, lesions within the optic fascicle were almost absent except for a small number (about 2 per cross section of one fascicle) of individual cells being swollen and degenerated (Figure 4I).
We investigated general activity, motor coordination and motor learning performance in juvenile (23 days old) and adult (three months old) mCx47M282T mice. Since PMLD is caused by an autosomal recessive mutation we also investigated whether heterozygous mice would exhibit an intermediate phenotype ranging between homozygous and wild-type mice. For this purpose pre-planned pair-wise comparisons between wild-type mice and homozygous as well as between wild-type mice and heterozygous mice were performed.
To address the question whether homozygous and heterozygous expression of mCx47M282T mutation affects coupling, dye transfer experiments were performed in the corpus callosum of acute coronal slices obtained from P10–15 mice. In each slice, single oligodendrocytes were dialyzed by whole cell patch-clamp with the gap junction permeable tracer biocytin, as previously described [23]. Membrane currents of the injected oligodendrocyte were recorded in response to a series of hyperpolarizing and depolarizing voltage steps (10 mV increment) from −170 mV to +50 mV for 50 ms (Figure 6A1, 6B1). The cohort of wildtype and Cx47−/− control mice was the same as previously published, since the experiments reported here were done in parallel with the former study [23].
We distinguished between oligodendrocytes participating in glial networks and uncoupled oligodendrocytes by biocytin injection into oligodendrocytes and subsequent streptavidin-Cy3 labeling (Figure 6C). The number of coupled cells and extent of tracer spread (µm) were determined for the networks identified (Figure 6D, 6E) and glial cells whithin these networks were further characterized by marker protein costainings (Figure 7). For a detailed numeric representation see Table 1.
In Cx47M282T/M282T mice, 67% of the single oligodendrocytes dialyzed with biocytin formed networks (10 out of 15 slices from 3 mice). In these ten networks the median value of biocytin-positive (coupled) cells was 6 (4–13; values indicate 25th–75th percentiles, respectively), with a significant reduction by 88% in the number of coupled cells per network as compared to wildtype mice (Kruskal-Wallis, p = 0.001, Mann-Whitney U-test, p<0.001). Measuring the largest distance between two cells within a network indicated that the mean extent of tracer spread for Cx47M282T/M282T mice compared to wildtype animals was not significantly different (Figure 6C–6E).
To determine whether homozygous mCx47M282T expression affected coupling of oligodendrocytes to neighboring oligodendrocytes and astrocytes, biocytin/streptavidin-Cy3 labelling was combined with immunostaining for the oligodendrocytic marker CNPase and the astrocytic marker GFAP. In Cx47M282T/M282T mice the majority of biocytin-positive cells coupled within a given network were CNPase-positive (100%; 86%–100%), with a significant increase in comparison to wildtype (see table 1; Kruskal-Wallis, p<0.01, Mann-Whitney U-test, p<0.01) (Figure 7A and 7C). GFAP-positive cells colabeled with biocytin/streptavidin-Cy3 (0%; 0%–6%) were found only in 3 out of ten networks (Figure 7A and C). However, the observed reduction was not significant compared to wildtype mice (see Table 1). Within a network analyzed we detected a population of CNPase- and GFAP-negative cells, which was significantly reduced to a median value of 0% (0%–7%) compared to wildtype animals (see table 1; Kruskal-Wallis p<0.05, Mann-Whitney U-test, p<0.01) (Figure 7A and 7C).
In heterozygous Cx47+/M282T mice (Figure 6B1–6B3) only 4 out of 15 biocytin injections into single oligodendrocytes revealed intercellular coupling (27% formed networks, 4 mice), with networks consisting of a median value of 36 cells (6–82) and tracer spreading up to 193 µm (average 131±24 µm) (Figure 6). The observed decrease in the number of oligodendrocytes forming networks was significantly different in comparison to wildtype mice (see table 1; Chi-Square Test, Chi Square = 12.523, p<0.001) [23]. However, the number of coupled cells within a given panglial network and the extent of biocytin spread were not significantly affected as compared to wildtype animals [23]. Similar to the networks detected in brain slices obtained from wildtype mice, the majority of biocytin-positive cells were CNPase-positive, 3% (0%–7%) coupled cells were CNPase- and GFAP-negative, while the remaining 11% (1%–18%) were GFAP-positive astrocytes (Figure 7B and 7C).
Thus, homozygous mCx47M282T expression resulted in a decrease in the number of coupled cells within the network of cells coupled to the injected oligodendrocyte. No significant difference in number of coupled cells between Cx47M282T/M282T and Cx47−/− mice was observed. Heterotypic coupling of oligodendrocytes to GFAP-positive astrocytes was impaired but not completely abolished and coupling to the population of CNPase- and GFAP-negative cells was significantly affected. Heterozygous mCx47M282T expression caused a decrease by 68% in the number of oligodendrocytes forming networks in comparison to wildtype animals, but did not impair the number of coupled cells within a given network. However, in Cx47+/M282T mice neither the number of coupled cells within a network nor coupling of oligodendrocytes to astrocytes and to the subpopulation of CNPase- and GFAP-negative cells was significantly affected.
In cerebellar slices obtained from P10 Cx47M282T/M282T and Cx47−/− mice vacuoles were occasionally present in the white matter tract and MBP stainings appeared inhomogeneous compared to wildtype and cerebella of Cx47+/M282T mice. Furthermore, Cx47M282T/M282T and Cx47−/− mice showed fewer fine fibers in the granular layer than wildtype littermates and these fibers did not reach the Purkinje cell layer in Cx47+/M282T, Cx47M282T/M282T and Cx47−/− as observed in Cx47+/+ slices (Figure 8A, 8D, 8G and 8J). Expression of the mCx47M282T protein resulted in conspicuous astrogliosis in the cerebellar white matter of Cx47M282T/M282T and Cx47−/− animals as indicated by up-regulation of GFAP expression, while Cx47+/M282T mice displayed only a mild increase in GFAP levels within white matter tracts (arrows). Furthermore, fewer GFAP-positive cells were visible in the cerebellar granular layer of Cx47M282T/M282T mice (Figure 8B, 8E, 8H and 8K). Elevated ionized calcium-binding adapter molecule 1 (Iba1) expression indicated accumulation/activation of microglial cells in the cerebellar white matter of Cx47+/M282T, Cx47M282T/M282T and Cx47−/− animals (Figure 8C, F, I and L). However, astrogliosis and activation of microglial cells could not be detected any longer in P90 Cx47+/M282T and Cx47M282T/M282T mice (Figure 9E, H, F and I). In the cerebellar granular layer of 3-month-old Cx47M282T/M282T mice, GFAP-positive cells were still diminished compared to corresponding Cx47+/M282T and Cx47+/+ mice (Figure 9B, F and H). MBP staining on cerebella slices of adult Cx47M282T/M282T mice revealed uneven signals in the white matter compared to wildtype littermates. Vacuoles could not be detected anymore but the fine myelin fibers pervading the granular layer were almost absent in cerebella of Cx47+/M282T and Cx47M282T/M282T mice (Figure 9A, 9D and 9G).
To ascertain the impact of mCx47M282T on oligodendrocyte differentiation, we examined the expression of the early oligodendrocyte marker 2′,3′-cyclic nucleotide 3′ phosphodiesterase (CNPase) and the marker for mature myelinating oligodendrocytes myelin basic protein (MBP) [29] at five stages (P10–P90) of postnatal development (Figure 10).
One way ANOVA of quantitative immunoblot analyses of cerebellar lysates revealed a significant reduction of CNPase in juvenile and young adult Cx47M282T/M282T mice (P10: 73±28%, n = 5, p<0.001; P16: 76±42%, n = 5, p<0.05 and P40: 84±17%, n = 5, p<0.05), while Cx47+/M282T mice displayed a significant reduction of CNPase levels to 67±45%, n = 5, p<0.05 only at P16, compared to wildtype littermates (Figure 10A, B and E). Analyses on lysates obtaind from P28 mice revealed levels of CNPase for both tansgenic genotypes which were not significantly altered compared to wildtype littermates (Cx47+/M282T: 119±39%, n = 3, p = 0.26 and Cx47M282T/M282T: 115±37%, n = 3, p = 0.2). Immunoblot analyses on cerebellar lysates obtained from P90 Cx47+/M282T and Cx47M282T/M282T mice revealed significantly elevated expression of CNPase with a mean level of 181±35%, n = 5, p<0.001, for Cx47+/M282T mice and 160±49%, n = 5, p<0.001, for Cx47M282T/M282T mice in comparison to wildtype animals (100%).
Even more pronounced, expression of MBP was significantly diminished in both Cx47+/M282T and Cx47M282T/M282T mice on days P10, P16, P28 and P40 (Figure 10C–E), as compared to wildtype littermates. This reduction was particularly obvious in P10 Cx47M282T/M282T mice (10±10%, n = 5, p<0.001) and in P10 Cx47+/M282T mice (13±10%, n = 5, p<0.001). Sixteen day-old heterozygous transgenic mice showed less severe but significant MBP reduction to 75±11% (n = 5, p<0.001) and homozygous mCx47M282T expressing mice still had intensly reduced MBP levels to 39±17% (n = 5, p<0.001). Similar results were observed for cerebella of 28-day-old mice with a mean reduction of MBP expression to 61±7% (n = 3, p<0.05) in Cx47+/M282T mice and to 47±22% (n = 3, p<0.001) in cerebella obtained from Cx47M282T/M282T mice. Although the MBP levels were still significantly reduced, the expression almost reached wildtype levels in lysates obtained from 40-day-old Cx47+/M282T mice (92±6%, n = 5, p<0.05) and Cx47M282T/M282T mice (84±21%, n = 5, p<0.05). After three months of postnatal development (P90) MBP expression reached wildtype levels in cerebella of Cx47M282T/M282T animals (97±39%, n = 5, p = 0.83), while beeing significantly elevated in Cx47+/M282T mice at the same developmental stage (145±37%, n = 5, p<0.001).
These data illustrate reduced formation of myelin during adolescence in cerebella of mCx47M282T expressing mice, which is compensated during adulthood.
Mice deficient for both Cx47 and Cx32 (Cx47−/−/Cx32−/−) display severe myelin malformations and die on average on day 51 [13]. In order to analyze whether mCx47M282T expression is sufficient for proper myelin maintenance or leads to malformations in accordance with Cx47−/−/Cx32−/− mice, we crossbred Cx47M282T/M282T male and Cx47+/M282T/Cx32+/− female mice to obtain Cx47M282T/M282T/Cx32Y/− male mice. These animals were Cx32-deprived and expressed the mCx47M282T mutation homozygously. Like Cx47−/−/Cx32−/− mice, adult Cx47M282T/M282TCx32Y/− displayed severe CNS myelin defects. Immunoblot analyses of cerebellar lysates obtained from P90 Cx47M282T/M282TCx32Y/− mice revealed a significant MBP decrease to residual 11.7±7.3% expression compared to Cx47+/+ mice, while CNPase expression was not significantly altered (Figure 10A, 10C and 10E). Furthermore, histochemical analysis yielded only faint MBP signals while increased GFAP and Iba1 expression was obvious in the cerebellar white matter of these mice (Figure 9M, 9N, and 9O). Only three out of eight animals reached three months of age but displayed severe motor impairment. Death of Cx47M282T/M282TCx32Y/− mice began at the age of 42 days. All dying animals displayed pronounced ataxia one to two days before death. Cx47+/M282TCx32Y/− littermates did not die and displayed staining patterns similar to Cx47+/M282T mice (Figure 9J, 9K, and 9L). These data further illustrate that the loss of Cx47 function in vivo is caused by the M282T missense mutation.
Here we have analyzed phenotypic abnormalities of a new transgenic mouse line, carrying a - loss of function - point mutation which leads to human inherited PMLD1. We decided to introduce only the M282T missense mutation into the mouse Cx47 gene, in order to avoid side effects of other amino acid differences between mouse and human Cx47 proteins.
Expression of the mCx47M282T missense mutation, whose orthologous hCx47M283T counterpart was homozygously found in patients, results in a complex but variable neuropathologic phenotype in juvenile, homozygous Cx47M282T mice.
Cx47 deficiency as well as homozygous Cx47M282T expression leads to similar phenotypes in mice, suggesting that the point mutation of Cx47 results in loss of function proteins in vivo. Accordingly, biocytin injections into corpus callosum oligodendrocytes participating in glial networks revealed a 88% decrease in number of coupled cells in Cx47M282T/M282T P10–P15 mice which was very similar to the 85% reduction observed with Cx47−/− mice. Furthermore, like Cx47 deficiency, homozygous Cx47M282T expression caused a significant reduction of the CNPase- and GFAP-negative population of coupled cells within panglial networks. We have previously characterized the majority of this cell population as oligodendrocyte precursors by expression of Olig2, a marker for both immature and mature oligodendrocytes and by activity of the NG2 promoter [23].
Immunofluorescence analyses revealed only few Cx47 positive puncta in brain slices of Cx47M282T/M282T mice, although the ß-galactosidase reporter protein, which is encoded by the same bicystronic mRNA was robustly expressed. Loss of mutant protein expression due to the bicistronic mRNA construct carrying IRES LacZ, in addition to the Cx47M282T coding sequence was ruled out by immunoprecipitation and subsequent detection via immunoblot. Furthermore, the IRES was introduced 87 nucleotides downstream of the first and 79 nucleotides upstream the second cistron to achieve optimal expression of both coding sequences and similar constructs did not result in loss of expression in recently published mouse lines [30]–[32].
Cell culture experiments revealed that transfectants expressing different Cx47 missense PMLD1 mutations showed accumulation in the endoplasmatic reticulum (ER) of most mutant connexins. This resulted in loss of punctate immunostaining in the plasma membrane and loss of channel function [8], [25]. In addition to ER retention, Cx47M283T connexins were occasionally located at the plasma membrane in HeLa cells [25]. We obtained similar results by expression of the orthologous mCx47M282T protein in HeLa cells (data not shown). However, we could not identify intracellular accumulations of the mutant protein in vivo and immunosignals in the plasma membrane were very rare in homozygous mice. This could be caused by attenuated signals due to trafficking abnormalities and dispersal of Cx47 connexons or by rapid degradation of the mutant protein, possibly ER associated [33].
Thus, loss of Cx47 function may be due to degradation, rapid internalization and closure of gap junction channels. As a first step to clarify this issue, quantitative Cx47 immunoblots are needed but all currently available antibodies yield prominent unspecific signals in Cx47 deficient brain lysates [34].
Expression of connexins was shown to be critical for cell migration in the CNS [35] but homozygous expression of the mutant mCx47M282T did not alter the typical distribution pattern of oligodendrocytes during brain development. On the other hand, ten-day-old homozygous mCx47M282T mice displayed a significant reduction of CNPase expression by 28% and more diminished MBP levels (by 90%), whereas 90-day-old mCx47M282T expressing mice showed increased levels of CNPase and almost normal levels of MBP. CNP expression appears early during oligodendrocyte differentiation compared to MBP which indicates mature oligodendrocytes in vivo [36], [29]. Thus, our data support the notion that Cx47 function in developing oligodendrocytes is needed for regular oligodendrocyte development and proper timing of myelination rather than for oligodendrocyte migration.
Since outright degenerations of white matter as described in Figure 4 are comparatively rare and appear to follow a highly focal distribution, it is unlikely that they contribute to any major degree to the functional alterations seen in these mice. Rather, they may be seen as reflecting local transport imbalances surpassing a critical lethal threshold, while the functional deficits are caused by more generalized transport deficiencies. This view is in line with data from heterozygous mice, in which these degenerative effects are largely absent despite their functional impairments.
Gap junctional conduits couple apposed cells and may form reflexive channels that connect myelin sheaths in oligodendrocytes and Schwann cells [22], [37]. Panglial networks built by astrocytic and oligodendrocytic gap junctions are considerd to play a key role in spatial K+ siphoning after neuronal activity [38]–[40]. Furthermore, gap junctional communication may also be needed to provide oligodendrocytes with metabolites in particular during differentiation and myelin formation. Insufficient supply of metabolites such as glucose, ATP, and possibly N-acetylaspartate (NAA, 173 Da molecular mass) due to loss of Cx47 channel function is likely to play an important role for the hypomyelinating phenotype of Cx47M282T expressing mice and may be the cause of PMLD1 in humans. Increased levels of NAA represent a typical feature of Canavan Disease, a fatal dysmyelinating disorder caused by mutations of the oligodendrocytic enzyme aspartoacylase (ASPA) [41]. Most children with Canavan disease die in the first decade of life, while mice with loss of Aspa function are able to survive beyond 12 months but display progressive vacuolation in myelin starting around P14 [42], [43]. PMLD patients showed normal to slightly elevated NAA levels but the N-acetylaspartylglutamate (NAAG) level is increased in the cerebral spinal fluid of PMLD patients harbouring Cx47 mutations [2], [44], [45]. NAAG is released by neurons under depolarizing conditions and degraded by the astrocytic enzyme glutamate carboxypeptidase II (GCP II) to glutamate and NAA [46], [47]. Intraastrocytic NAA may be transported to oligodendrocytes via gap junction channels where it is cleaved by ASPA to aspartate and acetate that may be used subsequently for myelin lipid synthesis [48]. Investigations on lipid composition, NAA, NAAG and aspartate levels in myelin and CSF of Cx47−/−, Cx47M282T/M282T and Cx47M282T/M282T/Cx32−/− mice should clarify whether these metabolites play a major role in PMLD1.
Besides loss of channel function, loss of intracellular protein interaction may also be relevant for the phenotype of Cx47M282T expressing mice and for PMLD. Cx47 was shown to interact with the tight junction adaptor protein zonula occludens 1 (ZO-1), the Y-box transcription factor zonula occludens 1 associated nucleic acid binding protein (ZONAB) which regulates expression of the proliferating cell nuclear antigen (PCNA), cyclin D1 and ErbB-2. [49]–[52]. Interaction of ZONAB and ZO-1 is associated with the transition from a proliferative to a differentiated state of epithelial cells [53], [54]. In Cx47 deficient mice punctate staining for ZONAB is lost from oligodendrocyte cell somata while ZO-1 signals remain [50]. Analyses on interactions of these proteins with the mutant Cx47M282T using this mouse model will help to gain further insight into intracellular mechanisms which might be influenced by the disease causing mutation.
Given that myelination in mutant mice was largely coequal to wildtype mice six weeks after birth, other oligodendrocytic connexins like Cx29 or Cx32 can probably account for compensatory effects in mice.
In normaly developed myelin Cx29 is not localized at oligodendrocyte somata but mainly at the periaxonal space and was recently shown to be incapable of forming functional intercellular gap junction channels [16], [24], [49], [55]. This indicates a distinct role of Cx29 in myelin function and makes it an improbable candidate for functional redundancy of Cx47 channels.
Mice deficient for both Cx47 and Cx32 (Cx47−/−/Cx32−/−) showed severe myelin abnormalities and premature death [13], [56] suggesting functional redundancy of both connexins. This led to the speculation that mutations of Cx47 might lead to a detrimental gain of function and inhibition of Cx32 gap junctional channels [1].
Although homozygous Cx47M282T expression results in diminished or total loss of Cx47 gap junction channel function, it does, like Cx47 deficiency, not lead to a complete inhibition of interoligodendrocytic coupling as recently described for Cx47−/−/Cx32−/− mice [23]. Furthermore, death of Cx47M282T/M282T/Cx32−/− animals began 42 days after birth and 90-day-old Cx47M282T/M282T/Cx32−/− mice displayed severe motor impairment, myelin malformations, vacuolation, pronounced astrogliosis and increased Iba1 expression, whereas P90 Cx47M282T/M282T mice showed minor affected myelin in the cerebellum. These data further corroborate that the M282T missense mutation results in loss of Cx47 channel function and underline the common role of Cx32 and Cx47 channels for myelin maintenance. In addition, these data show that Cx47M282T does not have a transdominant negative effect on Cx32 function in vivo.
Detailed immunofluorescence analyses have revealed a drafmatic increase of Cx32 levels along myelinated fibers in the CNS of Cx47−/− mice [50]. It seems probable that increased Cx32 expression or altered Cx32 channel localization can account for loss of Cx47 function compensation in Cx47M282T/M282T and Cx47 null mice but not in human PMLD patients.
Although PMLD1 is a recessive disease, heterozygous mCx47M282T expressing mice revealed an unexpected CNS phenotype. MBP expression was decreased, accompanied by mild astrogliosis in cerebella of ten-day-old mice. Furthermore, the number of oligodendrocytes forming networks was significantly reduced in heterozygous mice which was not detected in Cx47M282T/M282T or Cx47 null mice. Compared to wildtype littermates, adult Cx47+/M282T but not Cx47M282T/M282T mice showed significantly decreased locomotion, running speed and reduced rearing behavior in the open field test, but no significant alterations in the rotarod assay suggesting impaired neuronal function in motor circuits, independent of a cerebellar phenotype. Immunofluorescence analyses revealed a reduction of Cx47 positive puncta in myelin of Cx47+/M282T mice and the number of oligodendrocytes forming intercellular networks was also reduced in these mice. This reduction may have caused the mild myelin abnormalities and alterations in intercellular coupling found in young heterozygous mice. In contrast to Cx47M282T/M28T mice, high-resolution light microscopy based upon 1 µm thin sections did not reveal lesions in the cerebellum and at best very minor alterations in the optic fascicle. These findings indicate that the functional alterations seen in heterozygous mice are unrelated to myelin degeneration.
Furthermore, Cx47+/M282T/Cx32−/− mice did not die during adolescence and did not show the strong phenotype observed with Cx47M282T/M282T/Cx32−/− and Cx47−/−/Cx32−/− mice. This indicates that residual expression of wildtype Cx47 protein is sufficient to prevent early death in Cx47+/M282T/Cx32−/− mice, in accordance with recessiveness of human PMLD1.
The cause of the phenotypic abnormalities found in heterozygous mice is still enigmatic, but they suggest a gain of detrimental function by heterozygous Cx47M282T expression, possibly due to inhibited or disturbed adapter protein interaction or hemichannel dysfunction [8]. Furthermore, the phenotypes observed with heterozygous and homozygous animals were not completely congruous and some abnormalities aggravated with heterozygous expression. This suggests that compensatory effects in homozygous mice are lacking in heterozygous mice, at least in distinct cells.
The course of PMLD1 is progressive in humans in contrast to myelin pathology in Cx47M282T/M282T and Cx47−/− mice. One major difference between mouse and human CNS is the higher portion of white matter in the human brain. While in humans more than 50% of the volume consists of white matter, in the mouse it is only around 10%. The increased number of oligodendrocytes and their relatively extended distance to blood vessels may be one reason for the higher sensibility of the human brain regarding loss of Cx47 function. Furthermore, previous studies have demonstrated the strong ability of the mouse CNS to efficiently compensate for loss of myelin proteins e.g. PLP1 and Nogo-A, MAG or complete elimination of oligodendrocytes during the first weeks of postnatal life [57]–[59]. Cx47M282T/M282T and Cx47−/− mice apparently compensate for loss of Cx47 function within the critical time interval before axonal loss occurs [60].
Altogether, our results support the notion that loss of Cx47 channel function causes PMLD1. Furthermore, we consider young homozygous Cx47M282T mice, although displaying only transient myelin malformation, as a suitable mouse model to gain closer insights in the mechanisms that lead to inherited human PMLD1.
All mice used in this study were kept under standard housing conditions with a 12 h/12 h dark–light cycle and with food and water ad libitum. All experiments were carried out in accordance with local and state regulations for research with animals. All animals investigated harboured at least 96% C57BL/6 genetic background.
The hCx47M283T mutation from a human patient [1] was inserted by PCR mutagenesis in the orthologous mouse gene resulting in the mCx47M282T mutation, cloned into the pBluescript vector and sequenced in both directions by LGC Genomics (Berlin, Germany). The mutated gene was cloned into a vector containing the IRES and the nls-LacZ cassettes by BclI/HindIII and NsiI digestions and subsequent ligation of blunted ends. The IRES sequence (Clontech Laboratories, CA, USA) was cloned by SmaI digestion from IRES_pBluescript [30], into pBS_nls_LacZpA [61] digested with HindIII and subsequent blunt end generation by Klenow polymerase fill in.
The 17.6 kb targeting vector included two homologous regions with a 4667 bp 5′ intron fragment upstream and a 2479 bp 3′ fragment downstream the Cx47 coding DNA. The neomycin selection cassette including the phospoglycerate kinase promoter was flanked by frt sites and cloned into the 3′ homologous region downstream of the Cx47 polyA signal by SalI/XhoI ligation.
The final vector was partially sequenced by LGC Genomics (Berlin, Germany) after restriction analysis. The functionality of the frt sites was verified by transformation into Flp recombinase expressing E. coli bacteria [62]. The functionality of the mCx47M282T-IRES-nls_LacZ construct was tested in HeLa cells (not shown).
ES cell culture and transfection were performed as described previously [63]. The targeting vector DNA (200 mg) was linearized by SalI digestion prior to transfection. Neomycin resistant ES cell clones were screened for homologous recombination by PCR using an external primer specific to the Cx47 3′ region (Cx47-3-rev: CCA GGA TTC ATG TGA AGG AGA AGG G) and an internal primer specific to the neomycin resistance cassette (primer C1: CTC TGA GCC CAG AAA GCG AAG GAG). To exclude clones in which recombination had occurred partially, a second PCR analysis was performed with cells positive in the first one. Here, we used primers annealing in the coding region of Cx47 (primer B1: GAG GAG CGA GCG GAG GAT GTG GCT G and primer B2: GTG GCG CTG CCG GTT CCG GAA GCT AG) which flanked the BsmBI restriction site that was introduced by the mutation of the Cx47 gene. BsmBI digestion of the 700 bp PCR fragments yieded additional 350 bp fragments which indicated integration of DNA coding for the mutant mCx47M282T. Homologous recombined ES cell clones were subsequently confirmed by Southern blot hybridization. Therefore, DNA from the PCR-positive clones was digested with BamHI or EcoRV and DNA fragments were separated via agarose gel electrophoresis. After blotting onto Hybond N+ membranes (Amersham Biosciences, Buck, UK) and ultraviolet crosslinking for fixation, hybridization was performed under stringent conditions using the Quick Hyb solution (Stratagene, La Jolla, CA, USA) at 68°C for 1.5 h. A 454 bp BglII fragment was used as a 5′ external probe and hybridized with BamHI digested DNA fragments. EcoRV digested DNA was hybridized with an 824 bp 3′ external probe generated by BamHI/XmnI digestion and with a 624 bp HpaI Fragment obtained from the LacZ coding region serving as an internal probe.
The homologously recombined ES cell clones were injected into C57BL/6 blastocysts, as described previously [63]. Blastocyst injections of ES cells resulted in fur-color chimeric mice. Breeding of chimeric mice with C57BL/6 mice yielded agouti offspring and germline transmission of the recombinant allele was checked by PCR analyses of isolated tail DNA. Heterozygous Cx47+/M282Tneo mice were crossed to hACTB:FLPe mice expressing the Flp-recombinase and litters were backcrossed to C57Bl/6 mice to generate Cx47+/M282T mice. The heterozygous Cx47+/M282T mice were backcrossed with C57BL/6 mice to increase the C57BL/6 genetic background. Genotyping of mice was performed by three primer PCR (primer A1: GCA GCA GAG ACG GCA AGG CCA CC, primer AC2: CCG GTC GCT ACC ATT ACC AGT TGG and primer A3: CAG AGA GAG GAG CTG TTC TTG GTC C) and deletion of the frt site flanked neomycin resistance cassette was verified by PCR using primer C2: CTC AGC TGC AGT AAG GGA TCT CTC G, primer C1 and primer AC2 (described above). Presence of the mutation was verified by PCR and subsequent BsmBI digestion as described for screening of ES cell clones. The different allelic variants were further verified by Southern blot analyses on MfeI digested DNA and hybridization with an 845 bp EcoRI/XhoI fragment as a 3′ internal probe.
Mice were killed by injecting an overdose of anesthetic (combination of ketamin hydrochloride and xylazine hydrochloride) intraperitoneally and subsequent transcardially perfusion with 40 ml phosphate-buffered saline (PBS) followed by 25 ml 2% phosphate-buffered formaldehyde solution (Roti-Histofix; Roth, Karlsruhe, Germany) containing 0.2% glutaraldehyde. Tissues of interest were dissected from the mice and postfixed for 24 h at 4°C. 50 µm sagittal vibratome (VT 1200 S, Leica, Wetzlar, Germany) sections were obtained from brains of Cx47M282T/M282T mice for investigation of development-dependent LacZ expression. For immunohistochemistry combined with β-gal staining samples were cryoprotected with 30% sucrose in PBS over night and 40 µm sections were cut with a cryostat (Microm, Walldorf, Germany) and collected on SuperFrost Ultra Plus slides (Menzel, Braunschweig, Germany). β-Gal staining was performed using the substrate 5-bromo-4-chloro-3-indolyl-b-galactoside (X-Gal) as previously described [64]. For doublelabeling immunohistochemical analyses, sections were stained overnight with X-Gal, exposed to PBS containing 0.1% H2O2 and 10% methanol for 20 min, blocked in M.O.M. reagent (Vector Laboratories, Burlingame, CA, USA) and incubated over night at 4°C with primary antibodies in PBS containing 4% BSA. As primary antibodies monoclonal mouse anti-neuronal nuclei (NeuN; 1∶100, Chemicon, Millipore GmbH, Schwalbach/Ts., Germany), monoclonal mouse anti-CNPase (1∶200, Sternberger Monoclonals) and monoclonal mouse anti-GFAP (1∶400, Sigma-Aldrich) were used. Immunohistochemical analyses were performed using the M.O.M. kit, according to manufacturer's instructions. Development of the peroxidase reaction was carried out using 3,3′-diaminobenzidine tetrahydrochloride (Sigma-Aldrich) as a substrate. Slices were air dried and mounted with Entellan (Merck Chemicals, Darmstadt, Germany).
For double immunofluorescence labelling mice were transcardially perfused with PBS containing 1% formaldehyde and 0.1% picric acid, and brains were postfixed in 1% phosphate buffered formaldeyde solution for 2 h and cryoprotected with 30% sucrose in PBS over night. 16 µm cryosections were blocked with TBS (50 mM Tris, 150 mM NaCl, pH 7.5) containing 0.3% Triton X-100 and 5% NGS for 1 h at room temperature. Sections were incubated with both primary antibodies: mouse monoclonal anti-Cx47 (Zymed, 1∶250) and polyclonal rabbit anti-β-gal (ICN, 1∶500) at 4°C over night, washed with TBST (50 mM Tris, 150 mM NaCl, pH 7.5, 0.1% Triton X-100) and incubated for 1 h with appropriate goat polyclonal secondary antibodies conjugated to Alexa-488 and Alexa-594 (1∶500, Invitrogen, Karlsruhe, Germany) diluted in TBS containing 0.3% Triton X-100 and 5% NGS. After TBS washes, sections were mounted with PermaFluor (Thermo Fisher scientific, Waltham, MA, USA) and Images were taken with a Laser Scanning Microscope (LSM 510, Zeiss, Germany).
Dye coupling experiments were performed in parallel with experiments previously described [23] and we shared the controls between these two studies. Acute coronal brain slices containing the corpus callosum were prepared from postnatal days (P) 10–15 mice as previously described [65]. For patch clamp recordings slices were placed in a recording chamber and perfused continuously with artificial cerebrospinal fluid (aCSF) composed of (in mM): NaCl 134; KCl 2.5; MgCl2 1.3; CaCl2 2; K2HPO4 1.25; NaHCO3 26; D-glucose 10 and saturated with carbogen (95% O2, 5% CO2) to a pH of 7.4 at room temperature. Patch clamp recordings were performed as previously described [23]. Pipettes had a resistance ranging from 3 to 7 MΩ when filled with an intracellular solution containing (in mM): NaCl 4; KCl 120; MgCl2 4; CaCl2 0.5; Hepes 10; EGTA 5; D-glucose 5; 0.5% biocytin at pH 7.4. Alexa Fluor 594 (10 µg/ml, Invitrogen, Karlsruhe, Germany) was added to the pipette solution to confirm intracellular access. To improve voltage-clamp control, capacitance was compensated by TIDA software (HEKA Elektronik, Lambrecht, Germany). In each individual slice only a single cell was filled via the patch pipette during whole-cell recordings (20 min) [66]. Only cells with stable input resistance over the 20 min period were considered for data analysis. During recording, the membrane was continuously de- and hyperpolarized between −170 and + 50 mV from a holding potential of −70 mV (10 mV steps, 50 ms). Current signals were amplified (EPC 9/2 or EPC10 amplifiers, HEKA), filtered (3 kHz), sampled (5 kHz), and monitored with TIDA software (HEKA).
After recording, slices were fixed and processed for biocytin visualization with fluorochrome conjugated streptavidin combined with immunostaining for the oligodendrocyte marker CNPase and the astrocytic marker GFAP, as previously described [23]. Cy3-conjugated streptavidin (1∶200, Jackson ImmunoResearch/Dianova, Hamburg, Germany), mouse anti-CNPase (1∶200, Covance/HISS Diagnostic GmbH, Freiburg, Germany) and rabbit polyclonal anti-GFAP (1∶1.000, DAKO, Hamburg, Germany) were used. Primary antibodies were visualized by application of FITC-conjugated donkey anti-mouse IgG (1∶200) and Cy5-conjugated donkey anti-rabbit IgG (1∶200) or by FITC-conjugated donkey anti-chicken IgG (1∶200, secondary antibodies were purchased at Jackson ImmunoResearch/Dianova), respectively. No unspecific cross reaction between secondary antibodies was observed. Images were acquired by confocal microscopy (Leica TCS SP5, Leica, Solms, Germany) with Leica software (LAS AF Lite).
Electrophysiological data were analyzed and plotted using TIDA and Origin (MicroCal, Northampton, Massachusetts). Biocytin filled cells were counted on sequential confocal stacks (0.2–0.8 µm steps) of the 110 µm thick slices with Image-Pro plus (MediaCybernetics, L.P., Bethseda, MD, USA). The maximum extent of tracer spread, defined as the largest distance between two somata within a network, was measured with the same software used for image acquisition. Statistic analysis was performed with SPSS 11.5 for Windows (SPSS Inc., Chicago, Illinois). The cohort of controls used was the same as previously published since the dye coupling experiments described here were performed in parallel with the former study [23]. Differences between groups were evaluated with the Kruskal-Wallis nonparametric test followed by Mann-Whitney U-test for two independent samples, Bonferroni corrected for n-pair comparisons. All values are expressed as median, 25th and 75th percentiles. Values of specific glial cell types are given as percentage of the number of biocytin-positve cells per network. For comparisons between groups, the number of oligodendrocytes forming networks was evaluated with Chi Square-test. Data regarding the extent of tracer spread are expressed as mean ± standard deviation (SD) and were analyzed with one-way ANOVA, followed by Bonferroni post hoc analysis. P-values of <0.05 were considered statistically significant.
Mice were killed by cervical dislocation and whole brain or cerebella were quickly dissected out and flash frozen in liquid nitrogen. Tissues were stored at −80°C until extraction in homogenization buffer (20 mM Tris, 1% Triton X-100, 140 mM NaCl, 10% glycerol, 1 mM EGTA, 1.5 mM MgCl2, pH 8.0) containing protease inhibitors (Complete Mini; Roche Diagnostics). The samples were homogenized on ice with a glas pestle and subsequently sonicated two times for 15 s. After centrifugation for 15 min at 10.000 g and 4°C, the supernatant was retrieved in new 1,5 ml tubes and kept at −80°C or −20°C. HeLa-Cx47-eGFP [67] cell lysates were prepared in homogenization buffer. Cultured HeLa cells were harvested and lysed as described previously [68]. Samples were denatured at 60°C for 10 min (Cx47) or 95°C for 5 min containing 1× Laemmli buffer. Proteins (40–100 µg) were separated by electrophoresis on a 10–15% polyacrylamide gel and transferred to Hybond ECL membrane (Amersham Bioscience). Blots were preincubated in a blocking solution of 5% milk powder (MP) in TBST (50 mM Tris, 150 mM NaCl, pH 7.5, 0.1% Tween-20) for 1 h at room temperature, incubated with primary antibodies overnight at room temperature (5 h with anti-Cx47 antibodies) and after three wash steps, with horseradish peroxidase (HRP)-conjugated antibodies (1∶2.000–1∶10.000, Dianova, Hamburg, Germany).
Primary antibodies were polyclonal rat anti-MBP (1∶1.000 in 5% MP, Chemicon, Millipore GmbH, Schwalbach/Ts., Germany), monoclonal mouse anti-CNPase (1∶2.000 in Roti-block (Roth, Karlsruhe, Germany), Sigma-Aldrich) rabbit anti-Cx47 (1∶250 in 4% BSA, No. 36-4700, Invitrogen, Karlsruhe, Germany), rabbit anti β-gal (1∶500 in 5% MP, ICN) and monoclonal mouse anti-tubulin (1∶10.000 in 5% MP, Chemicon, Millipore GmbH, Schwalbach/Ts, Germany).
Protein bands were detected by incubation of the membranes with enhanced chemiluminescence reagents (Amersham Bioscience) and development on X-ray films. Densitometry analysis was performed with E.A.S.Y Win32 software and by normalizing the band intensities to tubulin values.
For immunoprecipitation, 250 µg protein of HeLa-Cx47-eGFP lysate and 600 µg protein obtained from whole brain lysate of wildtype Cx47+/+, Cx47−/− and Cx47M282T/M282T mice were incubated with 85 µl protein A sepharose CL-4B (Amersham Bioscience) each for 3 h at 4°C in TBS and subsequently centrifuged (12.000 g, 4°C, 30 min). Mouse Cx47 antibodies (No. 37-4500, Invitrogen, Karlsruhe, Germany, 2.5 µg) were incubated with 25 µl of protein A-Sepharose (Amersham Biosciences) on ice for 2 h. The precleared lysates were incubated with the protein A-Sepharose antibody complex overnight at 4°C and washed three times with TBST. The proteins were eluted in 18 µl of Laemmli buffer (60°C, 10 min) and separated by electrophoresis as stated above.
Mice were killed with an overdose of chloroform and sequentially perfused via the left ventricle with phosphate buffer with 1% procaine HCL and then 6% glutardialdehyde in phosphate buffer. Tissues of interest were dissected from the mice and embedded in epoxy resin (Epon 812/glycid ether) after postfixation in 2% OsO4. Sections were cut at a thickness of 1 µm heatmounted on aminosilane-coated slides and stained with toluidine blue and pyronin G.
At different ages mice were killed by injecting an overdose of anesthetic intraperitoneally. Mice were transcardially perfused with 40 ml phosphate-buffered saline (PBS) followed by 25 ml phosphate-buffered formaldehyde solution 4% (Roti-Histofix; Roth, Karlsruhe, Germany). The brains were rapidly prepared and postfixed in 2% phosphate-buffered formaldehyde solution for at least 48 h at 4°C. 25 µm vibratome sections were obtained (VT 1200 S, Leica, Wetzlar, Germany) and free floating slices were incubated with 0.1% H2O2 in 10% methanol/PBS (pH 7.4) for 30 min to inhibit endogenous peroxidase activity, washed in PBS and incubated in blocking solution (5% normal goat serum (NGS), 4% BSA, 0.3% Triton X-100, 0.01% NaN3 in PBS) for 2 h at room temperature to avoid unspecific crossreactivity. Primary antibodies were applied in blocking solution for 16 h at room temperature. After washing with 0.2% Triton X-100 in PBS, sections were incubated with the appropriate biotin-conjugated secondary antibody for 2 h at room temperature and washed again. Vectastain Peroxidase ABC reagent (Vector Laboratories, Burlingame, CA, USA) was applied following the manufacturer's protocol. After 30 min incubation in working solution free floating sections were washed in 0.05 M Tris for 20 min and then transfered to bidest for at least 5 min prior to NovaRed (Vector Laboratories, Burlingame, CA, USA) staining. Sections were mounted on glass slides, air dried at 45°C on a slide warmer and coverslipped with Entellan (Merck Chemicals, Darmstadt, Germany). As primary antibodies rat anti-MBP antibodies (1∶750, Chemicon, Millipore GmbH, Schwalbach/Ts, Germany) were used for myelin staining, rabbit polyclonal anti-Iba1 antibodies (1∶750, Wako Chemicals GmbH, Neuss, Germany) were used for staining of microglia and rabbit polyclonal anti-GFAP antibodies (1∶1.000, Dako, Carpinteria, CA USA) were used for detection of astrocytes.
The mice were housed individually in macrolone cages (Type 2, 22×16×13 cm) with metal covers and were given free access to standard rodent diet (Ssniff, Spezialdiäten GmbH, Soest, Germany) and water. They were maintained on a 12 h light-dark schedule with lights on at 7 a.m. and were tested during the light phase between 9 a.m and 6 p.m. All experiments were approved by the North Rhine Westphalia State Authority in accordance with the German legislation on animal experimentation (German Animal Welfare Act, TSchG).
Each animal was subjected to three trials in the open-field with an inter-trial interval of 24 h. The open-field was constructed of grey plexiglas and had the following dimensions: 30 cm (height)×30 cm (length)×30 cm (width). It was illuminated by diffuse white light providing a low illumination density of approximately 10 lux at the center and was placed in a sound-attenuated chamber. A camera mounted above the open field transmitted the digitized image of the open-field to a computer and a VCR. The base area of the open field was subdivided by software into nine equal sized quadrants, providing a 10×10 cm central quadrant and four corner quadrants. Each mouse was individually placed into the central quadrant of the open-field and allowed to freely explore the arena for 10 min. After each trial, the entire apparatus was cleaned with 50% ethanol solution and dried thoroughly. A semi-automated video tracking system (Ethovision Software, Noldus, The Netherlands) was used to quantify the animal's horizontal (the distance travelled in cm, mean running speed in cm/s), and vertical activity (number and duration (s) of rearings on the hind limbs).
Motor coordination and balancing was tested with an accelerating rotarod (TSE systems; Bad Homburg, Germany; model no.: 7650). The rotating rod was elevated 10 cm off the floor, had an axis diameter of 3.5 cm and a striated surface made of black rubber. During the acquisition phase, each mouse was given three trials (with an inter-trial interval of ≥25 to control for possible effects of physical exhaustion) per day for three consecutive days. On each trial, the mouse was placed on the inactive drum, with its head pointing towards the direction opposite to that of the rod's motion. The mouse had to move forward on the drum, which was rotating along its vertical axis, in order to avoid falling off. Over a period of 300 s, the rod step-wise accelerated to a speed of 50 rpm. As some mice tend to passively ride around the rod, especially at higher velocities, the duration (s) of active performance until the mouse fell off the drum was registered with a cutoff after 300 s. After retention delay of one week the long-term motor memory of the animals was evaluated during three trials.
Data are presented as mean ± standard error of mean (SEM). Data were analyzed by repeated measures ANOVA and by pre-planned Bonferroni tests between transgenic and wild-type mice. P-values lower than 0.05 were considered statistically significant.
|
10.1371/journal.pcbi.1003548 | Mechanisms of Zero-Lag Synchronization in Cortical Motifs | Zero-lag synchronization between distant cortical areas has been observed in a diversity of experimental data sets and between many different regions of the brain. Several computational mechanisms have been proposed to account for such isochronous synchronization in the presence of long conduction delays: Of these, the phenomenon of “dynamical relaying” – a mechanism that relies on a specific network motif – has proven to be the most robust with respect to parameter mismatch and system noise. Surprisingly, despite a contrary belief in the community, the common driving motif is an unreliable means of establishing zero-lag synchrony. Although dynamical relaying has been validated in empirical and computational studies, the deeper dynamical mechanisms and comparison to dynamics on other motifs is lacking. By systematically comparing synchronization on a variety of small motifs, we establish that the presence of a single reciprocally connected pair – a “resonance pair” – plays a crucial role in disambiguating those motifs that foster zero-lag synchrony in the presence of conduction delays (such as dynamical relaying) from those that do not (such as the common driving triad). Remarkably, minor structural changes to the common driving motif that incorporate a reciprocal pair recover robust zero-lag synchrony. The findings are observed in computational models of spiking neurons, populations of spiking neurons and neural mass models, and arise whether the oscillatory systems are periodic, chaotic, noise-free or driven by stochastic inputs. The influence of the resonance pair is also robust to parameter mismatch and asymmetrical time delays amongst the elements of the motif. We call this manner of facilitating zero-lag synchrony resonance-induced synchronization, outline the conditions for its occurrence, and propose that it may be a general mechanism to promote zero-lag synchrony in the brain.
| Understanding large-scale neuronal dynamics – and how they relate to the cortical anatomy – is one of the key areas of neuroscience research. Despite a wealth of recent research, the key principles of this relationship have yet to be established. Here we employ computational modeling to study neuronal dynamics on small subgraphs – or motifs – across a hierarchy of spatial scales. We establish a novel organizing principle that we term a “resonance pair” (two mutually coupled nodes), which promotes stable, zero-lag synchrony amongst motif nodes. The bidirectional coupling between a resonance pair acts to mutually adjust their dynamics onto a common and relatively stable synchronized regime, which then propagates and stabilizes the synchronization of other nodes within the motif. Remarkably, we find that this effect can propagate along chains of coupled nodes and hence holds the potential to promote stable zero-lag synchrony in larger sub-networks of cortical systems. Our findings hence suggest a potential unifying account of the existence of zero-lag synchrony, an important phenomenon that may underlie crucial cognitive processes in the brain. Moreover, such pairs of mutually coupled oscillators are found in a wide variety of physical and biological systems suggesting a new, broadly relevant and unifying principle.
| The study of large-scale brain dynamics, and the cortical networks on which they unfold, is a very active research area, providing new insights into the mechanisms of functional integration and complementing the traditional focus on functional specialization in the brain [1], [2]. Whilst progress towards understanding the underlying network structure has been impressive [3], [4], the emergent network dynamics and the constraints exerted on these dynamics by the network structure remain poorly understood [5]. The problem is certainly not straightforward, as the dynamics between just a pair of neural regions already depends critically on the nature of the local dynamics and the nature of the coupling between them [6]: Although non-trivial, a complete description of nonlinear dynamics between a pair of nodes is nonetheless typically possible [7]. However, aggregating such duplets into larger arrays and introducing noise and time delays leads to further challenges and prohibits an exact description of the precise functional repertoire, motivating recourse to the broader objective of finding unifying and simplifying principles [8]. Structural and functional motifs – small subnetworks of larger complex systems – represent such a principle [9]. As depicted in Fig. 1 a, they characterise an intermediate scale of organization between individual nodes and large-scale networks that may play a crucial role as elementary building blocks of many biological systems [10]. Motif distribution in cortical networks has also been shown to be highly non-random, with a small set of motifs that appear to be significantly enriched in brain networks [9]. The relative occurrence of 3-node motifs in three different anatomical networks of the Macaque brain and cat cortex (Figs. 1 b–e) is shown in Figs. 1 f–i. These motifs may play distinct roles in supporting various computational processes. In this report we examine the principles of neuronal dynamics that emerge on small motifs and consider their putative role in neuronal function.
The mechanisms supporting zero-lag synchrony between spatially remote cortical regions can be considered paradigmatic of those mediating between structure and function. Since first reported in cat visual cortex [11], zero-lag synchrony has been widely documented in empirical data and ascribed a range of crucial neuronal functions, from perceptual integration to the execution of coordinated motor behaviours [12]–[16]. In particular, zero-lag synchrony between populations of neurons (quantified through synchrony between the local field potentials) may play a crucial role in aligning packets of spikes into critical windows to maximize the reliability of information transmission at the neuronal level [17], and to bring mis-aligned spikes into the time window of spike-time-dependent plasticity [18]. The situation is particularly pertinent in sensory systems, where precise differences in the timing of inputs, between left and right cortex for example, may carry crucial information about the spatial location of the perceptual source [19]. However, the empirical occurrence of zero-lag synchronization is at apparent odds with the observation that two mutually coupled oscillators interacting through a time-delayed connection do not, in general, exhibit zero-lag synchrony [20]. Indeed, in many models of neuronal systems the presence of a reciprocal delay has been found to introduce a ‘frustration’ into the system such that zero-lag synchrony is unstable and out-of-phase synchrony is instead the preferred dynamic relationship [21]. In fact, this phenomenon occurs quite generally in systems of oscillators with time-delayed coupling [21], [22].
Complex dynamics in spatially embedded systems arise in a broad variety of physical and biological contexts. Arrays of coupled semiconductor lasers are a prominent example. Because of their extraordinary internal speed, even small time delays due to the finite speed of light are usually nonnegligible in arrays of coupled lasers [23]. Detailed analysis of delay-coupled laser systems has suggested that an intermediate and reciprocally coupled relay node in a motif of three nodes could represent a general mechanism for promoting zero-lag synchrony in delay-coupled systems [24]. In previous work, it was also shown that such motif arrangements also represent a candidate mechanism for zero-lag synchrony in delay-coupled neuronal systems [25]. This is encouraging because there exist several candidate neuronal circuits in the mammalian brain which are characterized by reciprocal coupling between an intermediate delay node, including corticothalamic loops and the hippocampus [26], [27]. There also exist strong reciprocal connections in the visual system, such as the heavily myelinated connections between primary visual cortex and the frontal eye fields. Indeed, the corresponding motif occurs disproportionally in mammalian cortex (Fig. 1), hence being embedded in many cortical subsystems [9].
The presence of a node that drives two common-driven nodes that reach zero-lag synchrony between them due to the driver's influence is intuitively appealing and finds anatomical support, for example, by shared input through bifurcating axons [13]. Certainly, a common-driving input of sufficient intensity can generate virtually perfect spike-time correlation, as long as the time delay to both driven nodes is identical. However, this scenario is not robust if the time delays lose symmetry or the coupling is not sufficiently strong. The common-driving setup is nonetheless a key prototype that offers insights into the synchronization between the driven nodes and the roles of the dynamics of the nodes [28]–[32]. Here we consider dynamics on the 3-node motifs that occur abundantly in large-scale networks of the brain (Fig. 1), adding connections to the prototypical common-driving motif. We confirm that common driving – a coupling arrangement that is widely invoked in the literature – is an ineffective means of inducing zero-lag synchrony in the presence of weak coupling (a neurophysiologically plausible regime). However, the additional incorporation of a single reciprocally coupled connection between the driver and an edge node – which leads to synchrony between that pair – is found to be a novel and efficient way of promoting zero-lag synchrony amongst other nodes in these small motifs. We demonstrate that this effect – which we term resonance-induced synchrony – arises consistently in candidate computational models at the neuronal, population and mesoscopic spatial scales and is robust to mismatches in system parameters and even time delays. Remarkably, we show that the resonance effects of a synchronized pair are not necessarily localized, but may instead propagate throughout the network. We hence propose resonance-induced synchrony as a general and unifying mechanism of facilitating zero-lag synchrony in the brain.
We studied zero-lag synchronization – quantified as the average zero-lag cross-correlation between two nodes A and B () – in a variety of different motifs involving a common driving node. We considered the dynamics of nodes expressing different neuronal systems across a hierarchy of scales. At the microscopic scale, each node was modeled to represent a single spiking Hodgkin-Huxley neuron; at the circuit scale, each node was taken to represent a population of 400 excitatory and 100 inhibitory randomly connected neurons described by the Izhikevich model; and at the mesoscopic scale each node was modeled as a neural mass model with chaotic activity. This last model permits systematic parameter exploration that is not possible with populations of spiking neurons. In all the three modeling levels, coupling between nodes was via excitatory chemical synapses (see Methods for details on models and integration scheme). For the sake of simplicity, we initially assumed homogeneous delays in the motifs, i.e., all connections between nodes had the same time delay. We later explored the robustness of the results when relaxing these assumptions in the section “Mismatch in the conduction delays”.
The notation we adopt for the motifs of three nodes follows the notation of Sporns and Kötter (2004) [9] who denoted all 13 possible connected subgraphs (motifs) composed of three nodes, denoted from M1 to M13 (Fig. 1). The genuine common-driving motif (illustrated in Fig. 2) is designated M3. Node 2 is the common driver whereas nodes 1 and 3 are the common-driven ones. In particular, we pay special attention to the cross-correlation between nodes 1 and 3. With the exception of illustrative time traces and their corresponding analysis, the results represent an average over 40 independent runs, unless otherwise stated, with different random initial conditions and with error bars given by the corresponding standard deviation. We characterize the synchronization in other motifs that represent structural variations of the M3 motif: the addition of one or more connections (M6, M8, M9, M13), or the addition of connections and nodes (e.g., M3+1). In particular, M9 is the prototypical dynamical-relaying motif [24], which has been previously shown to promote zero-lag synchronization in a variety of systems [33]–[38], including neuronal systems [25]–[27].
We first focus on the four motifs depicted in Fig. 2. The simple common driving motif (M3), in which node 2 drives the dynamics of nodes 1 and 3 was contrasted with three other motifs (M6, M9 and M3+1), which represent structural variations of M3. Because motif M3 lacks any feedback or cyclical structure, the conduction delay plays no role in the dynamics or in the synchronization between nodes 1 and 3: Hence the outer nodes passively receive the driver's input. Onto this “backbone”, motif M6 has a single feedback connection added, forming a reciprocal connection between nodes 1 and 2. Motif M9 has reciprocal connections between node 2 and nodes 1 and 3. Motif M3+1 possesses an extra node (4) reciprocally connected with node 2.
Biological systems are naturally diverse, and therefore, any relevant behavior should not be highly dependent on the fine-tuning of the delay – and particularly its symmetry. We next tested the generality of the zero-lag synchronization between nodes 1 and 3 with respect to delay mismatch in the different motifs containing the resonance pair. The connections preserved the conduction delay of except for a single feedback connection to the driver node 2 in motifs M6′, M9′ and M3+1′ in which we introduced a variable conduction delay in one direction (), as illustrated in Fig. 6 a. The three motifs exhibited zero-lag synchronization that was substantially larger than that of motifs M3 (black line) or even M3 plus a unidirectional input (yellow line) across a large region of the parameter space (Figs. 6 b–d). In the motifs of Hodgkin-Huxley neurons (Fig. 6 b), the behaviors of all three motifs are similar for . In contrast, for zero-lag synchrony decays in a similar way for motifs M6′ and M3+1′, whereas synchronization in motif M9′ is virtually independent of for up to fivefold (not shown in the plot). Supplementary Fig. S9 shows the analyses of the dynamics of motif M6′ in more detail: It shows that synchronization arises in M6′ only when the delay mismatch yields synchronization with the same phase relation as , which – in the case of – is anti-phase synchronization between neighboring neurons (see Fig. 3). The motifs of neural mass models show a systematic consistency of synchronization across for a biologically plausible range of delays (Fig. 6 c). However, a behavior similar to that observed in motifs of Hodgkin-Huxley neurons occurs for greater delay mismatches (Fig. 6 d). Such differences in the time scales are consistent with the different time scales of these systems: The Hodgkin-Huxley neurons oscillate with periods of about 15 ms, whereas the neural masses oscillate with periods of about 110 ms.
From herein, we focus on motifs of neural masses, exploiting their relative computational parsimony to gain deeper insight into the mechanisms of the resonance pair. In particular we studied the robustness of our findings with respect to the most salient parameters of the system, namely the coupling strength and the delay. As shown in Fig. 7, the strength of the synchronization in the motifs with a resonance pair, but not M3, show an increase as a function of coupling strength (panels a, b). Although an expected feature of the model [49], the emergence of synchrony even at very weak coupling () is somewhat surprising for a biological system. There are, however, some regions of complex dynamics (evidenced as large error bars) in which there is not a unique solution, thereby entailing significant trial-to-trial variability. At relatively weak coupling (c = 0.01), zero-lag synchronization between nodes 1 and 3 holds across a broad regime of physiologically plausible time delays (Fig. 7 c). Analysis of longer coupling delays (supplementary Fig. S10) reveals an influence on synchronization that resembles the system of Hodgkin-Huxley neurons (Fig. 3), albeit weaker and at a much longer time scale.
These analyses suggest a partition of the common-driving motifs into three distinct families: (i) The simple common-driving motif (M3) where synchronization at zero lag is not achieved in the weak-coupling regime, independent of the time delay; (ii) A ring of three mutually coupled systems (M13) or a common-driving motif that also contain direct coupling between the driven nodes (M8) require a relative strong coupling and negligible delay in order to promote synchronization (Figs. 7 d–f), because of the existence of frustration; and (iii) Common-driving motifs enhanced by active resonance pairs (e.g., M6, M9, M3+1) which exhibit zero-lag synchronization even for very small couplings, irrespective of the time delay (up to ). It is clear in these analyses that the increase in zero-lag synchrony in motifs with a resonance pair is not due to the additional coupling introduced by the backward connection, but rather through the placement of the additional edge. For example, the motifs with the greatest number of edges (M8 and M13) are amongst the most difficult to achieve zero-lag synchrony with an increase in coupling. Closing the outer nodes with two additional edges (going from M9 to M13) leads to a substantial decrease in zero-lag synchrony.
The preceding analyses show that the effect of the resonance pair can influence the common driving motif even when it is placed outside the motif itself (e.g. M3+1). Here we further investigate the propagation of the resonance pair effect by considering larger structures in which the resonance pair is distant from the driver node (2). This procedure is schematically shown in Fig. 8 a, and illustrated for a particular network of N = 7 nodes in Fig. 8 b. We are particularly interested to understand if the effects of the resonance pair are strictly local, and, additionally, on how the polysynaptic distance to the resonance pair influences the dynamics and synchronization.
We observe that zero-lag synchronization between the driven nodes 1 and 3 is virtually independent of the distance along a polysynaptic chain from the resonance pair (Fig. 8 c). For a fixed motif length (N = 7), we also characterized the zero-lag synchronization of different pairs of nodes that did not interact directly, but interacted indirectly through a common neighboring mediator (see Fig. 8 d). Apart from pairs 5–7, all such pairs correspond to a strict flux of information flow, mandated by the direction of the coupling. Thereby, the synchronization decreased with the distance from node 7, unless the system was set with a specific coupling (see arrow in Fig. 8 d) that gives rise to global synchronization. This corresponds to identical synchronization between nodes 2, 5 and 7, which are anti-phase synchronized to nodes 1, 3, 4 and 6 occurring at this particular coupling strength.
Finally, to highlight the influence of the resonance pair in the dynamics, we removed the feedback connection to node N (results shown as thin dotted lines in Figs. 8 c and d). By means of this control simulation, we find that: (i) Zero-lag synchronization between 1–3 is consistently reduced (Fig. 8 c); and (ii) Zero-lag synchronization between 5–7 (Fig. 8 d) completely disappears in the absence of a resonance pair.
We have denoted an active resonance pair as two mutually connected nodes that synchronize in the presence of appropriate time delays and coupling strength. This effect propagates through the motifs because the driven nodes show a strong tendency to synchronize with the driver node (hence promoting zero-lag synchronization between driven nodes). That is, the emergence of synchronization between the resonance pair then stabilizes synchrony amongst unidirectionally coupled nodes. The same phenomenon underlies the propagation down a polysynaptic chain (Fig. 8). Interestingly, the impact of the resonance pair extends beyond this propagation, giving rise to other dynamical effects for coupling delays in which anti-phase synchrony between neighbors prevails. Geometrical frustration is an example: In some motif configurations, anti-phase synchrony between pairs of mutually connected nodes (potential resonance pairs) is simply not a stable solution. In the case of motif M13 (illustrated in Fig. 7), for example, anti-phase synchronization between any pair is frustrated because the third node cannot be simultaneously synchronized in anti-phase with respect to the other two neighbor nodes. This situation illustrates that frustration can disturb potential resonance pairs. Large mismatches in the delays of the mutual connection between the pair can also disturb the effects of a resonance pair. As depicted in Fig. 6, both motifs M6′ and M3+1′ are similarly susceptible to mismatches in the reciprocal latencies.
Connectivity also plays a role on the onset of synchronization. We studied the temporal onset of zero-lag synchronization in neural mass models for different motifs by (1) examining the transient dynamics following random initial conditions, and (2) studying the response to a transient perturbation. An example is shown in Fig. 9 a, in which dynamics on M6 begin from random initial conditions, then approach synchronization between masses 1 and 3. The dynamics are then perturbed by a brief current from 800 to 1000 ms – that is distinct for each driven node – before rapidly regaining synchrony after a few hundreds of milliseconds. It is noteworthy that the approach to zero-lag synchrony in both scenarios is approximately exponential, with an exponent that can be used as a numerical estimate of the stability of the synchronous state (Fig. 9 b). In contrast, edge nodes on motif M3 remain unsynchronized. The dependence of the exponent with the coupling strength for the 1200 ms following offset of the transient perturbation is shown in Fig. 9 c. Motifs with resonance pairs (M6 and M9) showed a negative exponent, consistent with stable synchrony, whereas the exponent associated with motif M3 was positive throughout. Interesting, the coupling strength associated with the strongest synchrony (most negative exponent) occurred for a relatively weak coupling strength of c = 0.01.
Synchronization hence arises quickly in the presence of a resonance pair. Is it possible to adjust the dynamics of the driver node without such reciprocal coupling to induce synchronization? We next studied this possibility by fine-tuning the input current ( to the driver node (2) in motif M3, whilst keeping all other parameters fixed. As shown in Fig. 10 a, introducing a slight mismatch in the input current can indeed lead to large changes in the zero-lag synchronization between nodes 1 and 3. Crucially, careful fine-tuning of this current mismatch can lead to a near complete synchronization in motif M3 (A), or at least lead to a strong enhancement of synchronization (B and C). As depicted in Fig. 10 b, the maximum synchronization (A) occurs when the input current causes the driver node to exhibit the same oscillatory frequency as the driven edge nodes. The other local maxima occur when the driver node oscillates with a frequency that is an integer multiple of the driven nodes (2∶1 in B and 3∶1 in C). In contrast to this need for fine-tuning in motif M3, the resonance pair guarantees that node 2 oscillates with the same frequency as the driven nodes, with strong synchronization hence arising regardless of the coupling strength, as shown in Fig. 10 c for motif M3+1.
The effects of a resonance pair can enhance the synchronization locally and even propagate in a polysynaptic way to influence distant dynamics. Reciprocally connected nodes can also interact in a way that disturbs the synchronization if they introduce frustration as in motifs M8 and M13, as shown in Fig. 7. To more deeply understand the role of reciprocally connected nodes and loops, we studied resonance motifs that go beyond the resonance pairs. Starting with a common driving motif M3, we added chains of bi- or uni-directionally coupled nodes of varying sizes as shown in Fig. 11 a. Adding one node reciprocally connected to node 2 recovers the resonance pair, which is clearly a more effective way of synchronizing the driven nodes than adding one extra unidirectionally connected node (the blue dashed line of Fig. 11 b). The addition of two reciprocally connected extra nodes in a closed loop (resonance triplet) had an effect that was analogous to the resonance pair, and again far more effective than the counterpart of two extra unidirectionally connected nodes in a loop (green dashed line). The addition of three or more reciprocally connected extra nodes in closed chain had a similar effect to the resonance pair. However, the influence of the unidirectionally coupled loops gradually approaches that of their reciprocally connected counterparts, which have already attained the ceiling effect (magenta dashed line). Hence, the interaction of unidirectionally connected nodes in a loop gradually enhances the synchronization of the driven nodes as the size of the loop increases. Therefore, even in the absence of reciprocally connected nodes, synchronization between 1 and 3 can be enhanced by a loop of at least three extra nodes connected to the driver node. Interestingly, the addition of a single resonance pair is the most efficient means of achieving zero-lag synchronization compared to loops of any size.
Our final analysis concerns the synchronization properties of commonly driven nodes with higher polysynaptic orders (Figs. 12 a–c). In particular, we study the synchronization of the symmetrically located nodes n−n′ for the different connectivity states of the driver node A. Figure 12 a illustrates the case in which node A was part of a resonance pair together with node B; Fig. 12 b illustrates the case in which node A received a unidirectional input from node B; Fig. 12 c illustrates the case in which node A did not receive input from any neighboring regions. It can be seen in Figs. 12 d–g that only the motifs with the resonance pair (red line) yielded high correlation between nodes n and n′ (for n = 1,2,3,4). Interestingly, when the coupling strength is fixed (c = 0.024) and the number of elements further increased (Fig. 12 h), the cross-correlation coefficient remained quite high for the chain containing the resonance pair. A similar behavior occurred for the maximum cross-correlation coefficient (for all time delays) between node A and node n (Fig. 12 i): Again, the resonance pair was required for the propagation of synchronous activity.
Zero-lag synchronization between distant neuronal populations confers a number of important computational advantages, and finds broad empirical support. Here we report that common driving of passive nodes by a central “master” (motif M3), a scenario that is broadly assumed to underlie zero-lag synchrony, fails completely in the weak-coupling regime and is sensitive to parameter mismatch. However, the addition of one or more mutually coupled pairs fosters the emergence of zero-lag synchrony in the outer nodes of triplet motifs, and beyond. We find that this effect is robust to many of the particular details of the system, the spatial scale and parameter asymmetry, and can propagate through a multi-synaptic relay chain. In stark contrast, the further addition of a reciprocal connection between the driven nodes introduces frustration for delays that favor out-of-phase synchrony and fails to promote zero-lag synchronization. The disruptive effect of adding new edges that close the motif reinforces the observation that it is the topology (not the total amount of coupling) that determines the zero-lag synchrony. This is also evident by the fact that an increase in the coupling over two orders of magnitude in the unidirectional motif (M3) is less effective than adding a single feedback connection (where the effective coupling within that pair is simply doubled).
We have denoted this reciprocal pair a resonance pair because it can induce zero-lag synchronization between outer nodes. We find that an entire family of three- and four-node motifs exhibits zero-lag synchronization in the presence of such a resonance pair. Perhaps the archetypal motif in this family is M9 (see Fig. 1) also known as the dynamical relaying motif [24]–[27], [33]–[38], [49]. This motif contains two active resonance pairs (Fig. 1). Here we find that one feedback connection to the driver node can be removed (i.e., transforming the motif into M6) without compromising the synchronization between the outer nodes (confirming a recent observation in electronic circuits [50]). Similarly, the addition of one extra node mutually connected to the driver node, M3+1 (thereby comprising a resonance pair) causes robust zero-lag synchronization of the driven nodes where M3 alone fails. This indicates that a necessary condition for nodes 1 and 3 to synchronize is that the resonance-pair nodes also synchronize, regardless of their exact phase relationship. The synchronization of the resonance pair appears in turn to enhance its propensity to synchronize the driven nodes because when the driving node is synchronized its internal incoherence diminishes: This change in the regularity of the master node in turn enslaves the unilaterally driven node onto the synchronization manifold (Fig. 9). Thereby, we propose that the mechanism that promotes zero-lag synchronization in the dynamical relaying motif is indeed the resonance pair, in common to all other motifs in the broader family we examined.
We observed the effect of the resonance pair in a variety of different models (Hodgkin-Huxley neurons, populations of Izhikevich neurons, and neural mass models) and scales: motifs of neurons and motifs of cortical regions. The results are also robust with respect to the delay, the coupling strength, the oscillatory frequency band, and arise in autonomous, chaotic systems as well as noise-driven excitable dynamics. It seems reasonable to propose that resonance-induced synchronization will prove important for other neuronal systems, such as dendritic oscillations in single-neuron dynamics [51], and indeed other physical and biological systems of any domain characterized by weak interactions. Although the responses of neural populations to noisy inputs have been well studied [52], it remains to be seen if our results prove robust to further physiological details, including embedding stronger synaptic inputs into the noisy background [53] and stronger balanced background inhibitory and excitatory inputs [30]. We also note that although our study focused mainly on interactions with time delay, the resonance-induced synchronization can also occur in systems with no time delay (Figs. 7 b and c, and supplementary Figs. S3, S5, S7, S8 and S10).
Despite the robustness of the present effect in different classes of models and dynamical regimes, the universality and extent of the phenomenon remains to be clarified. Phase-resetting curves (PRCs) can be useful to predict whether phase or out-of-phase synchronization will arise [54]: This is a crucial factor in the dynamics because frustration does not occur in the case of in-phase synchronization. While usually studied in systems without delay, PRCs can also be used in systems in the presence of conduction delays [25], [55]. Analysis of the PRC can also be employed for formal stability analysis of synchronization of motif dynamics [25]. A second caveat, at least in the model of population of spiking neurons, is the type of dynamics studied – namely that in the dynamical regime studied here, neurons spike at least once per population cycle. An alternative approach would be to analyze synchronization in motifs of populations of spiking neurons in a sparsely synchronized regime [56] – that is when individual neurons spike less often than the background ensemble cycle. Further analysis is hence required to elucidate the extent to which our results translate to other physical and biological systems, perhaps focusing on canonical models that are more amenable to mathematical analysis such as the Kuramoto system.
Computational studies of anatomically derived brain networks have shown that motifs M9 and M6 are the first and second most abundant of all three-node motifs in the macaque visual cortex [9] and are among the most frequent motifs in other cortical networks (Fig. 1). Moreover, they appear to be clustered around the core “rich club” backbone of the structural connectome [57]. The presence of a resonant-pair in these motifs, and the robust zero-lag synchrony that they confer, may provide a dynamical advantage for these pairs. However, given the additional wiring cost, it is not clear why motif M9 is more common than M6. A possible explanation we provide derives from our observation that synchronization on motif M9 is robust to longer delays in one branch of the resonance pair in comparison to M6 (Fig. 6). Hence, the gain in robustness might overcome the cost of maintaining this extra feedback connection.
The influence of a resonance pair is not limited to local synchronization dynamics but also, through propagation, to larger networks, decaying only slowly with the polysynaptic distance (see Figs. 8 and 12). In a sufficiently sparse network like the brain, the number of neurons grows roughly exponentially with the inter-node distance. The coexistence of the slow decay (long correlation length) of the influence of the resonance pair, with rapid growth in the number of affected elements as a function of synaptic distance suggests that the zero-lag synchronization arising locally through a resonance pair has the capability to impact globally on network dynamics. Reframed in terms of a branching process, the slow decay of zero-lag synchronization and rapid growth of neuronal connectivity could lead to critical or supercritical propagation of zero-lag synchrony, consistent with prior theoretical considerations [58], and also suggesting a means for analytic extension of the present results.
The notion of motifs as fundamental building blocks of complex networks has yielded considerable prior success [9], [10], [53]. Degree distribution, the relative density of reciprocal synapses, convergence, divergence, and chains of synapses have been shown to play a crucial role in shaping the dynamics and synchronization properties of large networks [59]–[62]. In contrast to these studies, which focus on the global statistical features of large-scale networks, we have focused on particular features of small motifs. Future work, aimed at immersing these small motifs into larger networks, and focusing on the role of reciprocal nodes on the global synchronization properties of such networks, would be of significant interest. Our work confirms that the interplay between structural, functional and effective connectivity, while likely complex [63], may nonetheless be reliant upon a small number of unifying principles.
We simulated neuronal motif dynamics at different scales, and for different dynamical scenarios. First, representing the microscopic scale, each node was taken as a single neuron. For this endeavor we utilized the Hodgkin-Huxley model. Second, at the circuit scale, we took each node as a large population of spiking neurons. Third, at the mesoscopic scale, we considered a simplified coarse-grained version in which each population was taken as a neural mass model.
Each node was modeled by the well-known Hodgkin-Huxley equations [39]. The dynamics of the membrane potential depends on sodium, potassium, leaky, and synaptic (intra-motif and external) current components,(1)where is the membrane capacitance. The maximal conductances of the channels occur for completely open channels, with conductances given by , , and , and , , and stand for the corresponding reversal potentials. Generally, the voltage-gated ionic channels are not fully opened. The probability of finding them open depends on the gating variables. The channel depends on the combined effect of gating variables and , whereas depends on . They evolve according to the equations,(2)(3)(4)Hodgkin and Huxley set the empirical functions and to fit the experimental data of the squid giant axon,(5)(6)(7)(8)(9)(10)The synaptic current due to the interactions between neurons of the motifs are given by,(11)where , , and stands for the Dirac delta function. The summation over stands for the spikes of the presynaptic neurons (all excitatory). is the time at which the spike occurred. We varied the conduction delay . In agreement with the literature [40], the delay can shape the synchronization (Fig. 3). The external current incoming to each neuron is,(12)where , , runs over excitatory spikes, and corresponds to the spike times, modeled by an independent Poisson process for each neuron with rate . As shown in supplementary Fig. S11, nearly identical results can be also obtained by assuming the external current term as synaptic contribution and including it as an extra term in equation (4) with , and . The equations were integrated by the Runge-Kutta method of fourth order, with time steps of . Initial transient dynamics were discarded.
For this large-scale circuit model, each node represented populations of 500 randomly connected neurons described by the Izhikevich model [43]. 400 neurons were excitatory and 100 neurons were inhibitory. The neurons were described by the following equations:(13)where represents the membrane potential, represents the recovery variable, accounting for the and ionic currents, and is the total synaptic current. The neurons have a threshold at . Once this value is reached, is reset to and to . Following [44], we added dispersion to these four parameters (, , and ) to account for neuronal heterogeneity. Excitatory neurons have , and , where is a random number drawn from a uniform distribution in the interval [0,1]. Inhibitory neurons have , and .
Each neuron receives input from 80 neurons of the same population and from 25 excitatory neurons of each afferent population. The synaptic current is given by(14)where the dynamics of the excitatory and inhibitory synapses are described by(15) in the equations above stands for the Dirac delta function. The summation over () stands for the spikes of the presynaptic excitatory (inhibitory) neurons. () is the time at which the excitatory (or inhibitory) spike occurred. Conduction delays , associated with excitatory long-range connections, varied. We modeled short-range (intra-node) connections with negligible delays. Synapses were modeled by exponential decay functions [64], with time constants for excitatory and for inhibitory synapses. Each neuron was subject to an external driving given by independent Poisson spike trains at a rate of , which was also included in the sum over excitatory postsynaptic contributions ( index) of the equations above. With these parameters, individual neurons fire spontaneously, although not periodically.
The equations were integrated using a fixed-step first-order Euler method with time steps of 0.05 ms, starting with random initial conditions. To avoid spurious synchronization at the onset of simulations, neural populations were activated with random noise in 600 ms sequential windows (with a 500 ms overlap). The first transients of 1 s were discarded before further analysis.
The preceding large-scale circuit model is a high dimensional system. Whilst the dynamics are instructive, the large number of parameters and equations preclude an intuitive perspective of the system. We therefore additionally studied a reduced system [65], which represents the large cortical scale that permits characterization of the system dynamics with respect to the most salient parameters. In contrast to the previous models, the coupling is not through discrete pulses, but by means of smooth sigmoidal rate functions, which embody population-wide neuronal responses to synaptic inputs in the presence of parameter and state dispersion [66]. This also allows us to study the robustness of the resonance-induced synchronization in relationship to the precise details – and dynamical regime – of the models.
Each node represents the mean dynamics of an ensemble of neurons, with spontaneous dynamics arising from the interaction between excitatory and the inhibitory sub-populations. The model is derived from the biophysical Morris-Lecar model [45], extended to a neural mass model with passive diffusive chemical [46], then synaptic interactions [6] and subsequently extended to large networks to model whole brain activity [5]. We utilize this most recent approach developed by Honey et al. [5], [47] systematically varying the features of the connectivity: architecture, coupling strength, and delay.
This neural mass model comprises three state variables: The mean membrane potential of the excitatory pyramidal neurons, ; the mean membrane potential of the inhibitory interneurons, ; and the average number of open potassium ion channels, . Our main focus is on the dynamics of the pyramidal neurons. Their average membrane potential V depends on the passive leak conductance, and on the conductance of voltage-gated channels of sodium, potassium and calcium ions. The flow of current across the local pyramidal cell membranes, assumed as capacitors, governs its dynamics. In turn, the local activity of the inhibitory interneurons is course-grained modeled; its dynamics is modulated by the activity of the pyramidal cell. For each ensemble , the equations for the dynamics of the mean membrane potential of the neurons are given by(16)(17)
The fraction of channels open are the neural-activation function, whose shape reflects a sigmoidal-saturating grow with (18)The third differential equation of each node stands for the fraction of open potassium channels:(19)The neuronal firing rates (, and ) averaged over the ensemble are assumed to obey Gaussian distributions, thereby giving rise to the sigmoidal activation functions [66],(20)(21)
Our simulations employ the previously published parameter values: , , , , , , , , , , , , , , , , , , , , , , , , , , , , , were set to physiological values taken from [6]. These are associated with aperiodic fluctuations arising without external noise, but rather due to homoclinic chaos [6]. Equation 16 includes the other important parameters in our analysis: the presynaptic neighboring (afferent) regions of region ; c, the coupling strength between cortical regions; , the synaptic delay between cortical regions. The model was simulated in Matlab (Math Works) using the function dde23.
|
10.1371/journal.pntd.0001852 | Risk of Potentially Rabid Animal Exposure among Foreign Travelers in Southeast Asia | Each year millions of travelers visit Southeast Asia where rabies is still prevalent. This study aimed to assess the risk of rabies exposure, i.e., by being bitten or licked by an animal, among travelers in Southeast Asia. The secondary objective was to assess their attitudes and practices related to rabies.
Foreign travelers departing to the destination outside Southeast Asia were invited to fill out the study questionnaire in the departure hall of Bangkok International Airport. They were asked about their demographic profile, travel characteristics, pre-travel health preparations, their possible exposure and their practices related to rabies during this trip. From June 2010 to February 2011, 7,681 completed questionnaires were collected. Sixty-two percent of the travelers were male, and the median age was 32 years. 34.0% of the participants were from Western/Central Europe, while 32.1% were from East Asia. Up to 59.3% had sought health information before this trip. Travel clinics were the source of information for 23.6% of travelers. Overall, only 11.6% of the participants had completed their rabies pre-exposure prophylaxis, and 15.3% had received only 1–2 shots, while 73.1% had not been vaccinated at all. In this study, the risk of being bitten was 1.11 per 100 travelers per month and the risk of being licked was 3.12 per 100 travelers per month. Among those who were bitten, only 37.1% went to the hospital to get post exposure treatment. Travelers with East Asian nationalities and longer duration of stay were significantly related to higher risk of animal exposure. Reason for travel was not related to the risk of animal exposure.
Travelers were at risk of being exposed to potentially rabid animals while traveling in Southeast Asia. Many were inadequately informed and unprepared for this life-threatening risk. Rabies prevention advice should be included in every pre-travel visit.
| Rabies is a fatal disease most commonly transmitted through a bite or a lick of a rabid animal on the broken skin. Most deaths from rabies are reported in Asia and Africa where animal rabies is poorly controlled. Not only local people, but travelers in these areas are inevitably at risk also. In this study we surveyed foreign travelers just before they departed Southeast Asia at Bangkok International Airport. We aimed to determine the risk of possible rabies exposure and their attitudes and practices related to rabies. The risk of being bitten among 7,681 participants studied was 1.11 per 100 travelers per month and the risk of being licked was 3.12 per 100 travelers per month. Among those who were bitten, only 37.1% went to the hospital to get rabies post exposure treatment. Travelers with East Asian nationalities and who stay longer were more likely to be exposed to animals. The risk of animal exposure was not related with the reason for travel. These findings confirm that travelers in Southeast Asia were at real risk of possible exposure to rabies. However, most of them were inadequately informed and unprepared for this life-threatening disease. Rabies prevention advice should be given to all travelers in rabies endemic area.
| Rabies remains an important neglected disease worldwide. Approximately 50,000–55,000 people die from rabies each year [1]. Although most deaths are reported among local people in high endemic area especially in Asia and Africa [2], travelers in those areas are inevitably at risk if they are bitten by infected animals or if the saliva of an infected animal comes into contact with broken skin or mucosa.
Pre-exposure vaccination is an excellent preventive measure against rabies among travelers. However, it is not routinely recommended to all travelers in endemic areas. Its high price and cost-effectiveness are often debated as discussed in many papers [3]–[6]. Travel medicine practitioners should consider several factors, including the risk of being bitten or licked during trips, rabies endemicity and the availability of medical care at the travel destination and travelers' preferences before recommending a vaccine. Among those factors, the actual risk of animal exposure is thought to be a major one [5], [7], [8].
Southeast Asia is one of the popular tourist destinations for travelers worldwide. Each year, up to 60 million tourists visit Southeast Asia [9], where rabies is still endemic and stray dogs and cats are common. Information regarding the risk of rabies exposure among travelers in Southeast Asia is limited. Therefore, in this study, we aim to determine the incidence and risk factors of possible exposure to rabies, i.e., by being bitten or licked by animals, during their trips in Southeast Asia. The secondary objective was to assess their pre-travel preparation, vaccination rate, knowledge, and practices related to the risk of rabies.
This cross-sectional questionnaire based study was conducted in Suvarnabhumi International Airport in Bangkok. Data were collected from adult foreign travelers in the international departure hall. Only travelers who had completed their trip and were departing to the destination outside Southeast Asia were eligible to participate. Travelers of Southeast Asian nationalities or travelers who were just in transit were excluded. The study questionnaire was drafted, tested, and revised before the actual data collection. The final version of the questionnaire comprised of four parts, i.e., general information about the travelers, rabies pre-exposure preparations, knowledge about rabies, and the details of any animal exposure. Animal exposure in this study defined as being bitten or licked by mammals that potentially carry rabies virus. In this study, we considered all licked events were at potential risk of rabies exposure, since most travelers were unsure whether their skin was broken. Apart from English, the questionnaire had been translated into 3 more languages: Chinese, Japanese and Korean.
Data from previous studies showed that approximately 0.69–2.2% of travelers were bitten during their one-month stay in Thailand [10], [11]. Therefore, the sample size was calculated based on the assumed incidence of 1% with confidence interval of 0.75%–1.25%, together with the numbers and nationalities of travelers visiting Thailand in 2008 from Thai Immigration Department. To achieve a 95% confidence level, at least 6,081 travelers were required from all regions.
Since the number of travelers from different continents visiting Thailand were not equally distributed and the majority came from Europe and East Asia. To assure the representativeness of travelers from the different continents, quota sampling was implemented. Therefore, the proportions and numbers of participants required from each continent represented the actual annual travel population to Thailand.
During data collection, the investigator team invited any travelers in the departure hall to participate in the study. Eligible travelers who were willing to participate in the study filled out a questionnaire by themselves. The investigator team was available to help if they needed some assistance or clarification of the questionnaire.
The price per one dose of cell-cultured rabies vaccine in each country was obtained from travel medicine specialists through the EuroTravNet network, from personal communication and from other sources. The mean prices for each country or region was adjusted by using the gross domestic product (GDP) per capita, which were obtained from the World Bank. Then, cost index of rabies vaccine for each country could be calculated (mean price/gross domestic product per capita ×104). In this study, rabies vaccination rate was referred to the percentage of travelers who received any rabies pre-exposure vaccines (3 shots or 1–2 shots) over total number of travelers.
Statistical analysis was conducted using SPSS for Windows, version 10.0.7 (SPSS Inc, Chicago, IL) software.
Continuous data were presented as mean with standard deviation (for normally distributed data), or median with range (for non-normally distributed data). Categorical data were presented as numbers and percentage. The Student t-test was used to compare means of two groups, while the Chi-square test was used for categorical data, as appropriate. Relative risk (RR) and 95% Confidence interval were calculated to determine factors potentially associated with animal exposure and receiving pre-exposure vaccination. Factors with a p-value below 0.10 in the univariate models were considered eligible for the multivariate analysis. In this study, a p-value of <0.05 was considered as statistically significant.
The research protocol as well as the questionnaire was approved by the Ethics Committee of the Faculty of Tropical Medicine, Mahidol University (Approval No. MUTM 2010-015-02). Since this study was a voluntary, anonymous survey among adults and was non-experimental in nature; so the Ethics Committee had waived the written consent and approved to imply that filling the questionnaire represent their consent to participate in this study. All participants were informed of the study's objective and grants verbal consent before filling the questionnaires. No participant-identifiable data was recorded in the questionnaire to maintain confidentiality.
During the period from June 2010 to February 2011, 7681 questionnaires were collected and analyzed. The sex ratio of males to females of participants was 1.6 and the median age was 32 years. Approximately one third of the participants were from Western/Central Europe and one-third were from East Asia. The main reason for travel was tourism, followed distantly by business and visiting friends and relatives.
Approximately 60% of participants had sought travel health information before the current trip. The most common sources of information were the internet followed by general practitioners, travel clinics, friends and relatives, guidebooks and pharmacists. Only 12% of travelers had completed a course of pre-exposure rabies vaccinations (3 shots) before travel, 15% had received only 1 or 2 shots, while the majority had not been vaccinated for rabies at all. The complete demographic breakdown is shown in Table 1.
Of the 7,681 travelers studied, 1,809 (23.6%) had received pre-travel health advice from a travel clinic; 56% of the travelers in the travel clinic group had received information about rabies, which was significantly higher than travelers who sought pre-travel health advice from other sources (56.0% vs 37.5%, p<0.001). 21% of travelers in the travel clinic group had completed a course of pre-exposure rabies vaccine while only 8% of travelers in non-travel clinic group had completed their prophylaxis (21.4% vs 8.4%, p<0.001).
When the details of traveler knowledge about rabies was analyzed, it was found that most travelers knew that they could get rabies if bitten by an infected animal and that dogs could carry rabies. However, nearly one out of two travelers was not aware that cats could also carry rabies. Moreover, more than one-fourth of travelers thought that the bite of a healthy-looking dog or cat posed no risk of rabies.
Subgroup analysis also revealed that the travelers who had visited a travel clinic possessed some more specific knowledge items than those who did not visit the clinic including that being licked by an animal poses a risk of contracting rabies. The mean knowledge score for those who visited a travel clinic was significantly higher than the score of those who had not received pre-travel health advice from a travel clinic. The details are shown in Table 2.
Several factors including female sex, older age, longer duration of stay were found to be related with low vaccination rate. The rate of rabies vaccination also differed among travelers from different continents of origin. Travelers from North America or from Oceania had significantly lower vaccination rate when compare to travelers from Western/Central Europe while travelers from South Asia had significantly higher vaccination rate than travelers from Western/Central Europe. Details are shown in Table 3.
The actual cost of rabies vaccine and its cost index, which was adjusted by the GDP per capita, differed significantly from country to country as shown in Table 4. Travelers from countries where the vaccine cost index was <20 (n = 5556) were 1.4 times more likely to receive vaccination against rabies before travel, compared to those from countries where the cost index was > = 20 (n = 2125) (27% vs. 21%, RR 1.43, 95% CI 1.27–1.61).
Of 7,681 participants, sixty-six travelers (0.9%) had been bitten, while 185 travelers (2.4%) had been licked on the average stay of 23.2 days. Virtually all countries in Southeast Asia were reported as countries of exposures where travelers had been exposed to animals. The incidence of animal exposure (bitten or licked) varied from country by country ranging from 0.3% (1/325) among travelers in Malaysia to 3.6% (4/110) among travelers in Myanmar. The overall animal exposure rate in Southeast Asia was 2.8%.
Among those who were bitten, information regarding their actual practice after exposure was available in 35/66 travelers. Base on that data, 3/4 had cleaned the wound, but 2/3 did not seek medical care and did not receive post-exposure treatment. The animals most commonly encountered were dogs, followed by monkeys and cats.
Detail analysis was performed to determine risk factors that might be related to animal exposure. Age, gender, reason for travel and knowledge score had no influenced on animal exposure while the length of stay and continent of origin had some effects. Travelers from East Asia had a higher rate of exposure than Western/Central European (Adjusted RR 2.83, 95%CI 1.87–4.2). Conversely, travelers from South Asia were at lower risk (Adjusted RR 0.20, 95% CI 0.03–0.66). Apart from the nationality of travelers, the length of stay was found to be directly related with the risk of exposure. Travelers who stayed more than 20 days had a higher risk than travelers who stayed less than 5 days (5% vs 1.3%, Adjusted RR 7.78, 95%CI 4.71–13.01). Detailed of the results are show in Tables 5 and 6.
To our knowledge, this was the largest study that aimed to determine the risk of animal exposure among travelers. In our study, the risk of being bitten was 1.11 per 100 travelers per month and the risk of being licked was 3.12 per 100 travelers per month. These incidences were close to the overall estimation of risk published in one recent review. In that review, based on all available evidences [5], [10]–[13], it was estimated that 0.66% (0.02%–2.31%) of tourists will experience animal bite during one month stay [6].
It was not possible to compare our incidence rate directly with all previous studies since there were vast variations in term of the population studied, destination, definition of exposure and so on. However, several important points should be noted. Firstly, the highest incidence of animal exposure had been reported among travelers in Thailand in 1994 airport study. In that report, up to 1.3% of travelers had been bitten during an average stay of 17 days [11]. Compared to the 1994 study, our study found an approximately two-fold decrease in the risk of being bitten (1.1% per month VS 2.2% per month). The lower incidence of animal bite may result from better awareness of rabies among travelers which could by imply from the vaccination rate i.e only 1.1% of travelers in the previous study had received rabies pre-exposure prophylaxis while up to 25% of travelers in our study had received rabies vaccine before their trips.
Apart from risk of animal bite, the endemicity of rabies in the destination is also the major factor that determines the real risk of exposure to rabies virus. Fortunately, data from Thailand showed that local situation of rabies was much improved when compared to the last few decades. For example, the number of human rabies in Thailand cases had decreased from 185 cases per year in 1990 to 78 cases per year in 1994 and to less than 20 cases annually since 2001 [14]. Moreover the percentage of FAT positive animal specimens among those examined for rabies were also decline i.e. from 41% in 1990 to 28% in 2000 and to 12% in 2004 [15]. Several factors were contributed to this success such as the control of stray dogs and cats, vaccination programs in animals, mass campaigns to raise public awareness and better and more accessible post-exposure treatment [3], [14].
However it is important to note that, although the rabies situation in Thailand was much better and the risk of being bitten among travelers seemed to be lower than previous report, this risk was still high when compare to the other studies outside Southeast Asia [5], [13]. Partly, it may be due to the poor control of stray dogs and cats in many countries in Southeast Asia where more than 1 million people are estimated to be bitten annually [16]. Not only local people, but travelers in these areas are inevitably at risk also. Given that rabies is an untreatable disease once the symptoms develop, travelers in rabies endemic areas need a good basic knowledge regarding rabies risk and prevention.
Unfortunately, our study found that, travelers' attitudes and knowledge related to rabies risk were far from ideal. As seen in several previous reports [10], [17], [18], many misconceptions and misunderstandings were found among our participants, such as, up to 59% were not aware that they might get rabies after being licked by an infected animal and 50% did not know that they needed a booster vaccination once they were bitten. These misconceptions were critical and might lead to serious consequences if they actually had been exposed to the rabies virus. In our study, we also confirmed that the travelers' practice after being exposed to animal was poor i.e. one fourth of the responding travelers who were bitten had not cleaned the wound and two third of responding travelers did not go to the hospital to get a rabies vaccination. These were serious and dangerous misunderstanding. Therefore, travelers to rabies endemic areas should receive proper advice regarding rabies before their trip. Travel clinic might be a good source of information as found in several studies [10], [19], [20]. However, in our study, although travelers who had visited a travel clinic had higher mean knowledge scores than those who did not visit the clinic, some misconceptions were also found in comparable percentage between these two groups of travelers.
In this study, the length of stay in Southeast Asia was significantly related to higher rate of animal exposure. Age, gender, and travelers' knowledge, had no significant relationship to rate of animal exposure. Apart from length of stay, multivariate analysis indicated that the nationality of a traveler was related to the risk of animal exposure. Travelers from East Asia had a 2.8-fold higher risk than travelers from Western/Central Europe, while travelers from South Asia had a significantly lower risk. These differences might imply that travelers from different cultures might have different attitudes and different risk behaviors that can be related to a higher or lower risk of animal exposure. For example, travelers from South Asia where rabies was highly endemic might have higher rabies awareness than travelers from Europe, so they were less likely to risk encounter with an animal.
Through the analysis, we also found that the reason for travel was not related to the risk of animal exposure. Hence the magnitude of risk among tourists, businessmen and students in Southeast Asia could be considered the same. This finding might challenge the general belief that the activities of travelers play some role in terms of risk. Although it is logical to assume that, so far there was no available evidence to support this belief, at least in Southeast Asia. This may be in part be due to the fact that stray dogs and cats in Southeast Asia are not restricted to only certain areas, but rather can wander freely around in urban and rural areas. This might explain why, when compared to our recent study done in backpackers in Bangkok [10], the risk of being bitten in the backpacker group was even lower than that in general travelers in this study (0.69 per 100 backpackers per month VS 1.11 per 100 travelers per month). Similar findings were also reported in a study conducted in Nepal, where trekking did not increase the risk of animal exposure [5].
Although many authorities recommend pre-exposure rabies vaccination in high risk travelers [21]–[23], there was no consensus what defines “high risk”. In our study, twenty-seven percent of our participants received rabies vaccine before their trips. Several factors including male sex, younger age, travel for tourism and, surprisingly, a shorter length of stay were found to be correlated to higher vaccination rates. We also found that travelers from countries with a cost index <20 were more likely to receive the vaccine. As in many studies, this was confirmed that cost of the vaccine was an important factor that travelers consider before receiving the pre-exposure vaccines [10], [24], [25].
Our study had some limitations. Although we surveyed more than 7,000 departing travelers from Suvarnabhumi International Airport, which is the main airport hub in Southeast Asia, data from a single airport is not ideal for representing the whole of Southeast Asia. Our data should strongly represent travelers in Thailand and its neighboring countries such as Lao PDR, Cambodia and Vietnam, because most of them use Suvarnabhumi International Airport as a travel hub. But our data may underrepresent people who travel mainly in Indonesia, Singapore and the Philippines, since they may use other airports. Ideally, a multi-airport study could provide more comprehensive data.
Second, the language barrier may have led to selection bias in our study. In this study, apart from English, we translated our questionnaire to 3 different languages i.e. Chinese, Japanese, and Korean. However, the questionnaire were not translated into Arabic, Hindi, Spanish, or any African languages. So those travelers from the Middle East, India, Africa and Latin America, who did not understand English, had to be excluded from the study. It is possible that travelers from those areas who understood English and those who did not may have different risk characteristics.
Third, children, who represent a recognized at-risk population for animal bites and rabies, [1], [2] were not included in our survey, which may have biased the results.
We could conclude that travelers in Southeast Asia, regardless of their reasons for travel, had a significant risk of being bitten or licked by animals while traveling. A longer duration of stay was associated with a higher risk. However, it must be pointed out that 53.8% of travelers with exposure to potential rabies infected animals were actually exposed while traveling for less than 3 weeks. Many were inadequately informed and lacked a basic knowledge of this life-threatening risk. Rabies prevention advice should be included in every pre-travel visit.
|
10.1371/journal.pntd.0007330 | Whole genome sequence of Vibrio cholerae directly from dried spotted filter paper | Global estimates for cholera annually approximate 4 million cases worldwide with 95,000 deaths. Recent outbreaks, including Haiti and Yemen, are reminders that cholera is still a global health concern. Cholera outbreaks can rapidly induce high death tolls by overwhelming the capacity of health facilities, especially in remote areas or areas of civil unrest. Recent studies demonstrated that stool specimens preserved on filter paper facilitate molecular analysis of Vibrio cholerae in resource limited settings. Specimens preserved in a rapid, low-cost, safe and sustainable manner for sequencing provides previously unavailable data about circulating cholera strains. This may ultimately contribute new information to shape public policy response on cholera control and elimination.
Whole genome sequencing (WGS) recovered close to a complete sequence of the V. cholerae O1 genome with satisfactory genome coverage from stool specimens enriched in alkaline peptone water (APW) and V. cholerae culture isolates, both spotted on filter paper. The minimum concentration of V. cholerae DNA sufficient to produce quality genomic information was 0.02 ng/μL. The genomic data confirmed the presence or absence of genes of epidemiological interest, including cholera toxin and pilus loci. WGS identified a variety of diarrheal pathogens from APW-enriched specimen spotted filter paper, highlighting the potential for this technique to explore the gut microbiome, potentially identifying co-infections, which may impact the severity of disease. WGS demonstrated that these specimens fit within the current global cholera phylogenetic tree, identifying the strains as the 7th pandemic El Tor.
WGS results allowed for mapping of short reads from APW-enriched specimen and culture isolate spotted filter papers. This provided valuable molecular epidemiological sequence information on V. cholerae strains from remote, low-resource settings. These results identified the presence of co-infecting pathogens while providing rare insight into the specific V. cholerae strains causing outbreaks in cholera-endemic areas.
| Cholera affects more than 4 million people globally every year; people predominantly living in poverty or in resource-constrained conditions including political crises or natural disasters. Cholera’s typical presentation is characterized by rapid onset of acute watery diarrhea and vomiting which can progress from watery stool to shock in as little as four hours. Laboratory conditions needed for culture confirmation and strain preservation are rarely to never present in these affected areas. In fact, many cholera endemic areas in Sub-Saharan African are so remote that even treatment response alone is often challenging. Here we present the genomic analysis of DNA extracted from dried filter paper, which is a low-cost, low-tech and sustainable method. Previously this method has facilitated cholera confirmation by PCR, but we demonstrate that this method is also suitable for whole genome sequencing and subsequent strain characterization by presenting the analysis of samples from an outbreak in a remote area of Cameroon. This method will facilitate the understanding of the molecular epidemiology in cholera-prone areas, which were previously too challenging to attempt. It also introduces a method that can be used on a broader scale for diarrheal disease surveillance, including providing a window into co-infection and microbiome analyses.
| Global estimates for moderate to severe diarrhea are estimated to account for 1.6 million deaths annually worldwide and total a burden close to 75 million disability-adjusted life years (DALY) with costs approximating 3.11 billion USD in 2010 [1]. Recent studies have demonstrated the value of whole genome sequencing (WGS) in understanding the molecular evolution and transmission of the etiologic agents that cause moderate to severe diarrhea, including Vibrio cholerae. An analysis of > 700 V. cholerae isolate genomes originating from Asia, Africa, Latin America and the Caribbean spanning a period of more than half a century demonstrated that epidemics of cholera in Africa and the Americas stem from the introduction of a single pandemic lineage from Africa and South Asia [2]. This understanding of cholera was only possible because WGS data provided phylogenetically robust measures of relatedness. This analysis revealed that the currently circulating seventh pandemic of cholera can largely be attributed to a closely related genetic sublineage known as the seventh pandemic El Tor lineage (7PET) that forms a single branch within the diverse V. cholerae species phylogeny.
Over the past 20 years, WGS has improved our understanding of the molecular evolution and transmission of the etiologic agents that cause moderate to severe diarrhea, including V. cholerae [3]. Molecular epidemiology has become critical to determine the areas at highest risk for infections, to improve intervention measures such as vaccination campaigns but also to inform the development of new vaccines targeting the appropriate species and genotypes [4–6]. In addition, molecular epidemiological data also provides the information needed to combat the growing threats posed by antimicrobial resistance (AMR) spread [7]. Molecular epidemiology has played an important role in recent years for cholera. The OCV stockpile was created in 2013 and the need to target the doses available to those at greatest risk is important due to limits in vaccine availability[8–10]. Cholera often strikes in remote or resource-poor settings where laboratory capacity is limited, and therefore, it is not always possible to culture specimens from stool. Moreover, the time necessary from specimen collection to isolate preservation can take up to five days, if culture capability is available on site, and up to several weeks if specimens have to be stored on Cary Blair and subsequently transported to a central facility. In addition, specimen storage facilities are not available in every laboratory, and biosafety shipping of infectious strains to specialized genotyping laboratories may be very challenging from cholera-affected countries. Hence, the need for alternative approaches is critical to facilitate access to genomic information from samples collected in all cholera outbreak areas, particularly remote and resource-poor areas.
The primary analysis in this paper is based upon samples collected during a cholera surveillance and response program in the remote area of the Far North Region in Cameroon (FNC), which established and validated the use of dried filter paper for stool specimen preservation [11]. During an outbreak in the fall of 2014, cholera cases were primarily reported from an island in Lake Chad where specimen preservation is generally not possible due to lack of laboratory capacity and lack of cold chain transport. After prior stool enrichment in alkaline peptone water (APW), the enriched specimens were either directly spotted onto filter paper (APW-enriched specimen spotted filter paper) or cultured overnight, from which a single colony was picked and spotted onto filter paper (culture isolate spotted filter paper). The dried spot methodology, with or without prior culture step, was instrumental in providing the ability to preserve genomic material in order to characterize the outbreak. The study demonstrated that dried specimen spotted filter papers can be stored for up to 2 years at room temperature prior to DNA extraction and PCR amplification [12]. Once extracted, the DNA was analysed using multi-locus variable-number tandem repeat analysis (MLVA) and demonstrated that advanced molecular DNA methods could be used on dried specimen spotted filter papers preserved samples [12]. A similar study was performed in 2012 during a cholera epidemic in Sierra Leone, where watery diarrhea was directly sampled on filter paper without prior enrichment in APW, and stored for nearly three years at room temperature before successful MLVA genotyping [13]. Given the limitations of MLVA in providing detailed high-resolution molecular epidemiological features, we sought to determine if the same spotted filter paper material could also be used to perform WGS. This is a proof of principle study demonstrating that DNA extracted from simple, low-cost, APW-enriched and culture isolate spotted filter paper can generate high-quality accurate sequence data that has the potential to inform public health decisions by providing essential information on cholera genotype and co-infection burden.
APW-enriched specimen and culture isolate spotted filter papers included in the study were isolated from participants enrolled in the “Sustainable Cholera Surveillance for Cameroon” project. The Johns Hopkins Bloomberg School of Public Health Institutional Review Board reviewed and approved this study, IRB No. IRB00003981. Written informed consent was obtained from each participant or their caretaker prior to initiation of study activities.
The surveillance methodology, specimen collection, laboratory testing and findings have been previously reported [11,12]. Sixty-five isolates from two distinct outbreaks in the Far North of Cameroon were collected during this time.
Cameroonian stool specimens from 65 patients tested positive for V. cholerae by Crystal VCTM dipstick kit (Arkray Healthcare Pvt Ltd.,Surat, India) and subsequently culture-confirmed. Of these specimens, only 16 were processed according to two different protocols, called hereafter APW-enriched specimen spotted filter paper and culture isolate spotted filter paper.
The specimens referred to as APW-enriched specimen spotted filter papers are derived from stool samples enriched for 6-hours in 1X alkaline peptone water (APW) solution at room temperature. Two drops of the APW-enriched stool specimen were aliquoted onto two circles of a Whatman 903 filter paper card for preservation [12].
The specimens referred to as culture isolate spotted filter paper refers to the same 16 Cameroonian patients’ specimens from which an isolate was able to be cultured. Briefly, the APW-enriched stool specimen was transferred via Cary Blair media to the main health facility for microbiological culture. The specimen was streaked onto TCBS medium overnight at 37°C. From each TCBS culture, a single colony was selected, diluted in 50μL of phosphate-buffered saline (PBS) and aliquoted onto filter paper.
To evaluate spotted filter paper as a new sample preservation method, the viability of V. cholerae after drying on filter paper blots was tested. V. cholerae serogroup O1 was grown in liquid culture to a confluence of 1x108CFU/mL; 50μL of bacterial suspension was aliquoted onto Whatman filter paper and allowed to air dry overnight at room temperature (17h). Simultaneously, bacterial suspensions were aliquoted into 4 tubes for Vibrio viability experiments: heat-killing, ethanol-killing, bleach-killing and UV-light irradiation were evaluated for their sterilisation potential. After timed incubations with each potential killing agent, the bacterial suspension was spotted onto filter papers and allowed to air-dry overnight. A dried spot was excised from each filter paper and subsequently incubated in APW for 6 hours at 37°C. Following enrichment, specimens were tested via Crystal VC dipstick to assess for any V. cholerae growth. Concurrently, each specimen was streaked on both TCBS and Luria Agar (LA) plates for overnight culture.
For APW-enriched specimen and culture isolate spotted filter papers, a single spot of filter paper was excised at Johns Hopkins facilities, using scissors and inserted into a micro-centrifuge tube, washed twice with 1ml 1X PBS, and boiled with 200μL 1.5% Chelex solution for eight minutes. After a 1-minute centrifugation, the supernatant was transferred to a sterile micro-centrifuge tube. The presence of V. cholerae was confirmed via multiplex PCRs, first targeting an outer membrane protein, OmpW, in combination with primers targeting cholera toxin A, ctxA. A second PCR confirmed the presence of the rfb gene specific for the O1 serogroup following previously described methods [14,15]. To optimize for WGS following Chelex extraction, the DNA extracted from culture isolate spotted filter papers was subsequently purified by ethanol precipitation as described by Sambrook et al. [16]. The DNA was resuspended with ddH2O and sent for quantification by Qubit 2.0 Fluorometer (Thermofisher) and qPCR (StepOnePlus Real-Time PCR System) before WGS. Only samples with greater than a 0.001 ng/μLμl concentration of V. cholerae DNA were submitted for WGS.
WGS was performed at the Wellcome Trust Sanger Institute on an Illumina HiSeq 2500 platform to generate 100 bp paired-end reads. Short read data are available in the European Nucleotide Archive (ENA) database (S3 Table).
Sequence reads were mapped against reference genome V. cholerae O1 El Tor reference N16961 (accession numbers LT907989/LT907990) using SMALT v0.7.4 [17]. SMALT was used to index the reference using a k-mer size of 13 and a step size of 6 (-k 13 -s 6) and the reads were aligned using default parameters but with the maximum insert size (i) set as 3 times the mean fragment size of the sequencing library. PCR duplicate reads were identified using Picard v1.92 [18] and flagged as duplicates. A reference-based alignment was obtained by mapping paired-end Illumina reads to DNA sequenced from APW-enriched specimen spotted filter papers and isolate spotted filter papers to the V. cholerae O1 El Tor reference N16961. Automated annotation was performed using PROKKA v1.11 [19] and a genus-specific database from RefSeq [20].
Variation detection was performed using SamtoolsMpileup v0.1.19 [21] with parameters “-d 1000 -DSugBf” and bcftools v0.1.19 [22] to produce a BCF file of all variant sites. All bases were filtered to remove those with uncertainty in the base call. The bcftools variant quality score was required to be greater than 50 and the mapping quality to be greater than 30. If all reads did not give the same base call, the allele frequency, as calculated by bcftools, was required to be either 0 for bases called the same as the reference, or 1 for bases called as a single nucleotide polymorphism (SNP) (af1 < 0.95). The majority base call was required to be present in at least 75% of reads mapping at the base, (ratio < 0.75), and the minimum mapping depth required was 4 reads, at least two of which had to map to each strand (depth < 4, depth_strand< 2). Finally, strand bias was required to be less than 0.001, map bias less than 0.001 and tail bias less than 0.001. If any of these filters were not met, the base was called as uncertain. A pseudo-genome was constructed by substituting the base call at each site (variant and non-variant) in the BCF file into the reference genome and any site called as uncertain was substituted with an N. Insertions with respect to the reference genome were ignored and deletions with respect to the reference genome were filled with N’s in the pseudo-genome to keep it aligned and the same length as the reference genome used for read mapping. Mapping was visualised with Artemis [23] and ACT [24].
Short reads from the Cameroonian samples were assembled de novo using SPAdes v3.10.0 [25], reordered against the reference sequence with ABACAS [26], and then a metaSPAdes [27] was performed with parameters—meta -t 8 -m 15. Statistics from assemblies were extracted with metaQUAST with parameters—fast -no-check [28]. Genome completeness estimates and checks for contamination were performed using CheckM lineage wf with the following parameters -t 8 -x fa—reduced tree [29]. Kraken v0.10.6 [30] was used to assign taxonomic labels using default parameters and the database Refseq release 72 (27/08/2015). Annotation was performed using the RAST server [31].
A genome distance matrix was obtained by using MASH on SPAdes assemblies as previously described [32][5]. To accurately place our samples into a phylogenetic context, we supplemented our analyses with previously published genomes taken from Weill et al [2]. A Neighbor-Joining tree was generated based on the distance matrix using MASHtree [33]. An outgroup composed of M66, CNRVC960188 and CNRVC961190 was used to reroot the tree using Figtree. The resulting phylogenetic tree and corresponding metadata were visualized using Microreact [34].
Viability tests demonstrated that there was no viable V. cholerae after drying spotted filter paper overnight (17h), Viability was negative on all specimens evaluated via overnight culture on both TCBS and Luria Agar. This demonstrated that the specimens were no longer infectious for V. cholerae allowing safe shipping regarding biological risks.
Sixteen V. cholerae O1 positive specimen pairs were included in this study, comparing APW-enriched specimen spotted filter paper to culture isolate spotted filter paper, both derived from the same original stool specimen. As measured by RT PCR, total DNA concentration was on average 3 times higher when recovered from APW-enriched specimen spotted filter paper than from culture isolate spotted filter paper. Conversely the concentration of V. cholerae specific DNA was nearly 2.5 times higher from the culture isolate spotted filter paper compared to the APW-enriched specimen spotted on filter paper, as per the median and the mean reported in Table 1. In both cases, the quantity of DNA was sufficient to perform WGS (Table 1). Only two APW-enriched specimen spotted samples, 600064 and 600068, displayed higher V. cholerae DNA concentration than their respective culture isolate counterpart.
The mapping coverage for all samples ranged from 8.8x to 500x with an average of 128.4x (Fig 1A). The percentage of V. cholerae N16961 reference genome covered by reads ranged from 19.71% to 98.33% with an average of 68.44% (Fig 1B).
In both APW-enriched specimen spotted filter papers and culture isolate spotted filter papers, Spearman correlation test showed a positive correlation between the quantity of V. cholerae DNA and the percentage of V. cholerae genome covered by short reads in DNA extracted from both APW-enriched and culture isolate spotted filter papers (Spearman correlation score = 0.42, p < 0.1) (S2 Fig). A minimum concentration of 0.02 ng/μL of V. cholerae specific DNA in the sample generated more than 75% coverage when mapped against the reference genome. 43% of all spotted filter papers that showed successful mapping contained a concentration of V. cholerae DNA greater than 0.02 ng/μL. The percentage of the reference genome mapped appeared to plateau between a concentration of 0.2 and 0.3 ng/μL.
When comparing the mapping data according to the sample preparation protocol used, APW-enriched specimen spotted filter papers generated slightly higher mapping quality scores, specifically 148.7 ±144.03 mean depth and 70.52% ±26.81 reference genome covered compared to 105.86 ±81.4 mean depth and 66.14% ±25.52 reference genome covered for isolate spotted filter papers (Fig 1). The two protocols, APW-enriched specimen and culture isolate spotted filter paper were not statistically significant (Wilcoxon rank test, p-value = 0.5518 and p-value = 0.1965 for mean depth and reference genome covered respectively).
De novo assemblies using SPAdes produced assemblies with < 1000 contigs for 13 out of 16 isolate DNA samples and 6 out of the 16 APW-enriched specimen spotted filter papers. The best assembly with less than 100 contigs including some contigs larger than 50000bp and covering more than 97% of the genome was obtained from DNA extracted from the APW- enriched specimen spotted filter paper for specimen 600057 (Table S3).
APW-enriched specimen spotted filter paper exhibited larger contigs than culture isolate spotted filter paper assemblies (Fig 2A). The range in results was higher for the APW-enriched specimen spotted filter papers compared to the culture isolate spotted filter paper, as illustrated by the wide range of reference genome fraction covered by the assemblies of APW-enriched specimen spotted filter papers, varying from 0.007% to 97% (Fig 2B).
MetaSPAde generated assemblies for 15 of the 16 APW-enriched specimen spotted filter papers and 16 out of 16 culture isolate spotted filter papers, while SPAde only produced 24 assemblies out of 32 samples. One APW-enriched specimen sample, 500291, did not produce an assembly from either MetaSPAde analysis or SPAde analysis, likely due to limitations in the quantity of cholera-specific DNA available in the specimen. MetaSPAde analysis of APW-enriched specimen spotted filter papers allowed for the identification of the V. cholerae genome through sequence assemblies in a higher proportion of samples compared to culture isolate spotted filter papers.
Preliminary species diversity analyses of both APW-enriched specimen spotted filter papers and culture isolate spotted filter papers showed that V. cholerae reads represented one of the most abundant species in the majority of the samples (S3 Fig). Metagenomic analysis of APW-enriched specimen spotted filter paper showed the presence of a diverse microbial population (S3 Fig). As an example, bacteria belonging to the Enterobacteriaceae family such as various Shigella species or Escherichia coli strains, in addition to V. cholerae infection were identified in APW-enriched spotted filter papers. The use of MetaSPAde and MetaQUAST software allowed us to determine with higher accuracy the specific contribution of V. cholerae genome to the assemblies as well as the extent of bacterial diversity found in these samples (Fig 3). As expected, genomic diversity appears higher in enriched specimen spotted filter papers compared to culture isolate spotted filter papers. Similarly, V. cholerae genome is present in higher proportion in enriched specimen spotted filter papers compared to culture isolate spotted filter papers, which can be easily explained due to the lower quantity of DNA present in the isolate spotted filter papers (Fig 3).
Of the 32 APW-enriched specimen and culture isolate spotted filter papers samples sequenced, 20 failed to provide coverage above 50% or generate assemblies of greater than 1000 contigs. Therefore, these samples did not meet the criteria for further analysis and were excluded. Assemblies from both the APW-enriched specimen and culture isolate spotted filter papers samples of the same patient were simultaneously aligned with the V. cholerae reference genome N16961. Alignment comparison demonstrated that assemblies of all APW-enriched specimen spotted filter papers contained contigs outside of the V. cholerae genome and include sequences that show a high degree of similarity with other bacterial genomes (Fig 3).
The substantial level of coverage facilitated the alignment of short reads to V. cholerae O1 reference genome, confirming the molecular epidemiological characterization of these strains as V. cholerae O1. Further, several biologically relevant genes and genomic features of the V. cholerae genome could be identified in APW-enriched specimen spotted filter papers as well as in culture isolate spotted filter papers. Such examples of genes are ctxA and ctxB cholera toxin-encoded genes embedded into the integrated CTXΦ prophage; and tcpA, a gene of the Vibrio pathogenicity island. The strains were not part of O139 serogroup, demonstrated by the absence of genes including rstA, or wfbA of the rfb region (Fig 1C and S4 Fig)[35–38][35–38][34–37].
Variant calling was performed following SMALT mapping in order to identify single nucleotide variant sites (SNV) and sequence diversity within the V. cholerae genomes of each sample. Variant calling was not performed in samples with a DNA concentration below 0.02 ng/μL (S5A Fig).
Compiling all analyses, the highest quality samples were selected based on assembly criteria such as V. cholerae genome fraction covered (> 50%), number of contigs (> 100), largest contig (> 5000), total assembly length (> 2.2Mb) and NG50 > 500 (S3 Table). Genome distance subsequently estimated using MASH on the eight best spade assemblies of this study were analysed in the context of sequences published by Weill et al. representing V. cholerae samples spanning over the past century [2]. This data was clustered using a Neighbour-Joining tree demonstrating the characteristic waves reflective of the global phylogeny of the 7th pandemic as shown in (Fig 4) and https://microreact.org/project/S1OfV91PG. The phylogenetic analysis confirmed the affiliation of the Cameroonian V. cholerae strains extracted from APW-enriched and culture isolate spotted filter papers to the 7th pandemic El Tor. Importantly, the samples fit within the third wave of the global phylogenetic tree of V. cholerae, in close proximity to other Cameroonian samples of recent years such as those dated from 2005, and from 2010 and 2011.
Based on these observations, we have concluded that the following quality criteria need to be fulfilled for correct phylogenetic interpretation, namely a proportion of reference genome covering greater than 50%, a mean depth greater than 20x, V. cholerae DNA concentration greater than 0.02 ng/μL, and examination of assemblies (NG50).
In this study, we successfully sequenced DNA from two types of samples spotted onto filter papers for preservation. Not only were these preservation methods proven safe but effective for WGS quality standards after collection, storage and transport.
DNA was recovered from all Cameroonian spotted filter papers and WGS proved to be successful for all samples despite low quantity of DNA recovered. WGS results allowed for mapping short reads for the majority of APW-enriched specimen spotted filter papers and all but one of the culture isolate spotted filter papers. Despite high heterogeneity, the quality of the mapping for APW-enriched specimen spotted filter papers, when compared to culture isolate spotted filter papers, proved to be of variable but satisfactory quality. Quality was illustrated by several criteria including mean depth, proportion of reference genome covered, DNA concentration, and NG50. Mapping confirmed the identification of specific virulence genes and the absence of genes implicated in important biological pathways of V. cholerae, providing critical molecular epidemiological information to characterize cholera outbreaks in remote and/or unstable areas [39–41][38–40][39–41][35–37]. Furthermore, successful assemblies obtained from WGS of these samples were instrumental in identifying gene context and gene organisation within reconstituted genomes. Analyses suggested that the use of APW enrichment of stool rather than the more laborious selection of isolate spotted filter paper might be more efficient for reconstitution of V. cholerae assemblies. This data provides a source of information to develop informed experimental hypotheses that may reveal new biological mechanisms of V. cholerae bacteria.
The use of metagenomics software tools showed bacterial diversity in V. cholerae infected samples and highlighted the prospect for using spotted filter paper for routine metagenomics analysis. This finding highlights the presence of co-occurrences of potential gut bacterial colonisation or co-infection with other diarrheal pathogens. Novel biological interaction mechanisms may also be explored at the bacterial population level, such as the complexity of the gut microbiota in cholera infected patients [42]. Extracting DNA from APW-enriched specimen spotted filter papers revealed the potential for studying multiple bacterial populations through WGS. The ability to study the diversity of bacterial populations from spotted filter papers will facilitate the study and understanding of the microbiome in low-resource settings, not only as it pertains to cholera.
Mapping based on quality criteria such as the proportion of reference genome covered, mean depth, original concentration of DNA, and high-quality assembly criteria such as NG50, number and size of contigs allowed us to restrict our analysis to high quality assemblies only. This high-quality assembly data could be used to understand the genetic distance between V. cholerae strains and place the analyzed sample within a general phylogenetic context in the global history of cholera transmission. Pairwise mutation distance-based clustering results confirmed the low level of diversity expected in V. cholerae clinical samples of an epidemic outbreak. Phylogeny is a critical tool that has been proven to contribute to characterizing outbreaks and to provide evidence for global and local transmission. This tool will be of specific value in remote and resource-constrained settings such as regions where cholera is endemic or regions with elevated risk of cholera epidemics. It will facilitate testing and verification of experimental hypotheses related to the biology of V. cholerae in controlled laboratory settings where the opportunity may not be otherwise possible. This is a result of difficulties in specimen processing and preservation in remote and austere settings where cholera is often endemic. However, with the increasing affordability of sequencing and the recent development of affordable and compact sequencing technologies, such as ISeq and Oxford Nanopore MinION, access to these technologies in countries at high-risk for cholera is increasing. Together the use of dried specimens in combination with more affordable resources in country may facilitate informed decision-making for a timely response to cholera outbreaks in remote and low-resource areas.
There are several areas of improvement to be considered in future studies. First, since sequencing was not the original intention of the specimen preservation, we did not preserve non-enriched/direct stool samples on the filter paper. Currently, we are working to collect crude, APW-enriched specimen and culture isolate spotted filter papers in tandem to facilitate the comparison of DNA quality as well as sequence results across all potential specimen preservation types to optimize the method most applicable in a low resource setting. Second, the DNA extraction method is an important limitation in the comparison of APW-enriched versus isolate filter paper sequences in this paper. The use of ethanol precipitation likely greatly reduced the final DNA concentration available for sequencing from the isolate spotted filter paper, therefore a direct extraction comparison is warranted. Efforts are currently underway to actively improve the quality and quantity of DNA extracted from filter paper cards through protocol refinement for all specimen types. Subsequent to this study, we have also employed duplicated spotting of specimens at all of our study sites as it has shown to be advantageous providing additional specimen available at minimal cost for these advanced molecular studies. Finally, the wide array of filter paper technologies available will present options for further consideration to determine the specific type of filter paper best for use in subsequent work to optimize DNA preservation on filter paper.
In conclusion, we present a proof of concept for WGS of DNA extracted from APW-enriched specimen and culture isolate spotted filter papers specifically targeted, but not limited, to V. cholerae strains preserved on dried filter papers. We have determined the minimum methodological requirements allowing for successful WGS that would allow for the retrieval of biologically relevant genomic information. Until sequencing is widely accessible and affordable, the optimization of this method provides high-level molecular information at low cost and limited difficulty to countries at-risk. In conjunction with new sequencing technologies that may soon be available in low-resource settings, we may soon understand transmission patterns in-real time rather than post-outbreak characterization. The optimization of filter paper preservation for WGS will pave the way towards a better understanding of V. cholerae transmission and outbreak dynamics globally.
|
10.1371/journal.pntd.0007533 | Leishmania major degrades murine CXCL1 – An immune evasion strategy | Leishmaniasis is a global health problem with an estimated report of 2 million new cases every year and more than 1 billion people at risk of contracting this disease in endemic areas. The innate immune system plays a central role in controlling L. major infection by initiating a signaling cascade that results in production of pro-inflammatory cytokines and recruitment of both innate and adaptive immune cells. Upon infection with L. major, CXCL1 is produced locally and plays an important role in the recruitment of neutrophils to the site of infection. Herein, we report that L. major specifically targets murine CXCL1 for degradation. The degradation of CXCL1 is not dependent on host factors as L. major can directly degrade recombinant CXCL1 in a cell-free system. Using mass spectrometry, we discovered that the L. major protease cleaves at the C-terminal end of murine CXCL1. Finally, our data suggest that L. major metalloproteases are involved in the direct cleavage and degradation of CXCL1, and a synthetic peptide spanning the CXCL1 cleavage site can be used to inhibit L. major metalloprotease activity. In conclusion, our study has identified an immune evasion strategy employed by L. major to evade innate immune responses in mice, likely reservoirs in the endemic areas, and further highlights that targeting these L. major metalloproteases may be important in controlling infection within the reservoir population and transmittance of the disease.
| Our study discovered a highly specific role for L. major metalloprotease in cleaving and degrading murine CXCL1. Indeed, L. major metalloprotease did not cleave murine CXCL2 or human CXCL1, CXCL2 and CXCL8. CXCL1 is a critical chemokine required for neutrophil recruitment to the site of infection; thus, we propose that this metalloprotease may have evolved to evade immune responses specifically in the murine host. We have further identified that the C-terminal end on CXCL1 is targeted for cleavage by the L. major metalloprotease. Finally, this cleavage site information was used to design peptides that are able to inhibit CXCL1 degradation by L. major. Our study highlights an immune evasion strategy utilized by L. major to establish infection within a murine host.
| Leishmania spp. are unicellular eukaryotic protozoan parasites that are transmitted to mammalian hosts by sandfly (Phlebotomine and Lutzomyia spp.) bites [1]. Upon transmission of L. major promastigotes (the infectious stage for mammalian hosts with a long slender body and an anterior flagellin), the promastigotes are quickly taken up by neutrophils, macrophages and keratinocytes [2–6]. Within the macrophages, Leishmania spp. promastigotes hijack the phagocytic vacuole and transform into amastigotes (round body lacking an anterior flagellin) [7]. The Leishmania spp. amastigotes then proliferate within the vacuole and establish infection within the host [8, 9]. While a mammalian host-vector system is the major mode of Leishmania spp. transmission, several studies have reported a vertical transmission of these parasites in mammalian hosts, from a pregnant female to its offspring [10–13]. Specifically, Leishmania spp. infection has been found to be endemic in foxhound dog populations in the United States, where the vectors are not present [14–17]. Given that Leishmania spp. can infect several hosts, including rodents and dogs (in addition to humans) [18], these studies demonstrate how these parasites can remain endemic even with strategies to eradicate sandflies.
Our immune system is extremely efficient in killing pathogens. Professional phagocytes such as macrophages and neutrophils phagocytose and kill the invading pathogen in the intracellular phagosome-lysosome compartment. In addition, these phagocytes also respond to the foreign pathogens by secreting pro-inflammatory cytokines and chemokines to recruit neutrophils to the site of infection [19–21]. Leishmania spp. have evolved to evade the host immune response by using its armada of virulence factors to avoid host killing [22]. The two major virulence factors of Leishmania spp. include leishmanolysin metalloprotease glycoprotein 63 (GP63) and lipophosphoglycan (LPG), and both have been extensively studied for their roles in immune evasion [23, 24]. GP63 and LPG inhibit the formation of the membrane attack complex to evade complement-mediated lysis [25, 26], inhibit acidification of leishmania containing vacuoles [27, 28], and dampen host immune signaling pathways [29, 30] for establishing infection within the mammalian host. In addition, Leishmania spp. can hijack the host immune responses to establish infections as shown by Leishmania chagasi (L. chagasi)-mediated activation of TGF-β and Leishmania major (L. major)-induced activation of NLRP3 inflammasome, events that promote L. major survival and pathology [31–33]. Here, we have identified one such immune evasion mechanism employed by L. major.
L. major is the most common Leishmania spp. and a major cause of cutaneous leishmaniasis that affects 600,000–1,000,000 people globally [34]. Once deposited at the site of a sandfly bite, L. major promastigotes first come in contact with keratinocytes, which then secrete pro-inflammatory cytokines to recruit innate immune cells [3, 35]. Keratinocyte do not play a major role in phagocytizing L. major as less than 1–2% of human keratinocyte have been shown to internalize L. major in vitro; which are often non-productive infections [3, 35]. L. major promastigotes are taken up by the resident macrophages, which then secrete essential cytokines and chemokines to elicit an innate immune response [36]. One of the earliest chemokines that are produced in the skin in response to L. major include CXCL1 [37]. CXCL1 is a functional homolog of human IL-8 in mice and a potent neutrophil chemoattractant [38–42]. Keratinocytes [3], neutrophils [43] and macrophages [44, 45] have all been suggested to produce CXCL1 (or IL-8 in humans) in response to Leishmania spp. infection. Several studies have shown that neutrophils infiltrate the site of Leishmania spp. infection as early as 1-hour post-infection and are important for optimal resolution of the infection [21, 46]. Given the importance of CXCL1 in modulating the early innate immune response, we reasoned that L. major targets CXCL1 to evade early host innate immune responses.
In this current study, we have identified murine CXCL1 as a highly specific substrate for L. major metalloprotease and a possible immune evasion strategy employed by this parasite to establish a successful infection within the murine host. We further report that L. major promotes proteolytic cleavage of murine CXCL1 at the C-terminal end to initiate its degradation. Finally, we have designed a synthetic peptide spanning the CXCL1 cleavage site that inhibits L. major protease activity. In conclusion, our study has uncovered a specific mechanism employed by L. major to degrade murine CXCL1 which may help the parasite to establish infection within the murine host.
To investigate how L. major impacts innate immune responses elicited by macrophages, we stimulated bone marrow-derived macrophages (BMDM) with lipopolysaccharide (LPS) in the presence or absence of L. major (WHOM/IR/-173) infection following established protocol [32]. As detailed in the experimental outline (Fig 1A), BMDM were stimulated with 20 ng/ml LPS in the presence or absence of 20 MOI L. major promastigotes for 48 hours. LPS stimulation of BMDM induced robust production of IL-6, TNF and CXCL1. Simultaneous LPS stimulation and L. major infection did not impact the production of IL-6 and TNF by BMDM whereas CXCL1 levels were significantly blunted (Fig 1A). The reduction of CXCL1 was directly proportional to the concentration of L. major as demonstrated by L. major concentration dependent reduction in the levels of CXCL1 from LPS-stimulated BMDM (S1A Fig). These data suggest that L. major specifically targets LPS-induced CXCL1 production by BMDM.
Previous studies have shown that L. major targets signaling pathways to evade immune responses and establish infection [[29]]. To examine whether L. major specifically targeted CXCL1 expression and production, we designed an experiment whereby BMDM were infected with 20 MOI L. major for 6 hours, extracellular L. major were washed, then stimulated with 20 ng/ml LPS for the next 42 hours (Fig 1B). Interestingly, L. major-infected BMDM produced equal levels of CXCL1 when compared to controls (Fig 1B). These results show that inhibition of CXCL1 by L. major may not occur through modulation of intracellular signaling pathway that promotes CXCL1 expression and/or production by BMDM. L. major promastigotes once inside the macrophages transform into amastigotes. To this end, we tested the ability of lesional amastigotes to directly modulate rm-CXCL1 detection. Our results show that lesional L. major amastigotes prepared from L. major-infected footpads were able to inhibit rm-CXCL1 detection (S1B Fig). Altogether, these results suggest that both extracellular (promastigote) and intracellular (amastigote) forms of L. major are able to modulate CXCL1 levels.
Given that L. major pre-infection of BMDM prior to LPS stimulation did not affect CXCL1 production, we posited that L. major regulates secreted CXCL1 in the extracellular milieu. To this end, we used supernatants derived from LPS-stimulated BMDM as a source of CXCL1 and cultured 500 μl of these cell-free supernatants with 20 × 106 L. major promastigotes for 24 hours (Fig 2A). While levels of IL-6 and TNF detected in the supernatants remained unchanged, levels of CXCL1 were significantly reduced after addition of L. major to the supernatants. The reduced levels of CXCL1 were not due to its short half-life in the culture, as the levels of CXCL1 in the cell free supernatants were stable up to 48 hours (S1C Fig). Thus, L. major directly dampens CXCL1 detection.
We next examined whether the observed effect of L. major on CXCL1 detection was dependent on live parasites. To this end, we cultured supernatants from LPS-stimulated BMDM with L. major lysates (Lm lysate: generated by 3× freeze-thaw cycles of L. major) or supernatants from L. major culture (Lm sup: supernatant collected from stationary phase of L. major growth; Fig 2B). Similar to the addition of live L. major, addition of Lm lysate or Lm sup also reduced CXCL1 in the supernatants while IL-6 and TNF remained unaffected (Fig 2B). These results demonstrate that: first, L. major need not be alive to dampen CXCL1 detection; second, L. major lysate can dampen CXCL1 detection; and finally, L. major secreted factors dampen CXCL1 detection. Boiling Lm lysate or Lm sup for 20 minutes at 100ºC rescued CXCL1 detected in the supernatants, suggesting that the responsible L. major components are susceptible to heat treatment (S2 Fig). Given the sensitivity to heat treatment, we propose that the CXCL1 regulating L. major components are proteinaceous.
Macrophages secrete several hundreds of different proteins that may indirectly alter our observed effect of L. major components on CXCL1 [47]. To this end, we obtained recombinant murine CXCL1 (rm-CXCL1, Tonbo Biosciences, San Diego, CA) which was stable in culture up to 48 hours (S3 Fig). Importantly, Lm sup addition to rm-CXCL1 dampened its detection by ELISA in a time-dependent manner, and boiling Lm sup rescued rm-CXCL1 detection (S3 Fig).
To further investigate the specificity of L. major in reducing rm-CXCL1 levels, we examined rm-CXCL2, rh-CXCL1, rh-CXCL2 and rh-CXCL8 sequence homology by CLUSTAL-W alignment and found significant homology between these recombinant proteins (Fig 3A). Despite the significant homology, L. major failed to reduce levels of rm-CXCL2, rh-CXCL1, rh-CXCL2 or rh-CXCL8, demonstrating specific regulation of rm-CXCL1 (Fig 3B–3F). As expected, L. major did not inhibit rm-TNF levels (Fig 3G). While the C-terminal cleavage site of murine CXCL1 has high degree of sequence homology to murine CXCL2 and human CXCL1, CXCL2 and CXCL8; L. major fails to cleave murine CXCL2 or human CXCL1, CXCL2 and CXCL8. Several possibilities that may explain these discrepancies could be: 1) a requirement of specific sequence for Lm protease that is only present in murine CXCL1, and 2) the three-dimensional structure of murine CXCL1, specifically at the c-terminal end, is different from that of other CXCL homologs and orthologs. Given the specific nature of L. major (WHOM/IR/-170) in dampening rm-CXCL1 levels, we next determined whether this activity was specific to the WHOM/IR/-170 isolate. However, supernatant from L. major (IA0, isolated from a patient in Iowa who acquired L. major in Iraq [48]) was also able to inhibit rm-CXCL1 detection when compared to the WHOM/IR/-170 strain (Fig 3H). Furthermore, exosomes from the L. major supernatants from both isolates were able to reduce rm-CXCL1 levels, suggesting that the active components are also present in the L. major-derived vesicles (Fig 3H).
Our data show that L. major reduced CXCL1 levels as demonstrated by ELISA (Figs 1–3). Possible reasons why CXCL1 levels were reduced when incubated with L. major include: 1) L. major proteins bind to CXCL1, limiting the ability of anti-CXCL1 antibody in ELISA to interact with CXCL1 or 2) L. major proteases cleaves and degrades CXCL1. Considering both outcomes, it is possible that CXCL1, either masked or cleaved by L. major components, could still be biologically active. To this end, we performed a functional assay where rm-CXCL1 or rm-CXCL1+Lm lysate was used to stimulate BMDM (Fig 4A). As demonstrated previously, when rm-CXCL1 is incubated with L. major, rm-CXCL1 levels are significantly reduced (Fig 4B). Lm lysate alone has very little stimulatory activity as demonstrated by mRNA induction of Cxcl1, Tnf and Il6 (Fig 4C–4E). As expected, rm-CXCL1 induced modest increase of Cxcl1, Tnf and Il6 but rm-CXCL1+Lm lysate failed to induce Cxcl1, Tnf and Il6 mRNA (Fig 4C–4E). These results altogether demonstrate that L. major inactivates biological activity of rm-CXCL1.
To determine whether CXCL1 is degraded by L. major, we incubated rm-CXCL1 with Lm lysate for various time periods and examined rm-CXCL1 by silver staining (Fig 5A). Rm-CXCL1 was detected at approximately 10KDa by silver stain. Interestingly, when incubated with Lm lysate rm-CXCL1 showed two bands as early as 30 minutes after incubation, suggesting Lm lysate-mediated cleavage of rm-CXCL1 (Fig 5A and S4 Fig). By 2 hours, only the cleaved rm-CXCL1 was observed and this cleaved band intensity decreased over time, suggesting further degradation (Fig 5A). L. major-mediated cleavage of rm-CXCL1 was specific because similar experiments done with rm-CXCL2, rh-CXCL1 and rh-CXCL2 did not result in cleavage or degradation of these recombinant proteins (Fig 5B).
Based on the cleavage pattern of rm-CXCL1 (less than 1KDa shift), cleavage either at the N-terminal or C-terminal end would result in a large fragment (detected by silver stain as the lower band, Fig 5A) and a smaller fragment (which could not be detected by silver stain; Fig 5C). To identify the cleavage product and site, we processed the full-length rm-CXCL1 and Lm lysate-cleaved rm-CXCL1 bands by trypsin digestion and performed mass spectrometry to determine the sequence of the bands (S5 Fig). The tryptic peptide sequence readouts of full-length rm-CXCL1 covered the whole rm-CXCL1 peptide sequence (S5B Fig), while the peptide coverage of cleaved rm-CXCL1 covered all of the rm-CXCL1 peptide sequence except the last 7 amino acids (MLKGVPK) at the C-terminal end (S5C Fig). Thus, L. major cleaves rm-CXCL1 after lysine 65 (K65) residue that results in a large N-terminal rm-CXCL1 fragment lacking 7 amino acids at the C-terminal end (Fig 5D).
L. major has several proteases that enable it to survive within a cell and establish infection [49]. Metalloproteases and cysteine, serine and aspartic proteases are the major proteases described in L. major [49]. To examine whether cleavage activity was dependent on proteases, we first treated our rm-CXCL1 + Lm sup culture with pan protease inhibitor (Roche and Sigma) (Fig 6A–6C). Roche cOmplete protease inhibitor is a broad inhibitor of proteases including serine, cysteine and metalloproteases, and Sigma P8340 inhibitor is reported to inhibit serine, cysteine, acid proteases and aminopeptidases. However, the presence of protease inhibitors from Roche (cOmplete) or Sigma (P8340) did not inhibit Lm sup-mediated degradation of rm-CXCL1 as demonstrated by ELISA (Fig 6B and 6C and S6 Fig). Marimastat, a broad inhibitor of matrix metalloproteases [50], did not rescue the degradation of rm-CXCL1 by Lm sup (S6B Fig). EDTA treatment, which chelates metal ions such as Ca2+ and Fe3+ (and thus can inhibit certain metalloprotease), did not rescue rm-CXCL1 cleavage by Lm sup (Fig 6D and 6E). Interestingly, 1,10-Phenanthroline, a Zn2+ metalloprotease inhibitor [51], rescued rm-CXCL1 degradation by Lm sup (Fig 6F and 6G). While the addition of 1,10-Phenanthroline resulted in the upward shift of rm-CXCL1 bands in SDS-PAGE gels; importantly, no cleavage or degradation of rm-CXCL1 was observed (Fig 6G).
Several studies have shown that 1,10-Phenanthroline inhibits GP63, a Zn2+ metalloprotease present on all Leishmania spp and is often used in biochemical assays to inhibit non-specific proteolytic activity of GP63 [52]. To determine whether GP63 was involved in specific cleavage of rm-CXCL1, we immunoprecipitated (IP) GP63 from Lm lysate using anti-GP63 antibody (Fig 6H). Immunoblotting the GP63+ve and GP63-ve IP fractions for GP63 showed that GP63 was exclusively present in the IP+ve fraction and not present in the IP-ve fraction demonstrating successful immunoprecipitation of GP63 (Fig 6H). More importantly, when these fractions were cultured with rm-CXCL1, the IP-ve (Lm lysate lacking GP63) but not the IP+ve (i.e. GP63) fraction degraded rm-CXCL1 suggesting a role for GP63-independent metalloprotease in the specific degradation of rm-CXCL1 (Fig 6I).
Our data suggest that L. major Zn2+ metalloprotease specifically cleaves rm-CXCL1 after the K65 residue leaving a 7-mer amino acid sequence (MLKGVPK). To further examine the specificity of this L. major metalloprotease, we designed a blocking peptide that covered this cleavage site (i.e. the last 15 amino acid sequences of rm-CXCL1; Fig 7A). The addition of blocking peptide was able to rescue Lm lysate-mediated degradation of rm-CXCL1 in a dose-dependent manner (Fig 7B). The peptide sequence from the signal peptide region of rm-CXCL1 (peptide #1 and peptide #2, Fig 7A) did not inhibit Lm lysate-mediated degradation of rm-CXCL1, demonstrating the specificity of the blocking peptide (Fig 7B). These data demonstrate that a synthetic peptide spanning the murine CXCL1 cleavage site can competitively inhibit L. major-mediated rm-CXCL1 degradation.
CXCL1 is a potent neutrophil chemoattractant and is rapidly upregulated following Leishmania spp. infection in the skin [37]; however, the role of CXCL1 during Leishmania spp. infection of mice in vivo has been understudied. Charmoy et al showed that mice deficient in CXCL1 have a slight increase in lesion size and reduced numbers of neutrophil infiltrates in chronic lesions, but the overall pathology remained similar between WT and Cxcl1-/- mice [33]. However, it should be noted that the study used the L. major Seidman strain (LmSd) that causes a non-healing infection in C57BL/6 mice [33]. A recent study showed that Tlr2-/- mice produce significantly lower levels of CXCL1 following L. major infections and develop smaller lesions suggesting a pathogenic role for CXCL1 [35]. Furthermore, injection of recombinant CXCL1 to Tlr2-/- during L. major infection resulted in increased parasite numbers and disease pathology [35]. Thus, more thorough analyses are needed to determine the precise role of CXCL1 during L. major infection in mice and humans.
Importantly, even depletion of neutrophils using anti-Gr1 neutralizing antibody does not always lead to increased parasitic burden, lesion or pathology [21]. While some studies with acute depletion of neutrophils reported worsened pathology during L. major infections, other studies have shown no role or even amelioration of disease pathology [5, 53–61]. Thus, mouse strain, L. major strain, route of infection and timing of CXCL1 release may all contribute to the fate of the infection and overall pathology. In line with this thought, two independent L. major strains were tested in our study, and both strains degraded CXCL1 in our experimental systems. Whether this degradation of murine CXCL1 is a specific feature of L. major or a more general feature of all Leishmania spp. will need further investigation.
Our initial observation demonstrated that L. major in culture with macrophages effectively inhibited LPS-induced CXCL1 production. Further examination showed that intracellular L. major within the macrophages did not inhibit LPS-induced CXCL1 production. These experiments suggest that L. major does not interfere with signaling pathways that promote CXCL1 production but directly regulates secreted CXCL1 in the extracellular milieu. L. major are present as metacyclic promastigotes in sandflies [62]. When infected into the mammalian host, L. major infect macrophages and resides within the vacuole where they transform into amastigotes, multiply and establish infection [63]. Because L. major-infected macrophages which contain amastigotes did not inhibit CXCL1 production, it could be posited that only the promastigote form of L. major degrades CXCL1. However, genome microarray analysis shows that the majority of genes are constitutively expressed in both L. major promastigotes and amastigotes (>90%) and therefore the major genes and virulent factors expressed by these different stages of L. major may not be different [64]. Thus, both L. major promastigotes and amastigotes may similarly cleave and degrade murine CXCL1 in an in vitro assay. In addition, infected macrophages can secrete exosomes containing L. major proteins [65, 66], which could degrade murine CXCL1. We have consistently shown that L. major secreted microvesicles (i.e. exosomes) contain these metalloproteases that can degrade murine CXCL1. Physiologically, L. major promastigotes released during a sandfly bite may act locally to limit acute CXCL1 produced in response to the infection. Once L. major are phagocytosed and transformed into amastigotes, they are physically separated from the CXCL1 due to their location (amastigotes are in the vacuole while CXCL1 are secreted by the macrophages and are extracellular) and may only impact CXCL1 production through the L. major component laden exosomes.
Our studies suggest that a yet-unknown metalloprotease from L. major cleaves murine CXCL1. Given that 1,10-Phenanthroline (Zn2+ chelator) but not EDTA (Ca2+ and Fe3+ metal ion chelator) or Marimastat (Matrix metalloprotease inhibitor) rescued L. major-mediated CXCL1 degradation, it is possible that the unknown metalloprotease is a Zn2+ metalloprotease. Our studies also demonstrate that the L. major metalloprotease is highly specific for murine CXCL1 in that its closest murine homolog, CXCL2, or human homologs, CXCL1, CXCL2 or CXCL8, were not degraded. We hypothesize that this particular L. major metalloprotease may have evolved to specifically evade host immune response in rodents. Given that human CXCL1 homologs are not susceptible to this metalloprotease, one can argue that our results may not have any importance from a public health standpoint, but we reason otherwise. Our results are highly relevant to public health because rodents are ubiquitous, serve as a reservoir for Leishmania spp. and we highlight a rodent-specific L. major evolution in targeting murine CXCL1 [67].
Our studies demonstrate that the murine CXCL1 is first cleaved at the C-terminal end after K65 releasing a 7aa residue. CXCL1 cleavage occurs as early as 30 minutes of incubation with L. major and by 4 hours the cleavage is complete in that only the cleaved bands are observed. Interestingly, we did not observe any accumulation of the cleaved CXCL1 band, suggesting continuous degradation of the cleaved form. While our data clearly demonstrate that the unknown L. major metalloprotease mediates cleavage of murine CXCL1, how the cleaved CXCL1 is further degraded will be the subject of future investigation. The cleaved murine CXCL1 (lacking the C-terminal 7 aa) may be highly unstable and undergo spontaneous degradation overtime. Alternatively, the cleaved murine CXCL1 may be susceptible to additional L. major proteases which promote its subsequent degradation.
In conclusion, we have identified an immune evasion strategy utilized by L. major that is highly specific to murine CXCL1 (Fig 7C). Specifically, L. major-associated metalloprotease cleaved murine CXCL1 at K65 residue and released a C-terminal 7 amino acid fragment to promote its degradation. Finally, we have designed a peptide spanning the cleavage site of CXCL1 that inhibited murine CXCL1 cleavage by L. major. Our study altogether uncovered an immune evasion strategy employed by L. major that may have evolved in rodents and highlights how parasites may utilize diverse immune evasion strategy to establish infection within its diverse mammalian hosts.
Experimental procedures that utilized mice were all approved by the University of Iowa Animal Care and Use Committee (Approved Animal Protocol # 7042004 –PI Dr. Prajwal Gurung) and performed in accordance to the Office of Laboratory Animal Welfare guidelines and the PHS Policy on Humane Care and Use of Laboratory Animals.
BMDMs were prepared as described previously (32). Briefly, bone marrow cells were harvested from the hind limbs of BALB/c mice (Jackson Laboratory, Stock No. 000651) and cultured in L cell-conditioned IMDM medium supplemented with 10% FBS, 1% nonessential amino acid, and 1% penicillin-streptomycin for 5–7 d to differentiate into macrophages. BMDMs were counted and seeded at 1 × 106 cells in 12-well cell culture plates in IMDM media containing 10% FBS, 1% nonessential amino acids, and 1% penicillin-streptomycin. BMDMs were primed with LPS (20 ng/ml) and infected with 1, 2, 5, 10 or 20 MOI L. major promastigotes for 24 and 48 hours (S1 Fig). For some experiments, BMDMs were LPS primed to generate supernatant containing cytokines for in vitro biochemical analysis with L. major supernatant (Lm sup) and L. major lysate (Lm lysate). BMDMs were also treated with rm-CXCL1 (20 ng/ml) or rm-CXCL1+Lm sup preparations in some experiments.
L. major strains WHOM/IR/-173 [68] and IA0 [48] were grown in T-25 flasks with M199 media supplemented with 10% FBS, 5% HEPES and 1% penicillin-streptomycin at room temperature. BMDMs were infected with 20 MOI of L. major promastigotes for 48 hours. Conditioned L. major supernatant (Lm sup) was prepared after L. major reached the stable growth phase at approximately 20 × 106 / ml. Following centrifugation (3000×g) to pellet L. major, supernatants were collected and filtered using a 0.2μm vacuum filter to harvest Lm sup. Lm lysate was prepared by collecting the L. major pellet, washing 3 times with PBS followed by 3 freeze-thaw cycles of 7 × 109 L. major /ml in PBS. For the preparation of L. major exosomes, promastigotes were grown to stable phase and changed to serum-free media and incubated at room temperature overnight. Supernatant was obtained as above and then separated using a 100kDa molecular weight Amicon ultra-15 centrifugal filter (MilliporeSigma, Burlington, MA). The concentrated vesicles were washed twice with PBS by ultracentrifugation at 110,000xg at 4°C for 1 hour (Beckman Optma MAX-XP Ultracentrifuge, rotor TLA 120.2; Beckman Coulter Inc., USA). The final pellet (exosomes) was resuspended in PBS and used in in vitro biochemistry assays.
BALB/c mice were injected with 2 x 106 L. major promastigotes / footpad. 4 weeks post infection when footpad lesions were visible, mice were euthanized and footpad harvested and homogenized in PBS. L. major burden in the footpad were determined by limiting dilution assay as described previously [32]. Footpad lysates were centrifuged at passed through a 70micron filter to eliminate cellular debris. These footpad lysates were used as a source of L. major amastigotes. These footpad lysates with L. major amastigotes at a concentration of 2.5 x 108 and 1.25 x 108 (determined by limiting dilution assay) were incubated with 2ng rm-CXCL1.
For in vitro reactions, 125ng of total recombinant protein was separated on 4–20% or 10–20% Novex WedgeWell Tris-Glycine precast gels (Thermofisher, Waltham, MA). For silver stain, the Pierce Mass Spec compatible silver stain kit was used per manufacturer’s instructions (Thermofisher). Synthetic peptides were generated and purchased from Pierce Biotechnology (Thermofisher) and are composed of the following sequences: peptide#1: CAALLLLATSRLA; peptide#2: RLATGAPIAN; blocking peptide: LVQKIVQKMLKGVPK. For Western blot analysis a semi-dry transfer was completed using trans-blot turbo system (Bio-Rad, Hercules, CA) with Immobilon PVDF membrane. The membranes were blocked in 5% non-fat milk for 1 hour at room temperature. Membranes were probed with sheep polyclonal antibody Sp180 raised against L. chagasi promastigote GP63 at a concentration 1:1500 incubated overnight at 4°C. The membranes were then probed with HRP-tagged secondary antibodies at room temperature for 1 hour and developed using Immobilon Forte Western HRP Substrate (MilliporeSigma) and imaged with an Odyssey Fc Infrared Imaging System (LI-COR Bioscience, Lincoln, NE). Immunoprecipitation was performed using Protein G Dynabeads and Leishmania GP63 monoclonal antibody clone 96–126 (Thermofisher). As instructed by the manufacturer, 50μL beads were washed in PBS with 0.02% Tween 20 and bound to 5μg antibody for at least 10 minutes. Following a wash, the antibody bound beads were incubated with Lm lysate for another 10 minutes and finally either antigen-antibody binding was negated by acidic (pH 2.8) 50mM Glycine treatment or bead bound antibody-antigen complexes were used directly for biochemistry.
In-gel trypsin digestion: The gel was stained using a Pierce mass spec compatible silver stain kit (Thermofisher) per manufacturer directions. A procedure slightly modified than that described by Yu et al [69] was used for in-gel digestion. Briefly, the targeted protein bands from SDS-PAGE gel were manually excised, cut into 1mm3 pieces, and washed in 100mM ammonium bicarbonate:acetonitrile (1:1, v/v) and 25mM ammonium bicarbonate /acetonitrile (1:1, v/v), respectively, to achieve complete de-staining. The gel pieces were further treated with acetonitrile (ACN), to effectively “dry” the gel segments and then reduced in 50μl of 10mM DTT at 56°C for 60 min. Gel-trapped proteins were alkylated with 55mM chloroacetamide (CAM) for 30 min at room temperature. The gel pieces were washed with 25mM ammonium bicarbonate: acetonitrile (1:1, v/v) twice to remove excess DTT and CAM. 50μl of cold trypsin solution at 10ng/μl in 25mM ammonium bicarbonate was then added to the gel pieces and they were allowed to swell on ice for 60 min. Digestion was conducted at 37°C for 16 h. Peptide extraction was performed three times, adding 100μL of 50% acetonitrile/0.1% formic acid for 0.5 h, combining the supernatants. The combined extracts were concentrated in a lyophilizer and rehydrated with 15μl of mobile phase A.
LC-MS/MS: Mass spectrometry data were collected using an Orbitrap Fusion Lumos mass spectrometer (Thermofisher) coupled to an Easy-nLC-120 System (Proxeon P/N LC1400). The autosampler is set to aspirate 3μl (estimated 0.3ug) of reconstituted digest and load the solution on a 2.5cm C18 trap (New Objective, P/N IT100-25H002) coupled to waste, HV or analytical column through a microcross assembly (IDEX, P/N UH-752). Peptides are desalted on the trap using 16μl mobile phase A for 4 min. The waste valve is then blocked and a gradient begins flowing at 0.4μl/min through a self-packed analytical column, 10cm in length × 75μm id. The fused silica column was tapered from 100μm ID (Polymicro) to ~8μm at the tip using a Sutter P-2000 laser puller then packed with 2.7micron Halo C18 particles using a He-pressurized SS cylinder. Peptides were separated in-line with the mass spectrometer using a 70 min gradient composed of linear and static segments wherein buffer A is 0.1% formic acid and B is 95%ACN, 0.1% formic acid. The gradient begins first hold at 4% for 3 min then makes the following transitions (%B, min): (2, 0), (35, 46), (60, 56), (98, 62), (98, 70).
Tandem mass spectrometry on the Thermo Q-Exactive hf: Data acquisitions begin with a survey scan (m/z 380–1800) acquired on a Q-Exactive Orbitrap mass spectrometer (Thermofisher) at a resolution of 120,000 in the off-axis Orbitrap segment (MS1) with automatic gain control (AGC) set to 3E06 and a maximum injection time of 50 ms. MS1 scans were acquired every 3 sec during the 70-min gradient described above. The most abundant precursors were selected among 2–6 charge state ions at a 1E05 AGC and 70 ms maximum injection time. Ions were isolated with a 1.6 Th window using the multi-segment quadrupole and subjected to dynamic exclusion for 30 sec if they were targeted twice in the prior 30 sec. Selected ions were then subjected to high energy collision-induced dissociation (HCD) in the ion routing multipole (IRM). Targeted precursors were fragmented by (HCD) at 30% collision energy in the IRM. HCD fragment ions were analyzed using the Orbitrap (AGC 1.2E05, maximum injection time 110 ms, and resolution set to 30,000 at 400 Th). Both MS2 data were recorded as centroid and the MS1 survey scans were recorded in profile mode.
Proteomic searches: Initial spectral searches were performed with both Mascot version 2.6.2 (MatrixScience) and Byonic search engines (Protein Metrics ver. 2.8.2). Search databases were composed of the Uniprot KB for species 10090 (mouse) downloaded February 6, 2016 containing 58436 sequences. In either search, an equal number of decoy entries were created and searched simultaneously by reversing the original entries in the target database. Precursor mass tolerance was set to 5 ppm and fragments were searched at 10 ppm. A fixed 57 Da modification was assumed for cysteine residues while variable oxidation was allowed at methionine. A variable GG modification at lysine was set to monitor ubiquitylation and potential phosphorylation was accessed at Ser and Thr residues. The false discovery rate was maintained at 1% by tracking matches to the decoy database. Both Mascot and Byonic search results were combined and validated using Scaffold ver. 4.8.5 (Proteome Software). Protein assignments required a minimum of two peptides established at 70% probability (local FDR algorithm) and an overall 95% protein probability (assigned by Protein Prophet). Approximately 300 protein families (including common contaminants) were assigned at a total FDR to 1.2%. Proteins were annotated with GO terms from goa_uniprot_all.gaf downloaded on May 3, 2017.
Cytokine ELISAs and multiplex immunoassays were performed according to manufacturer instructions. Mouse and human CXCL1 and CXCL2 as well as human IL-8 (CXCL8) were obtained from R&D Systems (Minneapolis, MN) and multiplex CXCL1, IL-6, and TNF were completed using ProcartaPlex assays (Thermofisher) which were run on the BioRad Bio-Plex (Luminex, Austin, TX). We obtained our recombinant proteins from Tonbo biosciences which included mouse CXCL1, CXCL2, and TNF as well as human CXCL1, CXCL2, and CXCL8.
RNA was extracted from BMDM cultures in TRIzol reagent (Thermofisher) followed by chloroform extraction and isopropanol precipitation. The extracted RNA was reverse-transcribed into cDNA by using qScript Supermix (Quanta Biosciences, Beverly, MA). Real-time quantitative PCR was performed on Eppendorf realplex EPgradient S Mastercycler (Eppendorf, Germany) using PerfeCTa SYBR Green SuperMix ROX (Quanta Biosciences) and the appropriate primers. The sequences of the quantitative RT-PCR primers are as follows: mCXCL1: 5’-CAATGAGCTGCGCTGTCAGTG-3’, 5’-CTTGGGGACACCTTTTAGCATC-3’; mTnf: 5′-CATCTTCTCAAAATTCGAGTGACAA-3′, 5′-TGGGAGTAGACAAGGTACAACCC-3′; mIL-6: 5’-CAAGAAAGACAAAGCCAGAGTC-3’, 5’-GAAATTGGGGTAGGAAGGAC-3’. Threshold cycle Ct values were normalized to GAPDH, and gene fold change was determined by the relative comparison method, relative to the 0 h time point.
GraphPad Prism 8.0 software was used for data analysis and figure presentation. Data are shown as mean ± SEM. Statistical significance was determined by t tests (two-tailed and Mann-Whitney) for two groups, and one-way ANOVA (with Dunnett’s or Tukey’s multiple comparisons tests) for three or more groups. For the purpose of alignment presentation, sequences were obtained from the protein manufacturer or the NCBI protein database and figures were prepared using Jalview 2 [70].
|
10.1371/journal.pntd.0006877 | Seroprevalence for the tick-borne relapsing fever spirochete Borrelia turicatae among small and medium sized mammals of Texas | In low elevation arid regions throughout the southern United States, Borrelia turicatae is the principal agent of tick-borne relapsing fever. However, endemic foci and the vertebrate hosts involved in the ecology of B. turicatae remain undefined. Experimental infection studies suggest that small and medium sized mammals likely maintain B. turicatae in nature, while the tick vector is a long-lived reservoir.
Serum samples from wild caught rodents, raccoons, and wild and domestic canids from 23 counties in Texas were screened for prior exposure to B. turicatae. Serological assays were performed using B. turicatae protein lysates and recombinant Borrelia immunogenic protein A (rBipA), a diagnostic protein that is unique to RF spirochetes and may be a species-specific antigen.
Serological responses to B. turicatae were detected from 24 coyotes, one gray fox, two raccoons, and one rodent from six counties in Texas. These studies indicate that wild canids and raccoons were exposed to B. turicatae and are likely involved in the pathogen’s ecology. Additionally, more work should focus on evaluating rodent exposure to B. turicatae and the role of these small mammals in the pathogen’s maintenance in nature.
| In arid regions of the southern United States and Mexico, tick-borne relapsing fever is primarily caused by Borrelia turicatae. The tick vector, Ornithodoros turicata, feeds indiscriminately on a variety of vertebrates; however, it is unclear which animals are competent hosts for B. turicatae. This study evaluates the exposure of small and medium sized mammals in Texas to B. turicatae and identifies likely hosts for the pathogens. This work will provide insight regarding mammals to target for surveillance to identify endemic foci and to better prevent human exposure.
| Tick-borne relapsing fever (RF) is primarily caused by spirochetes in the genus Borrelia and the pathogens are transmitted when infected Ornithodoros ticks feed on a competent vertebrate host. In the United States and Mexico, there is an association between Ornithodoros ticks and RF spirochete species where Ornithodoros hermsi, Ornithodoros parkeri, Ornithodoros turicata, and Ornithodoros talaje transmit Borrelia hermsii, Borrelia parkeri, Borrelia turicatae, and Borrelia mazzottii, respectively [1]. Furthermore, these tick species involved in human disease are distributed in varying ecological niches. For example, the ecology of O. hermsi is associated with coniferous forests at elevations above 900 meters throughout the western United States and Canada [2–6]. Ornithodoros parkeri has also been collected in semi-arid regions of the western United States at elevations from sea level to over 2,000 meters [7, 8]. Ornithodoros turicata is found in arid regions of Mexico, the mid- and southwestern United States from California to Texas, and a population exists in Florida [9–11]. The ecology of O. talaje overlaps that of O. turicata and collections have occurred in Mexico and in Texas [1, 12, 13]. Of the Ornithodoros species that transmit RF spirochetes, O. turicata and O. talaje are currently the only ones known in Texas, yet there are few records of O. talaje collections and B. mazzottii has not been isolated in the laboratory.
The biology of both RF spirochetes and their tick vector have posed challenges in defining the pathogens’ ecology. Ornithodoros species are rapid feeding ticks that reside in cavities including wood crevices, dens, nests, and karst formations [6, 7, 9, 14]. Thus, the vector is rarely found attached on the vertebrate host. Moreover, in the vertebrate host, spirochetes replicate in the blood reaching densities of ~1 x 107 bacteria per ml before being cleared by an antibody mediated response [15]. The pathogens undergo antigenic variation and subsequently repopulate the blood [15]. This dynamic between antigenic variation and the host antibody response can continue for two to three months in a competent host [11, 16]. The cyclic nature of RF spirochetes within a competent host poses challenges when attempting to directly detect the pathogens in the blood of wild caught animals because there are quiescent periods when the spirochetes are undetectable. Since RF spirochetes induce a robust IgG response [17–19], serological surveillance is a practical approach toward defining the pathogens’ ecology given the temporal persistence of generated antibodies in the host’s blood.
The ecology of B. turicatae is poorly defined and in this current study we utilized a diagnostic antigen, the Borrelia immunogenic protein A (BipA), which has been used to assess canine, rodent, and human exposure to the pathogen [17, 20]. Aside from the closely related B. parkeri, BipA is highly variable between species of RF spirochetes [20]. Moreover, a BipA homologue has not been identified in other viral, parasitic, or bacterial pathogens [17]. Utilizing this diagnostic antigen, we evaluated the exposure of wild and domestic canids, raccoons, and rodents to B. turicatae. Serum samples were collected from 23 counties in Texas and screened against B. turicatae protein lysates and recombinant BipA (rBipA). Most rodents were also identified to species by morphology and molecular sequencing of the cytochrome B gene. Our findings indicate that Canis latrans (coyote), Urocyon cinereoargenteus (gray fox), Procyon lotor (raccoon), and Peromyscus leucopus (white-footed mouse) may be vertebrate hosts for B. turicatae in nature.
Rodent collections were approved by the Institutional Animal Use and Care Committee at Mississippi State University (IACUC protocol #11–091) and Texas Parks and Wildlife (Scientific Research Permit #SPR-0812-958). Collections of coyotes, gray fox, and raccoon serum samples originally occurred as part of the rabies surveillance program by the Texas Department of State Health Services. Collection of shelter canine serum samples were approved by the University of Texas Health Science Center Animal Welfare Committee (AWC-07-147 and AWC-03-029).
Animal samplings occurred between 2005 and 2018. Rodents were captured alive using Sherman live traps (H.B. Sherman Traps, Tallahassee, FL). Traps were placed in and around houses, barns, and fields in the late afternoon and baited with dried oats. The following morning traps were collected, and the animals processed. Animals were euthanized by inhalation of isoflurane and exsanguinated by cardiac puncture. A drop of blood was placed on a microscope slide and the presence of spirochetes was evaluated by dark field microscopy. Peripheral blood smears were also made on microscope slides. The remaining blood was centrifuged at 1,000 x g and serum separated from the blood clot. Animals were evaluated for argasid ticks.
Shelter dogs and wild canids and raccoons were also sampled. Serum samples from stray domestic dogs located in Brownsville, TX were collected, as previously described [21]. Coyote, gray fox, and raccoon serum samples were collected as part of the Texas Department of State Health Services rabies surveillance program. The animals were captured in Tomahawk traps, terminally sampled, and serum samples stored at -20°C.
Canids and raccoons were identified to the species level using morphological characteristics. Rodents were identified by morphological characteristics and molecular analysis of the cytB gene. For rodent morphological characteristics, body weights were recorded with Pesola spring scales (PESOLA SG, Baar, Switzerland), gender determined, and body and tail measurements recorded. Photographs of each animal were obtained for future reference. For molecular analysis, a 3-mm tissue biopsy was collected from each animal, stored in 90% ethanol, and DNA extracted using the DNeasy Blood and Tissue kit (Qiagen Sciences, Inc., Germantown, MD). PCR was performed using forward (5’-CCATGAGGACAAATATCCTTCTGAGGG-3’) and reverse (5’-GCCCTCAGAAGGATATTGTCCTCATGG-3’) primers for cytB, and sequencing performed as previously described [19, 22]. Sequences were assembled into overlapping contiguous DNA segments (contigs) using Vector NTI 11.0 software (ThermoFisher Scientific, Waltham, MA). Contigs were evaluated using BLASTn on NCBI.
Immunoblotting was performed to evaluate seroconversion against B. turicatae protein lysates and rBipA, as previously described [17]. Briefly, protein lysates from 1 x 107 B. turicatae spirochetes and 1 µg of rBipA were loaded into the wells of Mini-PROTEAN TGX precast gels (Bio-Rad, Hercules, CA). Gels were run for 1.5 hours and proteins were transferred onto Immobilon PVDF membranes (Millipore, Billerica, MA). Membranes were blocked overnight with Tropix Iblock (Thermo Fisher Scientific, Waltham, MA) and then probed for one hour at room temperature with serum samples diluted 1:200. The secondary molecule was HRP-conjugated protein G (Thermo Fisher Scientific, Waltham, MA) for canids and rodents at a 1:4,000 dilution. Raccoon serum samples were probed with a goat anti-raccoon IgG-HRP conjugated antibody (Alpha Diagnostics Intl. Inc., San Antonio, TX) at a 1:4,000 dilution. The substrate used to detect binding was Amersham ECL Western Blotting Detection Reagent (GE Healthcare, Buckinghamshire, UK). A sample was considered positive for B. turicatae if we detected reactivity to at least five proteins in the B. turicatae protein lysate and rBipA.
For visualizing the ecoregions of Texas we obtained a shapefile from the United States Environmental Protection Agency, which included the following 12 ecoregions: Arizona/New Mexico Mountain, Central Great Plains, Chihuahua Deserts, Cross Timbers, East Central Texas Plains, Edwards Plateau, High Plains, South Central Plains, Southern Texas Plains, Southwestern Tablelands, Texas Blackland Prairies, and Western Gulf Coastal Plains [23]. These ecoregions were defined based upon several biotic and abiotic factors such as climate, vegetation, soil type, geology, land use, wildlife, and hydrology [23]. This shapefile was then imported into ArcMap and we overlaid each county where collections occurred in Texas noting the taxa group (coyote = C, Dog = D, gray fox = GF, raccoons = RA, and rodents = R) and the number of collections (Fig 1).
Statistical analysis was performed using R 3.3.1 (R foundation for Statistical Computing, Vienna, Austria, https://www.R-project.org/). The 95% confidence intervals (CI) were determined for each group of samples tested that had at least one positive sample, using the proportions test. A binomial distribution was assumed with determining CI.
One to four field sites were sampled in 23 counties of Texas between 2005 and 2018. Sites included private property that was accessible through the Texas Ecolab Program and Texas Parks and Wildlife Management Areas. Counties where samples were collected were within the following Texas ecoregions: Central Great Plains, Chihuahuan desert, Cross Timber, East Central Texas Plains, Edwards Plateau, South Central Plains, Southern Texas Plains, Southwestern Tablelands, Texas Blackland Prairies, and Western Gulf Coastal Plains (Fig 1). A total of 463 canids were sampled and included 185 shelter dogs (Canis lupus familiaris), 220 coyotes (C. latrans) and 58 gray foxes (U. cinereoargenteus) (Fig 1). Serum samples were also collected from 25 raccoons (P. lotor) and 263 rodents (Fig 1). Argasid ticks were not detected on the animals.
Animals were considered susceptible to infection by B. turicatae based on serological reactivity to at least five bands in B. turicatae protein lysates and rBipA. Assessing serological responses of canids (Fig 2) indicated a total seroprevalence of 5.4% (CI = 3.6–8.0%) (Table 1). None of the 185 shelter dogs that were screened had a detectable antibody response in the diagnostic assay, while 10.9% and 1.7% (CI = 7.3–16.0% and 0.09–10.5%) seroprevalence was detected in coyotes and gray fox, respectively. Webb County had the highest number of seropositive coyotes with a total of 10 animals. Presidio and Zapata County each had five seropositive animals, while Dimmit County had four. In the animals exposed to B. turicatae, a gender difference was not detected. In El Paso County, there was a single juvenile male gray fox that was seropositive, resulting in 1.7% (CI = 0.7–53.3%) prevalence among gray fox.
Species within seven genera of rodents were collected between 2012 and 2015 including Peromyscus maniculatus, Peromyscus leucopus, Chaetodipus hispidus, Sigmodon hispidus, Neotoma albigula, Perognathus, and Dipodomys species. Evaluating serological responses (Fig 2) indicated that 0.4% (CI = 0.02–2.4%) were seropositive (Table 3). Peromyscus leucopus was the only positive animal and originated from Edwards County.
In this study, we began to define the ecology of B. turicatae in Texas by assessing serological responses as an indicator of host competency. While RF spirochete infections can persist for several months in a competent host [11], the pathogens’ life cycle is recurrent and direct detection of infection can be challenging because of the brevity of time when spirochetes are detectable in the blood. To circumvent this, we indirectly detected exposure to B. turicatae by assessing the vertebrate antibody response. Our findings indicate that wild canids are likely a host for B. turicatae in west Texas. These studies were also the first known serological evaluation of rodents and raccoons to B. turicatae and provided verification that exposure is occurring in this tick-host-pathogen relationship.
Rodents and insectivores are known reservoir hosts for at least two species of RF spirochete [3, 19, 24, 25], but the role of these small mammals in the ecology of B. turicatae is vague. In high elevation regions of the Western United States, sciurid rodents are the primary vertebrate host for B. hermsii, while the pathogens have also been detected in Neotoma macrotis [3, 24]. In regions of western Africa, Borrelia crocidurae is maintained in Mastomys and Crocidura species [19]. Previous tick transmission studies of B. turicatae to laboratory mice suggest that wild rodents may be susceptible to infection [17, 26, 27], and the identified seropositive P. leucopus from this current study indicated that white-footed mice are a potential competent host. However, we sampled the field site where this animal was collected three more times from 2012 to 2014 and failed to identify other positive rodents. Additional studies are needed to investigate the life cycle of B. turicatae in rodents to determine whether the pathogen attains densities in the animals that will facilitate spirochete acquisition and colonization of the tick vector.
There is mounting evidence that canids likely support the maintenance and dissemination of B. turicatae in nature. For example, the competency of domestic canines for B. turicatae was demonstrated as nearly half of the B. turicatae isolates have originated from sick dogs [28]. Moreover, successful infection of B. turicatae to a laboratory dog by tick bite suggested that the spirochetes attained sufficient densities in the blood to infect ticks [17]. In this current report, B. turicatae positive coyote and gray fox serum samples originated from Dimmit, Presidio, Webb, Zapata, and El Paso County, all of which border Mexico. With the broad home range of coyotes, these mammals are likely circulating B. turicatae between the United States and Mexico.
Coyotes possess highly organized social systems even in urban settings and are classified as transient or resident based on their territorial range [29, 30]. Transient coyotes are typically solitary subordinate young adults with a home range of 40 km2 to 395 km2. Resident coyotes have a home range of 8 km2 to 29 km2 and are part of the larger pack that include breeders, juveniles, and pups [30]. Coyote dens are often found in or around urban settings and with the expansion of these areas in Mexico and the United States, coyotes and humans are commonly in contact [30, 31].
A knowledge gap in the ecology of B. turicatae is a poor understanding regarding the dissemination of the vector in nature. Ornithodoros turicata are rapid feeders, completing a bloodmeal within five to 60 minutes after attachment [26]. However, it is unclear whether some proportion of ticks remain on the wild vertebrate host after engorgement, either attached or unattached, allowing for increased dissemination. Interestingly, we have collected engorged Carios kelleyi nymphs, which are rapid feeding argasid ticks of bats [9]. This suggests that some argasid species may remain on the vertebrate host for a duration of time after feeding. Population genetic studies are needed to evaluate the genetic diversity between O. turicata populations at different spatial scales collected in the United States, to estimate dissemination patterns of both vector and pathogen.
A limitation of our study is the likely circulation of additional uncharacterized RF spirochete species in the southern United States. While BipA is highly divergent between most species of RF spirochete and the recombinant protein can discriminate between B. hermsii and B. turicatae infections [17], additional work is needed to obtain novel spirochete species circulating in nature. For example, Ornithidoros talaje was recently collected in Texas [1], and while we failed to detect Borrelia DNA in these the ticks, the circulation of Borrelia mazzottii in the state exists. In 1955, B. mazzottii was reported to be transmissible by O. talaje ticks that were collected in northern Mexico [12], but since then reports of the disease have been absent. Recently, RF spirochetes were detected in a blood smear of a sick patient in Sonora, Mexico, but the species was unidentified [32]. Furthermore, Candidatus Borrelia texasensis was initially isolated in medium from an adult ixodid tick, Dermacentor variabilis, which was feeding on a coyote collected in Webb County, Texas [33]. The spirochete was initially cultured and grouped with RF spirochetes, but Lin and colleagues were unable to revive frozen stocks and an isolate does not exist. While it is unclear whether coyotes are a competent host for Candidatus Borrelia texasensis, the findings suggest that the mammals may be exposed to additional species of RF spirochete.
We recommend increased surveillance of small and medium sized mammals within metropolitan areas of Texas. San Antonio, Austin, and Dallas, Texas are in the top 11 most populated cities in the United States, are rapidly expanding, and evidence indicates that B. turicatae may be emerging in these areas. In 2017 there was an outbreak of TBRF among conference attendees in Austin, Texas [34], which is located in Travis County. This outbreak was in a densely populated area of the city and B. turicatae infected ticks were collected from rodent dens at a public park near the conference site. In our current report, there was little overlap between the Texas ecoregions that were sampled for the different vertebrate species (Table 1), and only three rodents were collected in Travis County. Future studies should focus on small and medium sized vertebrate sampling in regions where B. turicatae is emerging, and investigate host competence for the pathogen. As these studies are conducted, a refined understanding of the vertebrate hosts that support the ecology of B. turicatae will be attained, and surveillance and countermeasures can be implemented to improve public health.
|
10.1371/journal.ppat.1000985 | Integration Preferences of Wildtype AAV-2 for Consensus Rep-Binding Sites at Numerous Loci in the Human Genome | Adeno-associated virus type 2 (AAV) is known to establish latency by preferential integration in human chromosome 19q13.42. The AAV non-structural protein Rep appears to target a site called AAVS1 by simultaneously binding to Rep-binding sites (RBS) present on the AAV genome and within AAVS1. In the absence of Rep, as is the case with AAV vectors, chromosomal integration is rare and random. For a genome-wide survey of wildtype AAV integration a linker-selection-mediated (LSM)-PCR strategy was designed to retrieve AAV-chromosomal junctions. DNA sequence determination revealed wildtype AAV integration sites scattered over the entire human genome. The bioinformatic analysis of these integration sites compared to those of rep-deficient AAV vectors revealed a highly significant overrepresentation of integration events near to consensus RBS. Integration hotspots included AAVS1 with 10% of total events. Novel hotspots near consensus RBS were identified on chromosome 5p13.3 denoted AAVS2 and on chromsome 3p24.3 denoted AAVS3. AAVS2 displayed seven independent junctions clustered within only 14 bp of a consensus RBS which proved to bind Rep in vitro similar to the RBS in AAVS3. Expression of Rep in the presence of rep-deficient AAV vectors shifted targeting preferences from random integration back to the neighbourhood of consensus RBS at hotspots and numerous additional sites in the human genome. In summary, targeted AAV integration is not as specific for AAVS1 as previously assumed. Rather, Rep targets AAV to integrate into open chromatin regions in the reach of various, consensus RBS homologues in the human genome.
| This is the first unbiased genome-wide analysis of wildtype AAV integration combined with a thorough bioinformatic analysis of preferred genomic motifs and patterns in the neighbourhood of the integration sites identified. The preference of Rep-dependent AAV integration near multiple consensus Rep-binding sites was lost in the case of AAV vector integration in the absence of Rep expression. Our findings challenge the commonly accepted notion of site-specific AAV targeting to AAVS1 on chromosome 19q13.42. Although AAVS1 contains a canonical Rep-binding site, numerous additional sites including the newly identified hotspots AAVS2 on chromosome 5p13.3 and AAVS3 on chromosome 3p24.3 harbour functional Rep-binding sites suitable for AAV integration. AAV vectors are quickly moving forward in the clinic and Rep-dependent vector targeting strategies are being actively pursued. Detailed information of AAV wildtype versus recombinant AAV vector integration sites and preferences are needed to evaluate the safety profile of AAV vectors in gene therapy.
| The family of adeno-associated virus (AAV) represents defective, helper-dependent viruses that need to establish latency to ensure persistence in their primate hosts [1]. Upon natural infections in humans AAV genomes were shown to persist mainly as episomes and integrated AAV genomes were rarely detected [2]. The molecular mechanisms leading to integration have only been characterized for AAV type 2 that prefers integration near a site on human chromosome 19q13.42, called AAVS1 [3]. The specificity of AAV integration is mediated by the large regulatory AAV proteins, Rep78/68 [4]. During productive AAV replication in the presence of either adeno- or herpesvirus as a helper virus, Rep78/68 is required for AAV gene expression and DNA replication. The AAV origins of DNA replication reside in the 145 bp inverted terminal repeats (ITRs) that flank the 4.7 kb single-stranded AAV genome. Rep78 and/or Rep68 are expressed from the AAV p5 promoter and were shown to bind to the Rep-binding site (RBS) within the AAV-ITRs [5]. Rep unwinds the DNA and introduces a single-strand nick at the adjacent terminal resolution site (trs) [6]. The AAV-ITRs also serve as cis elements for chromosomal integration [4]. A RBS homologue present in the AAV p5 promoter was shown to mediate AAV integration in the absence of the ITRs [7]. DNA sequences homologous to the RBS and a nearby trs element were also found in AAVS1 [8], [9] and, in vitro, ternary complex formation of Rep68 with the AAV-ITR and AAVS1 was shown [10]. A 33 bp sequence of AAVS1 spanning the RBS and the trs element was sufficient to mediate AAV integration in vivo [4], [11]. AAV integrated at variable distances from the RBS in AAVS1 and sequence rearrangements were frequently found at AAV-chromosome junctions [8], [9], [12], [13], [14], [15]. Quantitative real-time PCR analysis of AAVS1-specific AAV-2 integration within hours after AAV-2 infection and at increasing MOIs showed that 10 to 20% of infected cells displayed AAV integration within a 4 kb region of AAVS1 on chromosome 19q13.42 [16], [17]. In AAV-infected and subsequently selected cell clones up to 80% of AAVS1-specific integration had been described before [18].
Although AAV has not been associated with disease in humans, it is well established that AAV Rep78/68 induces DNA damage, cell cycle arrest [19] and apoptosis [20]. In addition, AAV Rep interferes with helper adenovirus- [21] herpes simplex virus replication [22]. AAV holds much promise as a vector for gene therapy. As a rule, recombinant AAV vectors persist as non-integrated, nuclear episomes. AAV vectors lack the integration promoting rep gene and therefore only occasionally integrate into the host cell genome. The preferred integration of wildtype AAV-2 in chromosome 19q13.42 is unique and is commonly viewed as a specifically evolved virus-encoded targeting mechanism. Multiple attempts were published that aim to exploit Rep-mediated targeting specificity for chromosome 19q13.42 for the specific integration of gene therapy vectors [23], [24], [25], [26], [27], [28]. Yet chromosome 19q13.42 is not the only target region. The presence of alternative integration sites has long been postulated and in silico analysis detected numerous consensus Rep-binding sites in the human genome. Many of these bound Rep in vitro [29] but their in vivo accessibility for AAV integration has not been explored so far. From an evolutionary standpoint the assumption that AAV latency is ensured by more than one target site or mechanism appeared reasonable.
This study was designed to close the knowledge gap between AAVS1-specific and assumedly non-AAVS1-specific wildtype AAV integration and to compare the identified genomic sites to those preferred upon AAV vector transduction. An open survey of chromosomal integration preferences for wildtype AAV-2 was conducted and complemented by the bioinformatic analysis of genomic motifs and patterns in the genomic regions surrounding the integration loci.
The genomic structure of latent AAV in infected cells is highly variable. Wildtype AAV-2 was shown to integrate into the host cell genome, as well as persist as extrachromosomal, nuclear episomes [2], [30]. In either case multicopy, concatemeric structures predominate and often lead to unpredictable rearrangements involving the 145 bp inverted terminal repeats (ITRs). Therefore the retrieval of AAV-chromosome junctions suffers from the inherent problem of inefficient PCR reads through the hairpin ITR into the adjacent chromosomal sequences. This leads to a predominance of rearranged AAV genomes lacking chromosomal junctions in previous PCR-based studies [31], [32], [33]. Furthermore, previously cloned junctions often displayed unknown intervening sequences of varying lengths between AAV and the identified chromosomal sequence [12], [15], [16], [27], [34], [35], [36]. Therefore, unambiguous assignment of the AAV-derived and chromosome-derived parts of junctions requires sufficient DNA sequence lengths.
Several methods to identify virus-chromosome junctions have been developed to study retrovirus integration, where generally a single proviral copy per chromosomal site is found [37], [38]. The ultimate structure of the integrated long terminal repeat (LTR) is generally predictable in a way that allows an integration-specific PCR design. Linear amplification mediated (LAM)-PCR was initially designed to retrieve rare retroviral vector integration sites from small, clinical sample sizes [38]. We established a LAM-PCR with AAV primers in the “D” element of the AAV-ITR, the innermost and sole ITR region without internal inverse repetitions (Figure 1A). Unfortunately, pure AAV sequences with rearranged ITRs predominated, AAV-chromosome junctions were rare and the chromosomal DNA part often too short for unambiguous assignment to a unique genomic site. We then tested ligation-mediated (LM)-PCR that had been employed for broad surveys of lentivirus (HIV) or γ-retrovirus (MLV) integrations [39], [40], [41]. LM-PCR relies on a first LTR-specific primer. A linker is ligated to the first PCR strand that typically ends at the chosen restriction site within the unknown chromosomal sequence. A primer complementary to this linker ensures second strand synthesis and retrovirus-chromosome junctions are amplified by using a combination of retrovirus LTR-specific and linker-specific primer sets.
For this study a variation of LM-PCR, named linker-selection-mediated (LSM)-PCR was developed which enriched for bona fide AAV-chromosome fusion sequences. The genomic DNA of AAV-infected cells was cleaved with restriction enzymes that lead to sufficiently sized DNA segments to allow unambiguous genomic assignment of the chromosomal junction (Figure 1B). DNA sequences were amplified with one primer for a unique AAV-sequence, either of the p5 promoter or of the cap gene. The other primer binds to the linker DNA attached to the unknown chromosomal site. The structure of the linkers forces the PCR to initiate within the AAV genome, thereby suppressing amplification of chromosomal DNAs lacking integrated AAV. The use of non-cut enzymes for AAV-2 DNA helped to circumvent the problem of ligating linkers to episomal, non-integrated AAV DNA sequences. To further enrich for AAV-chromosome junctions a biotin tag was attached to the 5′-end of the linker primer. Thus, chromosome-derived PCR products could be enriched by streptavidin-mediated magnetic bead selection. This lead to PCR products selected for both, the presence of AAV and of an unknown chromosomal DNA sequence.
Using LSM-PCR a total of 1700 cloned PCR fragments were screened for DNA inserts of a minimal fragment size (>500 bp) to insure unambiguous detection of AAV-chromosome junctions. Out of 350 DNA sequence runs a total of 129 unique junction sites could be assigned to the human genome. Of these, 109 fulfilled the criteria outlined in the methods for unambiguous assignment of a single chromosomal site. Junctions were retrieved with non-cut enzymes for AAV-2, PvuII or EcoRV or with DraI, which cuts once in AAV-2 DNA outside of the region covered by the PCR. In addition, 43 wildtype AAV-2 infected Hela-derived single cell clones were generated of which eight harboured AAV-chromosome junctions that fulfilled the criteria outlined in the methods.
DNA sequence analysis revealed that AAV-2 wildtype integration sites were scattered over the entire human genome. The chromosomal distribution pattern is displayed in Figure 2A. Over one third of AAV integration sites were clustered at hotspots on chr. 19q13.42, on chr. 5p13.3 and on chr. 3p24.3 (Figure 2B–D). Infection with AAV in the absence of a helper virus leads to transient, low Rep expression. Many previous AAV integration studies used plasmid transfections of wildtype or vector AAV constructs often in combination with a high-level Rep expression construct. To evaluate whether high Rep expression influenced the target site preference of AAV, the sequence data of previously published transfection-based AAV integration sites [42] were reevaluated with the more stringent criteria outlined in the method. Of 157 DNA sequences retrieved after cotransfection of a rep-expression construct and an AAV vector plasmid 47 junction sequences fulfilled our criteria for unambiguous assignment of AAV to a unique chromosomal site (Table 1).
For AAV wildtype 10% of all retrieved junctions were detected at the hotspot on chr. 19q13.42 spread over a total of 33 kb around AAVS1 (Figure 2B). Only one out of twelve chr. 19q13.42-specific AAV junctions was located within the 4 kb region of AAVS1, where a consensus Rep-binding site and an adjacent trs site had been defined [4] The reevaluated distribution pattern of junctions generated by transfection of AAV vector- and Rep expression plasmids [42] was similar (Figure 2B). Latently AAV-infected Detroit 6 cells [43], [44] were analyzed as control. Using cap-specific primers the junction was detected within AAVS1 at nucleotide position 60,319,992. A second hotspot named AAVS2 was detected on the small arm of chr. 5p13.3 within an intergenic region, where ten independent integration sites were detected within 8 kb (Figure 2C). In seven of these junctions clustered within 14 bp AAV had integrated directly into a consensus Rep binding site. The reanalyzed chromosomal integrations from AAV plasmid transfection [42] displayed a similar pattern with six integrations within 16 bp of the consensus RBS (Figure 2C). The third hotspot named AAVS3 was found on chr. 3p24.3 (Figure 2D). Out of 13 sites detected on chr. 3, three integrations were clustered in a 8 kb region where a consensus Rep binding site GAGT GAGT GAGT GAGC GAGC was detected on the complement strand (Figure 2D).
To evaluate the binding affinities of Rep to the consensus RBS of the hotspots on chr.5 and chr. 3 compared to the RBS of chr. 19 or within the AAV genome, double-stranded oligonucleotides spanning the respective RBS regions (Figure 3) were submitted to mobility shift assays (EMSA) with increasing amounts of purified MBP-Rep78. Since it was previously shown that GAGG repeats are deficient in binding to Rep [10], [45], a mutated oligo derived from the RBS of AAVS2 displaying GAGG GAGG GAGC GAGG was used as a control. As an additional control, a random oligonucleotide of similar length was used. As shown in figure 4, the RBS of AAVS3 contained five instead of four GAGY repeats and bound Rep with a two-fold higher affinity than the oligonucleotide spanning the AAVS1 RBS and trs (Figure 4B). The RBS of AAVS2 showed 76% of the Rep-binding affinity of the AAVS1 sequence (Figure 4C). In contrast, the relative binding affinity normalized to the AAVS1 sequence dropped to 13% with the mutated AAVS2 oligonucleotide, which was in the range of the random oligonucleotide (Figure 4C). These findings confirm the importance of the GAGY repeats in Rep binding. As expected, Rep78 displayed the highest affinities for oligonucleotides spanning the A-stem of the AAV-ITR or the AAV p5–promoter (Figure 4A, 4D). In summary, the newly discovered hotspots for AAV integration, AAVS2 on chr. 5 and AAVS3 on chr. 3 display RBS similarly proficient for Rep-binding as AAVS1.
To evaluate whether AAV-2 wildtype prefers specific motifs or genomic features for chromosomal integration the detected chromosomal junctions were compared to integration sites described for infection of human cells with a rep-deleted AAV-2 based vector [46]. The published DNA sequence files were reanalyzed using the criteria as outlined in the methods. This led to 450 junctions that could be included as an AAV vector-specific data set (Table 1). The preference for integration next to selected genomic features was analyzed for rep-positive AAV wildtype and for rep-deficient AAV vectors (Table 2). The data showed that the integration frequency of AAV wildtype in genes was higher than expected by chance (Table 2). The frequency was comparable to that of rep-deficient AAV vectors, thus confirming the findings by Miller et al. [46].
To analyze the effect of epigenetic modifications on AAV integration the association of integration sites with histone modifications as markers for open or closed chromatin were assessed by chromatin immunoprecipitation sequencing (ChIP-Seq) analysis as outlined in the methods. Trimethylated lysine 27 of histone 3 (H3K27me3) is correlated with gene repression (closed chromatin) [47], while methylation of lysine 4 in H3K4me3 and H3K4me1 is indicative of promoter or enhancer regions (open chromatin) [48]. As shown in table 2 the association of AAV wildtype with open chromatin regions is significantly higher than expected from random controls. Conversely, the respective association with closed chromatin is significantly reduced. In summary, AAV wildtype prefers integration into open chromatin whereas closed chromatin was avoided.
A series of publications have shown that fused combinations of two to four GAGC motifs bind to Rep78/68 of AAV-2 [4], [49], [50], [51], [52], [53]. Moreover, in vitro ternary complex formation of Rep68 with the AAV-2 ITR and AAVS1 of chr. 19q13.42 [10] led to the concept of Rep acting as an adapter that targets AAV to the human genome. Although only AAV-2 has been analyzed for chromosomal integration so far, all known AAV serotypes displayed various combinations of GAGC and/or GAGT motifs in the ITR and the p5 promoter. An alignment of these AAV elements to the integration hotspots AAVS1, AAVS2 and AAVS3 is displayed in Figure 3.
Based on these data we hypothesized that AAV-2 wildtype, due to the presence of Rep, prefers integration at chromosomal sites in closer proximity to consensus Rep binding sites than would be expected from control sites. The hypothesis was tested with the three sets of junctions derived from: 1. Infection with AAV-2 wildtype, 2. Cotransfection of plasmids coding for an AAV vector and a constitutive Rep-expression cassette, and 3. Infection with Rep-deficient AAV vectors (Table 1). The distances between any one integration site and its nearest Rep-binding site were determined in the human genome and compared to similarly determined distances of individual control sites to the nearest Rep-binding sites. Calculations were repeated using various combinations of RBS as displayed in Figure 5.
The choice of randomly generated genomic control sites was considered optimal for comparative analysis of the three sets of data. Yet, a concern was the choice of restriction endonucleases for the identification of the wildtype AAV-2 integration sites by LSM-PCR. To control a bias introduced by a conceivable non-random genomic distribution of the restriction sites, the average distance of PvuII, EcoRV, or DraI-generated restriction sites to putative Rep-binding sites was compared to the average distances of random sites to Rep-binding sites. PvuII restriction sites were found to be closer to Rep-binding sites than random control sites (Figure S1). This was assumedly due to the high G+C content of the PvuII recognition sequence and of the consensus Rep-binding sites. Both EcoRV and DraI sites were found further apart from Rep-binding sites in accordance with their high A+T content (Figure S1). To circumvent any bias arising from the use of PvuII, the data set for AAV wildtype infection was calculated against the data set of random control sites as well as against the data sets for the restriction site–related controls. Since not more than two thirds of sites were generated with PvuII, the PvuII-related control sites would at most underestimate the association to Rep-binding sites and was therefore used as the most stringent control set. In addition all calculations were also performed with the set of random controls leading to similar findings (Figure S2).
The bioinformatic calculations with GAGC GAGC as a minimal Rep-binding site strikingly confirmed our hypothesis that integration of wildtype AAV takes place close to Rep-binding sites with very high significance (p <0.0001). A comparable effect was seen with the data set for AAV vectors in the presence of Rep (p<0.001). Most importantly, the set of integration sites for AAV vectors in the absence of Rep did not show any difference of integration site preference compared to random control sites (Figure 5A). With a frequency of 15,707 sites per human genome the Rep binding motif GAGC GAGC occurs sufficiently frequent to lead to a mean distance of around 50 kb to the next AAV integration site in the presence of Rep. In the absence of Rep the mean distance to AAV (vector) integration sites rises to around 130 kb (Figure 5A). To ensure that the presence of repetitive DNA in the random controls did not lead to a bias in the analysis, an independent control calculation was performed for AAV wt data using AAV vector infection data as background. The high significance level was maintained (data not shown). The significance of the Rep-associated preferential integration near GAGC GAGC sequences was further underlined by the results of similar calculations for the putative Rep-binding motif GAGT GAGC, where no such association was found. Only in the presence of presumably large amounts of Rep (AAV vector transfection, Rep+++) a small effect was seen (Figure 5B). Obviously the GAGT GAGC motif is not sufficient to attract Rep and the AAV genome for integration. When an additional GAGC repeat is added (GAGY GAGC GAGC) the integration preferences of AAV wildtype and Rep-expressing AAV vectors shifted to closer proximity to Rep-binding sites (p<0.0001). This is especially surprising since only 616 sites per human genome are found for GAGY GAGC GAGC (Figure 5C). To allow more potential Rep-binding site permutations, calculations were repeated with the consensus GAGC GAGC GAGC with one or two random mismatches. This led to a significantly decreased mean distance to AAV junctions in spite of the fact that up to 100-fold more genomic hits were found for the motifs (Figure 5D; E). A single nucleotide exchange in the GAGY GAGC GAGC motif (Figure 5F, GAGY GAGC GAGA) on the other hand led to a complete loss of association to AAV integration sites. This is surprising in view of the reported in vitro binding of Rep to this motif [45] and supports the assumption that the C at the 3′ end of the Rep binding motif is relevant for Rep-binding in vivo. Motifs GCCC GAGT GAGC and GAGT GAGC ACGC are part of the RBS in the viral p5 promoter. The individual motifs are found at very low frequency (n = 85, or n = 82, respectively) in the human genome, so that either no RBS was found in the same contig or the distance to the next RBS was more than several thousands kb. For these reasons we did not proceed with calculations for these motifs. To further exclude the possibility that the calculated associations with Rep binding sites were predominantly based on sequences assigned to the hotspots AAVS1 and AAVS2, the significance of the associations was re-evaluated with data sets omitted for the hotspot sequences (Table 3). The robustness of the data becomes evident by the fact that the highly significant association of AAV junctions to motifs GAGC GAGC and GAGY GAGC GAGC is maintained. In summary, AAV prefers integration sites in the vicinity of consensus Rep-binding elements, most prominently on chr. 19q13.42 (AAVS1), chr. 5p13.3 (AAVS2), and chr. 3p24.3 (AAVS3). But even in the absence of hotspots AAV still shows a highly significant integration preference for Rep-binding motifs at numerous additional sites in the human genome.
This study represents the first genome-wide survey of wildtype AAV-2 integration in the human genome combined with a thorough bioinformatic analysis of the surrounding genome. We show here that wildtype AAV-2 infection leads to preferential integration in the vicinity of consensus Rep-binding sites (RBS) at defined hotspots as well as at numerous additional genomic sites. In contrast, AAV-2 vectors in the absence of Rep-expression integrate without discernable preference for consensus Rep-binding sites.
At the hotspot on chr. 19q13.42, up to 10% of all AAV junctions were scattered over a region of 33 kb, mostly in centromeric direction with regard to the previously defined core 4 kb AAVS1 site. AAV vectors in the absence of Rep expression do not show any preference for chr. 19q13.42 [46]. The here identified, novel hotspot AAVS2 on chr. 5p13.3 displayed roughly 8% of all junctions retrieved from wildtype AAV-2 infection and 23% of those retrieved from cotransfection of AAV vectors in the presence of Rep distributed over a region of 14 kb. A cluster of 13 independent junctions was found within 14 bp of the AAVS2 RBS that was shown to be similarly proficient in binding to Rep in vitro as is the RBS of AAVS1 (Figure 4). The high in vivo integration numbers may in part be due to the choice of HeLa as target cells. These are hypertriploid with up to 12 copies of the p-arm of chr. 5 [54]. The extra gain of integrations within the described 8 kb region is however unique for the AAVS2 site and not accompanied by a parallel increase of integrations at additional sites on the overrepresented p-arm of chr. 5, where 201 additional GAGC GAGC repeats and three additional GAGY GAGC GAGC repeats were counted. The only fourfold tetranucleotide repeat on the chr.5 p-arm is found in AAVS2 (GAGT GAGT GAGC GAGC; Figure 2C). In addition, junctions of rep-deficient AAV vector were reported to be underrepresented on chr. 5 [46].
A major difference between the hotspots on chr. 5 and chr. 19 concerns the presence of genes. The junctions identified on chr. 19 span the region of the transcribed gene for protein phosphatase 1, regulatory subunit 12C (PPP1R12C). The 8 kb AAVS2 sequence identified on chr. 5p13.3 represents an intergenic region to the best of current knowledge. It is well known that Rep expression leads to extensive rearrangements of AAVS1 [18], [55], [56]. Apparently, PPP1R12C is essential, since the majority of latently infected cell lines display gene duplications [57] and simultaneous AAV integrations in both alleles have never been reported. A currently unresolved question concerns the presence of a terminal resolution site (trs) next to the RBS of AAVS2 and AAVS3. In AAVS1 the spatial configuration of RBS and trs resembles that of the AAV-ITR. The trs element lies next to the RBS and serves as a nicking site for Rep [4]. In AAVS2 and AAVS3 the nearest perfect trs elements (5′-GTTGG-3′) are 400 and 500 bp away from the RBS, which represents the mean statistical occurrence for this motif. Unfortunately, the consensus nucleotide requirements for a functional trs element are not defined well enough to conduct a meaningful bioinformatic search. Therefore, the presence of nicking sites next to the RBS in AAVS2 or AAVS3 remains open at present.
Besides the identified integration hotspots numerous additional chromosomal junction sites were found for integrated wildtype AAV-2, scattered over the human genome. From the bioinformatic calculations it appeared that the perfect tetranucleotide repeat GAGC GAGC represented the minimal requirement for Rep-dependent targeted integration, and GAGY GAGC GAGC represents the optimized in vivo target sequence for wildtype AAV-2. Hotspots AAVS1, AAVS2, and AAVS3 display this core sequence fused to additional imperfect GAGY repeats. Other AAV serotypes display RBS sequences with similar numbers of GAGC and/or GAGT repeats, extended by additional imperfect repeats. AAV5 Rep co-crystallised with the hairpin-structured AAV5-ITR revealed that five Rep monomers bind to five consensus tetranucleotide repeats of the RBS, each of which was contacted by two Rep monomers from opposite faces of the DNA [58]. AAV2-Rep78/68 was shown to simultaneously bind to the RBS of the AAV-2 ITR and to that of AAVS1 [10]. Although it is currently unknown whether other AAV serotypes integrate at all, this is highly likely in view of the ability of both AAV-2 Rep and the relatively distant AAV-5 Rep to multimerize and simultaneously bind to clustered GAGY repeats.
In the initial descriptions of AAVS1, site-specific nicking of the trs by Rep bound to the adjacent RBS was viewed as preferred entry site for AAV recombination [4]. Meanwhile the majority of AAV integrations on chr. 19q13.42 were found many kb away from the RBS-trs combination, and neither AAVS2 or AAVS3 display obvious trs homologues next to the RBS. Therefore alternative explanations for RBS-dependent AAV integration should be considered. The potential use of preexisting chromosomal breakage sites recalls a mechanism already proposed for the integration of rep-deficient AAV vectors [34], [59]. Alternative integration concepts include the use of imperfect trs elements for nicking as shown in vitro [4], [60], [61], or the ability of Rep78 to induce DNA damage in vivo by single-strand nicking of cellular chromatin [19]. It is conceivable that the introduction of single-strand nicks occurs anywhere in accessible chromatin, even if the nicking site is hundreds or thousands of bp apart from the RBS on an extended DNA strand. HMGB1, an ubiquitous architectural protein that serves as key component of the chromatin remodelling complex may be of help [62]. Its long-known in vivo interaction with Rep [63] may help remodel the chromatin to make it accessible for nicking by Rep. Rep was also shown to contact other key players of the nucleosome remodelling complex as components of the transcription- or DNA replication machinery [64], [65], [66]. Any of these mechanisms can be exploited to open the chromatin for AAV integration. In summary, Rep with its combined DNA-binding and endonuclease activity appears to serve as a relatively imprecise targeting tool for AAV integration preferably in open chromatin regions in the reach of consensus Rep-binding sites prevalent in the human genome.
The early finding that Rep would mediate site-specific AAV integration on chr. 19q13.42 had immediate implications for gene therapy. A variety of concepts were devised to incorporate Rep as an adapter to target AAV-ITR flanked transgenes to a specific site [26], [27], [28], [57], [67]. In the majority of cases appropriate cell selection or PCR for AAVS1 led to cells displaying the desired integration. The reported high frequencies of integration into AAVS1 are difficult to reconcile with our findings, unless the level of Rep expression is considered to have an impact on target site choice. Upon AAV infection Rep is only moderately expressed due to autoregulation of the AAV p5 promoter. Rep-dependent AAV vector transductions typically use strong heterologous promoters that lead to high and sustained Rep expression levels. Increasing Rep levels may increase the overall probability for integration anywhere in the genome, including at hotspots. Under these conditions AAVS1-specific integration will be detected more readily. This appears however to come at the price of genomic rearrangements in reach of alternative Rep-binding sites. Therefore, it is plausible that in the absence of any selection AAV integration into AAVS1 is typically unstable and difficult to detect.
In summary, Rep expression increases the probability for integration next to one of several genomic hotspots. However, the net genotoxic effect is unpredictable both with respect to the integrity of the AAV integration locus itself and with respect to the numerous additional sites where Rep binds and initiates chromosomal damage. Therefore, the current concept of a relatively precise site-specific targeting of AAV should be extended to a concept of a relative preference for accessible chromatin regions in the neighbourhood of any of the numerous consensus Rep-binding sites. More recent approaches for site-specific vector targeting try to exploit DNA sequence-specific zinc-finger nucleases to target a genomic sequence of wish [68]. Although zinc-finger nucleases are not free from off-target genotoxicity, at least the genomic targeting site for the transgene can be more precisely defined, a goal that appears to be inherently unachievable using Rep as an adapter molecule.
Detroit 6 cells harbouring latent AAV-2 genomes and HeLa cells were grown in Dulbecco's modified Eagles's medium (Gibco) supplemented with 10% fetal calf serum, penicillin (100 U/ml), and streptomycin (100 µg/ml).
Viral stocks of wildtype AAV-2 with infectious titers of 5×109 i.u./ml were prepared on HeLa cells as described before [16]. For the analysis of AAV integration sites 1.7×106 HeLa cells were seeded overnight on 10 cm diameter dishes and infected with AAV-2 at a MOI of 500. Cells were harvested at 96 hours post infection (p.i.) for the extraction of genomic DNA. The period of cell growth after infection was minimized to reduce the chances of selection of particular integration sites during cell proliferation. Alternatively, AAV-infected HeLa cells were seeded to microtiter plates at a dilution of 60 cells per plate and grown up as single-cell clones without drug selection.
Plasmid pTAV2-0 covers the AAV-2 wildtype genome (GenBank accession number AF043303), pRVK the 4 kb fragment of the AAVS1 locus on chromosome 19 (GenBank accession number S51329), and pAAVS1-TR covers an AAV-ITR/AAVS1 junction [16]. Plasmid pMBP-Rep78 encoding Rep78 fused to maltose-binding protein (MBP) was described before [69].
MBP-Rep78 encoding Rep78 fused to maltose-binding protein was expressed und purified essentially as described [69]. Briefly, E.coli strain BL21 transformed with pMBP-Rep78 was grown at 30°C to an OD600 nm of 0.6 to 0.8. Production of MBP-Rep78 was induced with 0.3 mM IPTG for 3 h at 30°C. Cells were harvested by centrifugation and lysed by sonication for 2 min (30% duty cycle) in lysis buffer of 50 mM phosphate pH 7.8, 300 mM NaCl, 1% (v/v) Triton X-100, 0.1 mM PMSF. Cell debris was removed by centrifugation at 6500×g for 20 min at 4°C. The supernatant was adsorbed to amylose resin (New England Biolabs) in a batch process and the resin was washed extensively (5 washes with about 100 volumes of the resin) with lysis buffer. The adsorbed proteins were eluted with lysis buffer containing 10 mM maltose and analyzed for purity by SDS-polyacrylamide gel electrophoresis.
Binding of MPB-Rep78 fusion protein to 32P- labeled double-stranded oligonucleotide probes was detected by altered mobility of the probes in nondenaturating polyacrylamide gels essentially as described previously [70]. Briefly, oligonucleotides of 46–49 nt length were end-labeled with T4 polynucleotide kinase and annealed. EMSA reactions were performed for 20 min at 20°C as follows: 0.015 pmol of labeled DNA substrate was incubated with the indicated amounts of MBP or MBP-Rep78 in a binding buffer containing 25 mM HEPES-KOH (pH 7.8), 10 mM MgCl2, 40 mM NaCl, 1 mM DTT, 2% glycerol, 12.5 µg/ml BSA, 0,01% Nonidet P40 and 5 µg/ml salmon sperm DNA. The following oligonucleotides were used:
AAV-ITR (nucleotide position 85–133): GCCTCAGTGAGCGAGCGAGCGCGCAGAGAGGGAGTGGCCAACTCCATCA;
AAV-ITR complementary strand: TGATGGAGTTGGCCACTCCCTCTCTGCGCGCTCGCTCGCTCACTGAGGC
Chr. 19q13.42 (AAVS1): TGGCGGCGGTTGGGGCTCGGCGCTCGCTCGCTCGCTGGGCGGGCGGGC
Chr19 (AAVS1) complementary strand: GCCCGCCCGCCCAGCGAGCGAGCGAGCGCCGAGCCCCAACCGCCGCCA
Chr. 5p13.3 (AAVS2): AGCTGGACCCCACGCTCGCTCACTCACTCTCCCCTCACCGCTTTGT
Chr. 5 (AAVS2) complementary strand: ACAAAGCGGTGAGGGGAGAGTGAGTGAGCGAGCGTGGGGTCCAGCT
Chr. 3p24.3 (AAVS3) GCTTCCCAAGGGGAATGAATGTGCGCTCGCTCACTCACTCACTCCTCAC
Chr.3 (AAVS3) complementary strand: GTGAGGAGTGAGTGAGTGAGCGAGCGCACATTCATTCCCCTTGGGAAGC
Chr. 5MUT (AAVS2 mutated): AGCTGGACCCCACCCTCGCTCCCTCCCTCTCCCCTCACCGCTTTGT
Chr.5MUT (AAVS2 mutated), complementary strand: ACAAAGCGGTGAGGGGAGAGGGAGGGAGCGAGGGTGGGGTCCAGCT
AAV p5 (nucleotide position 245–292): TCACGCTGGGTATTTAAGCCCGAGTGAGCACGCAGGGTCTCCATTTTG
AAV p5 complementary strand: CAAAATGGAGACCCTGCGTGCTCACTCGGGCTTAAATACCCAGCGTGA
random control: CAGAGCAGCAGCACAGACGCTAGCAGATCTCCTGCGACCGGAGATGTG
random control, complementary strand: CACATCTCCGGTCGCAGGAGATCTGCTAGCGTCTGTGCTGCTGCTCTG
Total genomic DNA was extracted by SDS/proteinase K digestion followed by repeated phenol/chloroform extractions and ethanol precipitation, as described before [71]. High molecular weight DNA (2 µg) was digested with restriction enzymes that lead to a mean genomic fragment size of around 4 kb and produce blunt-ends ready for linker/adapter ligation. Non-cut enzymes for AAV-2 DNA were preferred, PvuII, EcoRV. Additional junctions were retrieved with DraI (one cut in AAV-2 wildtype DNA). Digested genomic DNA was purified by repeated phenol-chloroform extractions and precipitated with ethanol.
A linker-based strategy described in [39], [40] and outlined in more detail in the manual of the GenomeWalker kit (Clontech) was modified as outlined in Figure 1B. The following oligos were used for linker construction: “Linkerlong” (5′GTA ATA CGA CTC ACT ATA CGG CAC GCG TGG TCG ACG GCC CGG GCT GGT 3′) and “linkershort” (5′ACC AGC CC 3′modifikation: 2′,3′-dideoxyC). Equal amounts of “linkerlong” and phosphorylated “linkershort” (100pmol each) were annealed and ligated to restriction enzyme-digested genomic DNA.
PCR-primers: The linker-primers were “P linker outside” with biotin attached to its 5′ end (5′-GTA ATA CGA CTC ACT ATA CGG C; Tm = 58.4°C) and “P linker nested” (5′-ACT ATA CGG CAC GCG TGG T; Tm = 58.8°C). Two AAV-2-specific primer sets were used. The first primer set covered the AAV p5 promoter: “AAV2p5” (5′-TCA AAA TGG AGA CCC TGC GTG CTC A; Tm = 64.6°C, AAV-2, nt 293–269), primer “AAV2p5 nested” (5′-TAA ATA CCC AGC GTG ACC ACA TGG TG; Tm = 64.8°C, AAV-2, nt 260–235). The other primer set is located in the cap gene region, as described before [2]: “CAPgsp1” (5′-GTC TGT TAA TGT GGA CTT TAC TGT GGA CAC; Tm = 65.4°C, AAV-2, nt 4320–4349) and “CAPgsp2” 5′-GTG TAT TCA GAG CCT CGC CCC AT; Tm = 64.2°C, AAV-2 nt 4357–4379).
The PCR reaction contained 0.2 mM dNTPs, linker primer and AAV specific primer (0.25 µM, each), 2.5 U proofreading hot-start polymerase (Herculase) in reaction buffer, as provided by the supplier (Stratagene). Of the preceding linker-ligation reaction 1–5 µl was added to a final volume of 50 µl. PCR conditions were as follows: 3 min at 98°C, followed by 10 cycles of 40 sec at 98°C, 30 sec at 65°C, and 4 min at 72°C, followed by 20 cycles of 40 sec at 98°C, 30 sec at 65°C, and 4 min + 10 sec per cycle at 72°C, terminated by an extension period of 10 min at 72°C. Biotin-labelled PCR products were further enriched on streptavidin-labelled Dynabeads M-280, as outlined by the supplier (Invitrogen). Subsequent nested PCR used conditions identical to the first round but with pairs of the nested PCR primers, as outlined above. Finally, to add overhangs of multiple As, PCR products were incubated with 1 U Taq polymerase (New England Biolabs).
Products of LSM-PCR reactions were separated on agarose gels. To ensure sufficient chromosomal fragment lengths, PCR bands of a calculated minimal length (>500 bp) were excised and purified by the QIAEX II Gel extraction kit (Qiagen, Hilden, Germany). TOPO-TA cloning was performed as described [72]. Colonies were PCR-screened with the M13 forward (-20) and reverse primer pair (0.4 µM, each) with 0.2 mM dNTP, 2 U Taq polymerase (New England Biolabs) at the following conditions: 10 min at 94°C, followed by 30 cycles of 30 sec each at 94°C, 52°C, and 72°C, followed by 10 min at 72°C. Column-purified PCR products were submitted to DNA sequencing using the primer provided by the TOPO-TA cloning kit. DNA sequences were run on a CEQ2000 genetic analysis system (Beckman) using the CEQ Dye Terminator Cycle Sequencing Quick start kit (Beckman) and the run method LFR-a. Cycling conditions were as follows: 1 min at 96°C, followed by 30 cycles 20 sec at 96°C, 20 sec at 50°C and 4 min at 60°C.
The genomic positions of AAV integration sites in the human genome (assembly from March 2006, hg18) were determined using the BLAT tool from the UCSC Genome Browser web site (http://genome.ucsc.edu/cgi-bin/hgBlat) [73]. A match was defined as a BLAT search result fulfilling all of the following criteria:
In addition to the LSM-PCR derived sequences, the original DNA sequence files of 157 chromosomal junctions [42] kindly provided by Dr. G.W. Both, North Ride, Australia were reanalyzed applying the above inclusion criteria. This led to 47 DNA sequences suitable for our analysis (Table 1). In their study, HeLa cells had been cotransfected with plasmids for constitutive RSV-promoter-driven Rep78 expression and for recombinant AAV vectors expressing a SV40-promoter-driven neomycin gene [42]. Furthermore, 1100 DNA sequences from a published analysis of rep-deficient AAV vector integration sites in diploid human cells [46] were reanalyzed. Since the PCR methods employed in our study and in the one by Drew et al. [42] cannot detect the matching left and right junction sites generated by one AAV integration event, only one chromosomal junction was analyzed per rescued provirus. The original DNA sequence files (DU711025.1 to DU709854.1) of Miller et al. [46] were downloaded from the Genome Survey Sequences (GSS) Database of NCBI (http://www.ncbi.nlm.nih.gov/sites/entrez?db=nucgss) and reanalyzed using the UCSC March 2006 human genome build. The analysis led to a total of 450 junction sequences that fulfilled all of the above inclusion criteria for bioinformatic comparisons. For the subsequent data analysis we implemented software in C++ using the software library SeqAn [74] and several Python scripts.
For different Rep binding motifs, we computed the average distance of virus integration sites to the closest occurrences of Rep binding motifs within the genome. We supposed that the observed integration events were independent from each other and the sample size was high enough for assuming the distance to be normally distributed. To assess whether these distances differ significantly from expectation, several background models were generated:
The generation of both, the data analysis and the background model was confined to those genomic contigs that contained at least one Rep binding motif, since otherwise the distance to the “closest Rep binding motif” would not be defined. A given set of AAV integration sites was considered to be significantly closer to Rep binding motifs than expected by chance, if the significance was calculated for all relevant background models. Data sets of AAV vectors were analyzed with the “random” background model. We applied a Z-test for determining statistical significances for the distances of integration sites to Rep binding sites. For comparing integration sites from AAV wildtype infection sites against those from rep-deficient AAV vector infection we applied the Student's t-test.
AAV integration sites were examined for the occurrence of various genomic features using tables available in the UCSC database. For the determination of significant divergences from expectations, we compared the actual integration sites with a set of 100,000 randomly chosen control sites in the human genome using a two-tailed binomial test.
Chromatin immunoprecipitation sequencing (ChIP-Seq) data were used to define the state of histone modifications in genomic regions of AAV integration. H3K27me3 domains determined by Cuddapah et al. were used as markers for closed chromatin (http://www.wip.ncbi.nlm.nih.gov/projects/geo/query/acc.cgi?acc=GSM325898) [75]. Regions enriched for H3K4 methylation (open chromatin) were determined as follows: The raw ChIP-Seq reads by Robertson et al. [76] (http://www.bcgsc.ca/data/histone-modification) were mapped to the human genome using Bowtie [77], and peaks were called using MACS [78]. H3K4me1/3 domains are then defined as 5 kb windows around the centers of the peaks.
|
10.1371/journal.pcbi.1002171 | High-Precision, In Vitro Validation of the Sequestration Mechanism for Generating Ultrasensitive Dose-Response Curves in Regulatory Networks | Our ability to recreate complex biochemical mechanisms in designed, artificial systems provides a stringent test of our understanding of these mechanisms and opens the door to their exploitation in artificial biotechnologies. Motivated by this philosophy, here we have recapitulated in vitro the “target sequestration” mechanism used by nature to improve the sensitivity (the steepness of the input/output curve) of many regulatory cascades. Specifically, we have employed molecular beacons, a commonly employed optical DNA sensor, to recreate the sequestration mechanism and performed an exhaustive, quantitative study of its key determinants (e.g., the relative concentrations and affinities of probe and depletant). We show that, using sequestration, we can narrow the pseudo-linear range of a traditional molecular beacon from 81-fold (i.e., the transition from 10% to 90% target occupancy spans an 81-fold change in target concentration) to just 1.5-fold. This narrowing of the dynamic range improves the sensitivity of molecular beacons to that equivalent of an oligomeric, allosteric receptor with a Hill coefficient greater than 9. Following this we have adapted the sequestration mechanism to steepen the binding-site occupancy curve of a common transcription factor by an order of magnitude over the sensitivity observed in the absence of sequestration. Given the success with which the sequestration mechanism has been employed by nature, we believe that this strategy could dramatically improve the performance of synthetic biological systems and artificial biosensors.
| Here we recreate in vitro the sequestration mechanism thought to underlie the extraordinary sensitivity (the steepness of the input/output function) of a number of genetic networks. We do so first using fluorescent molecular beacons, a well-established, DNA-based biosensor architecture, as our model system. The experimental parameters that define this in vitro model can be controlled with great precision, allowing us to dissect and test a quantitative model of sequestration in unprecedented detail. Following on this we employ the sequestration mechanism to steepen the binding-site occupancy curve of a common transcription factor by an order of magnitude over the sensitivity observed in the absence of sequestration. Our study thus highlights the versatility with which this approach can be used to improve the performance of both synthetic biological systems and artificial biosensors.
| In order to test the extent to which we understand complex biochemical systems -and to improve our ability to exploit them in man-made technologies (e.g., synthetic biology; biosensors)- it is important to reconstruct these processes in the laboratory. Illustrative examples of this include recent demonstrations of synthetic genetic networks in which genetic elements are “mixed and matched” in order to create artificial bistable “toggle switches,” genetic oscillators and other complex, non-linear input/output behaviors (e.g., [1]-[5]). Other examples include the recent de novo design of proteins, including enzymes, unrelated to any naturally occurring sequences (e.g., [6]-[9]) and the artificial selection of new proteins [10]-[11]. And while these studies do not (and cannot) prove that our knowledge of, for example, genetic regulatory networks and protein folding and evolution is exhaustively complete, they nevertheless suggest that our understanding of these naturally occurring systems is sufficient to enable the design of similarly complex, artificial systems [12].
Motivated by the above philosophy, here we recreate in vitro the “sequestration” mechanism thought to underlie the extraordinary sensitivity (the steepness of the input/output function) of a number of genetic networks (e.g., [5], [13]-[18]). In the sequestration mechanism, low concentrations of a given target molecule are sequestered by binding to a high affinity (low dissociation constant) receptor that acts as a “depletant,” which serves as a “sink” that prevents the accumulation of free target without generating an output signal (Figure 1a). When the total target concentration surpasses the concentration of the depletant (saturating the sink), a threshold response is achieved in which the addition of any further target produces a large rise in the relative concentration of free target (Figure 1b, top). The rapidly rising concentration of free target then binds to –and thus activates– a second, lower affinity (higher dissociation constant) receptor (or “probe”) that, unlike the depletant, generates an output signal. This threshold effect generates a “pseudo-cooperative” dose-response curve, which is much more sensitive (much steeper) than the hyperbolic “Langmuir isotherm” produced by simple, single site binding (Figure 1b, Bottom) [17]-[19].
(At this juncture we must note an important semantic distinction. The sensitivity of biological systems, such as metabolic networks or signal transduction pathways, is defined as the ratio of the relative change in output to the relative change in input [15], [19]. The term ultrasensitivity thus describes systems for which the upstroke of the input/ouput function is steeper than the simple, hyperbolic curve obtained for single site binding, such as is observed for “classic” Michaelis-Menten enzymes [15], [18]. This definition of sensitivity is distinct from “analytical sensitivity,” which represents the smallest input (rather than the smallest change in input) that the method is capable of resolving above the noise floor. Indeed, as we show here, ultrasensitive behaviour is often produced at the cost of a reduced analytical sensitivity as a steeper input/output function is often achieved at the cost of increasing the smallest input signal that can be robustly detected. In this paper we use only the former, steep-input/output-function definition of the terms “sensitivity” and “ultrasensitivity.”).
Sequestration is thought to underlie the ultrasensitive responses of many cellular processes. An example is the depletion of specific messenger RNA by the binding of small regulatory RNA, which generates ultrasensitive thresholds leading, in turn, to the highly sensitive regulation of gene expression [20]-[24]. The binding of many transcription factors is likewise thought to be rendered ultrasensitive via a sequestration mechanism in which high affinity “decoy” binding sites scattered across the genome (or inhibitory proteins that compete for the transcription factor [25]-[26] act as depletants leading to a steep, effectively “all-or-none” transcriptional response [13], [27]-[28]. As a test of this hypothesis, Buchler and co-workers have recently recreated the sequestration mechanism in vivo in a synthetic genetic circuit that converts a graded transcriptional response into an ultrasensitive response via the addition of a depletant [16]. Here we build on their study by recreating the sequestration mechanism in vitro using molecular beacons, a well-established biosensor architecture, as our model system. The experimental parameters that define this in vitro model can be controlled with great precision, providing a means to dissect and test a quantitative model of sequestration in unprecedented detail. Following this we have adapted the sequestration mechanism to steepen the concentration-occupancy relationship for the binding of a common transcription factor by an order of magnitude over the sensitivity observed in the absence of sequestration.
Molecular beacons, synthetic biomolecular switches developed by Kramer and coworkers for the detection of specific DNA or RNA sequences [29], are now widely used in the diagnosis of genetic and infectious diseases [30]-[32]. Consisting of a stem-loop DNA modified with a fluorophore/quencher pair, molecular beacons are quantitatively described by a simple three-state population-shift model in which the equilibrium between a non-binding, non-signaling state and the binding-competent, signaling state is pushed towards the latter upon target binding [33]. This linkage between a conformational equilibrium and target binding allows us to rationally tune the affinity of molecular beacons –without affecting their specificity– by altering the stability of the stem. Indeed, using this approach we have previously shown that it is possible to tune the affinity of molecular beacons across more than 4 orders of magnitude [33]. For the studies reported here we have used a set of six molecular beacons sharing a common recognition element but spanning this same 10,000-fold range of target affinities (Table 1). The input/output function of each of these six molecular beacons is well described by the hyperbolic curve expected for single site binding,(1)
in which F[T] is the fluorescence output as a function of target concentration, [T], F0 and FB are the fluorescence of the unbound and bound states respectively, and Kdprobe is the dissociation constant of the probe/target duplex.
We introduce sequestration into molecular beacons by combining a relatively low affinity signaling probe (i.e., fluorophore/quencher labeled) with an excess of a higher affinity (but unlabeled and thus non-signaling) stem-loop that serves as the depletant (dep) (Figure 1c). Doing so, we convert the hyperbolic binding curve associated with a traditional molecular beacon (Eq. 1) into a much steeper, ultrasensitive response (Figure 2).
A physically reasonable description for the proposed sequestration mechanism is easily derived from the above hyperbolic binding curve by replacing [T] with the effective concentration of free (unbound, un-sequestered) target. As per Buchler and Louis [15], this concentration goes with: (2)
where [T]t is the total amount of target added and Kddep is the dissociation constant of the depletant/target complex. By combining Eq. 1 and Eq. 2 we see that the ratio of the depletant concentration to the probe affinity ([dep]/Kdprobe) is a crucial determinant of ultrasensitivity. To validate this we employed the relatively low affinity molecular beacon 2GCprobe (Kdprobe = 310 nM; table 1) as our probe and the higher affinity unlabeled molecular beacon 0GCdep (Kddep = 5.2 nM) as our depletant. When we do so we observe ultrasensitivity as soon as the depletant concentration rises above the probe dissociation constant (i.e., as [dep]/Kdprobe increases above unity; Figure 2, left). With further increases in depletant concentration the steepness of the dose-response curve increases monotonically to the highest [dep]/Kdprobe ratios we have investigated. Moreover, by applying equations 1 and 2 to these data we see that the sequestration model fits these ultrasensitive responses quantitatively (R2 ≥ 0.998) without the use of any fitted parameters (solid lines, Figure 2). That is, we can quantitatively fit our observations with this model using parameters values –the affinities and fluorescence signals of the two molecular beacons- determined independently in previous studies [33].
The Hill coefficient is commonly employed to describe ultrasensitive systems in biochemistry [34]. And while it is not a physically correct description of sequestration (as it was originally derived to describe allosteric cooperativity), we find that the pseudo-Hill coefficient we obtain by fitting the Hill equation to our data provides a convenient way of comparing the ultrasensitive behavior generated by the sequestration mechanism. As expected, we observe a pseudo-Hill coefficient near unity (0.90±0.02) for a binding curve obtained in the absence of depletant (dotted lines in Figure 2, left). Upon the addition of depletant this value climbs, reaching 1.3 at a [dep]/Kdprobe ratio of 0.3 before ultimately reaching a value of 9.4 at a ratio of 80, the highest [dep]/Kdprobe ratio we have investigated (Figure 2, right). That is, sequestration ultimately compresses the normally 81-fold dynamic range of a molecular beacon (i.e., the transition from 10% to 90% target occupancy spans an 81-fold change in target concentration; see ref. 17) into a 1.5-fold dynamic range, significantly increasing the steepness of the input/output function of the molecular beacon and, in turn, improving its ability to detect small changes in relative target concentration.
Our in vitro model also provides an opportunity to explore, for the first time, the extent to which sequestration-derived ultrasensitivity depends on other parameters, including, for example, the relative affinities of the depletant and the probe (i.e., Kdprobe/Kddep). To do so, we varied the depletants affinity (using Kd ranging from 5.2 nM to 3 µM) at a constant [dep]/Kdprobe ratio of 3.2 (Kdprobe = 310 nM –molecular beacon 2GC- with and a [dep] of 1 µM). As expected we find that, while high affinity depletants (i.e., 0GCdep, 1GCdep) produce clear ultrasensitive responses (pseudo-Hill coefficients of 3.9 and 3.6 respectively), depletants with affinities similar to (i.e., 2GCdep, 3GCdep) or poorer than (i.e., 4GCdep and 5GCdep) those of the probe produce only minor improvements in sensitivity (pseudo-Hill coefficients <1.3) (Figure 3, left). Again, all of the data so obtained are well modeled by Eq. 1 and 2 without the use of any fitted parameters (i.e., by fixing all parameters to the values obtained from independent experimental conditions), providing another high precision test of the quantitative sequestration model (Figure 3, right).
While the two ratios described above, [dep]/Kdprobe and Kdprobe/Kddep, play crucial roles in generating ultrasensitivity they do not work independently of one another. For example, if [dep]/Kdprobe falls well below unity we will not obtain ultrasensitivity no matter how high the Kdprobe/Kddep ratio climbs. This occurs because, when the probe dissociation constant is significantly higher than the concentration of the depletant, the free target concentration at the “threshold” is too low to saturate the probe, leading to a more-or-less hyperbolic response approximating that seen in the absence of depletant (Figure S1). To better understand the interplay between these two ratios (i.e., to illustrate the parameter space over which ultraensitive behavior is obtained) we can plot the pseudo-Hill coefficient as a function of [dep]/Kdprobe and Kdprobe/Kddep (Figure 4). Doing so we find that, when [dep]/Kdprobe falls below 0.91 it is not possible to achieve ultrasensitivity with any reasonable value of Kdprobe/Kddep. Similarly, if Kdprobe/Kddep falls below 0.94 we do not generate a pseudo-Hill coefficient above 2 even at the highest depletant concentration we have employed.
Finally, the ease with which we can manipulate our in vitro system renders it possible to also characterize the effects of varying Kdprobe at a constant depletant concentration. To do so, we increased the length of our target sequence, which lowers the dissociation constants of the probe and depletant for the target by the same extent. Using targets ranging from 13 to 17 nucleotides (and thus increasing the [dep]/Kdprobe ratio from 0.3 to more than 100), we observe the expected monotonic increase in sensitivity (Figure S2). Moreover, these data too, fit equations 1 and 2 quantitatively (R2>0.995) without the use of any adjustable parameters.
In the above studies molecular beacons served as a convenient, synthetic toolkit to quantitatively and precisely test the sequestration model. Obviously, however, molecular beacons are not themselves of any specific biological relevance. We have thus also developed an in vitro system to test the extent to which sequestration can cause pseudo-cooperative, highly sensitive changes in the concentration of free transcription factor, thus increasing the sensitivity with which a transcription factor-binding site is occupied. A difficulty in demonstrating this with precision is that traditional methods for quantifying the occupancy of a transcription-factor binding site, including gene activation, gel shift assays and ELISAs, provide only relatively “low resolution” measurements of site occupancy. As our read-out we have thus instead employed a recently developed, highly precise optical reporter for transcription factor binding activity termed “transcription factor beacon” (Figure 5, left) [35]. Specifically, in order to detect the binding of our transcription factor, TATA binding protein, we have used a transcription factor switch that exhibits a 45 nM dissociation constant for this target, reporting its binding via a large change in fluorescence output. As our depletant we have employed a hairpin DNA that contains TATA binding protein’s double stranded recognition sequence but that lacks a fluorophore/quencher reporting pair. Unlike the transcription factor switch, the hairpin does not undergo any binding-induced conformational change and thus its affinity for TATA binding protein is, as required by the sequestration mechanism, significantly greater than that of the reporting probe. Using a 1∶10 mixture of this probe/depletant pair we achieve a pseudo-Hill coefficient of 4.3, compressing the normally 81-fold psuedolinear range of the occupancy of this transcription factor binding site to a mere 4-fold and thus significantly increasing the sensitivity with which it is occupied (Figure 5, right).
Using molecular beacons and transcription-factor binding as model systems we have recreated, in vitro, the sequestration mechanism that Nature employs to generate ultrasensitive behavior in many genetic networks. Doing so we have demonstrated that the simple, quantitative model proposed by Buchler and Louis accurately and precisely predicts the relationships between ultrasensitivity and the concentrations and affinities of the depletant and probe. We have also demonstrated, more generally, the utility of employing DNA-based in vitro models in the high precision testing and dissection of biologically important regulatory mechanisms.
As noted above, Buchler and co-workers [15], [16] were among the first to test the sequestration mechanism using a synthetic genetic circuit in vivo that they converted from a graded transcriptional response into an ultrasensitive output via the addition of a depletant [16]. Their work confirmed earlier suggestions that sequestration could underlie bistable or oscillatory circuits in natural regulatory systems. It also highlighted the important determinants of the sequestration mechanism. Due to the intrinsic complexity of in vivo systems, however, it proved difficult to use this model system to test all the determinants of sequestration with high precision. Buchler and co-worker, for example, were unable to evaluate the range of Kdprobe/Kddep over which various degree of ultrasensitive behavior could be observed. Here, in contrast, we have employed molecular beacons, a well-defined, easily controllable, in vitro system, as a tool to dissect the sequestration mechanism [15], [16] in more detail and with greater precision than has hitherto proven possible.
While the impressive specificity, affinity and versatility of biomolecular recognition have motivated decades of research in the development of sensors and other biotechnologies based on this effect [36], the hyperbolic –and thus not particularly sensitive– concentration/occupancy curves characteristic of single site binding limits their precision. This, in turn, limits the utility of these biotechnologies in many applications. Given this we believe that the use of sequestration in vitro may be of use in increasing the sensitivity of synthetic biosystems, such as biosensors, in vitro. Specifically, we have shown that it possible to narrow the 81-fold pseudo-linear dynamic range of the well-known molecular beacon platform by almost 2 orders of magnitude and of a transcription factor switch sensor by a factor of 20. The modified sensors so produced exhibit ultrasensitive responses equivalent to Hill coefficients of greater than 9 and greater than 4 respectively, converting them into high precision analytical approaches. Given that sequestration requires only the availability of depletants that bind the target in question with greater affinity than that of the signal-generating probe, we anticipate that the mechanism can be adapted to many other biotechnologies, an argument bolstered by the frequency with which this mechanism is employed in the cell [15], [27], [28]. Moreover, the sequestration mechanism appears more straightforward to implement than the other mechanisms Nature has employed to achieve improved sensitivity. It appears far easier to implement, for example, than positive allosteric cooperativity as the later would involve the design of probes containing two or more precisely interacting sites for target binding [37]-[38]. Despite these advantages, the sequestration strategy is not without a limitation: the generation of ultrasensitive response is achieved at the cost of a reduced affinity, which shifts the minimum target concentration producing a detectable signal (the detection limit) towards higher concentrations.
The extremely steep input/output functions demonstrated here would appear to open the door to new applications across biosensors and synthetic biology. Perhaps the most obvious application would be in the monitoring of, for example, drugs with such narrow therapeutic windows that only high precision measurements of their concentration achieve clinical relevance. More speculatively, the availability of sensors that, in contrast to the graded (analog) outputs of most biosensors, produce an effectively “all-or-none” (digital) response may be useful in the development of molecular logic gates [39]-[43]. These, in turn, may enable the development of molecular-scale computers and “autonomously regulated” chemical systems, ideas that have attracted significant recent interest [44], [45].
The following HPLC-purified constructs were purchased from Sigma-Genosys and used as received (the bases in italic are those constituting the stem):
1GCprobe: 5’-(FAM)-A-CTATT-GATCGGCGTTTTA-AATAG-G -(BHQ)-3’
2GCprobe: 5’-(FAM)-A-CTCTT-GATCGGCGTTTTA-AAGAG-G -(BHQ)-3’
3GCprobe: 5’-(FAM)-A-CTCTC-GATCGGCGTTTTA-GAGAG-G -(BHQ)-3’
where FAM and BHQ represent 6-carboxyfluorescein and Black Hole Quencher respectively.
The following HPLC-purified, un-modified depletants and targets were purchased from Sigma-Genosys and were used as received (the bases in italic are those constituting the stem):
0GCdep: 5’-A-TTATT -GATCGGCGTTTTA-AATAA-G -3’
1GCdep: 5’-A-CTATT -GATCGGCGTTTTA-AATAG-G -3’
2GCdep: 5’-A-CTCTT -GATCGGCGTTTTA-AAGAG-G -3’
3GCdep: 5’-A-CTCTC-GATCGGCGTTTTA-GAGAG-G -3’
4GCdep: 5’-A-CTCGC-GATCGGCGTTTTA-GCGAG-G -3’
5GCdep: 5’-A-CGCGC-GATCGGCGTTTTA-GCGCG-G -3’
13-base target: 5’-TAAAACGCCGATC-3’
15-base target: 5’-TTAAAACGCCGATCA-3’
17-base target: 5’-TTTAAAACGCCGATCAA-3’
HPLC purified DNA modified with FAM and internal BHQ-1 inserted on a thymine residue was purchased from Biosearch Technologies (Novato, CA) and has the following sequence:
TFprobe: 5’-(FAM)-TACTT TTATATAAAT AAGT T(BHQ) GTGA TTTTTATATATT TCAC -3’
The following HPLC-purified, un-modified depletant was purchased from Sigma-Genosys and was used as received:
TFDep: 5’-CGTATATAAAGG TTTTTTT CCTTTATATACG -3’
This protein was expressed recombinantly, purified, and characterized as previously described [46].
All fluorescent experiments were conducted at pH 7.0 in 50 mM sodium phosphate buffer, 150 mM NaCl, at 45 °C. For all experiments with TATA binding protein, the buffer was supplemented with 5 mM MgCl2, as it is essential for efficient DNA binding and the measurement were conducted at 25 °C. Equilibrium fluorescence measurements were obtained using a Cary Eclipse Fluorimeter with excitation at 480 (± 5) nm and acquisition at 517 (± 5) nm. Binding curves were obtained using solutions of 10 nM of labeled molecular beacons (or TF switch) and varying concentrations of unlabeled beacons (or DNA binding protein recognition sequence) as depletant and by sequentially increasing the target concentrations via the addition of small volumes of solution with increasing concentrations of target. Dissociation constants of the labeled beacons were obtained from the literature [33].
|
10.1371/journal.pntd.0003053 | Visceral Leishmaniasis and HIV Co-infection in Bihar, India: Long-term Effectiveness and Treatment Outcomes with Liposomal Amphotericin B (AmBisome) | Visceral Leishmaniasis (VL; also known as kala-azar) is an ultimately fatal disease endemic in the Indian state of Bihar, while HIV/AIDS is an emerging disease in this region. A 2011 observational cohort study conducted in Bihar involving 55 VL/HIV co-infected patients treated with 20–25 mg/kg intravenous liposomal amphotericin B (AmBisome) estimated an 85.5% probability of survival and a 26.5% probability of VL relapse within 2 years. Here we report the long-term field outcomes of a larger cohort of co-infected patients treated with this regimen between 2007 and 2012.
Intravenous AmBisome (20–25 mg/kg) was administered to 159 VL/HIV co-infected patients (both primary infections and relapses) in four or five doses of 5 mg/kg over 4–10 days. Initial cure of VL at discharge was defined as improved symptoms, cessation of fever, improvement of appetite and recession of spleen enlargement. Test of cure was not routinely performed. Antiretroviral treatment (ART) was initiated in 23 (14.5%), 39 (24.5%) and 61 (38.4%) before, during and after admission respectively. Initial cure was achieved in all discharged patients. A total of 36 patients died during follow-up, including six who died shortly after admission. Death occurred at a median of 11 weeks (IQR 4–51) after starting VL treatment. Estimated mortality risk was 14.3% at six months, 22.4% at two years and 29.7% at four years after treatment. Among the 153 patients discharged from the hospital, 26 cases of VL relapse were diagnosed during follow-up, occurring at a median of 10 months (IQR 7–14) after discharge. After accounting for competing risks, the estimated risk of relapse was 16.1% at one year, 20.4% at two years and 25.9% at four years. Low hemoglobin level and concurrent infection with tuberculosis were independent risk factors for mortality, while ART initiated shortly after admission for VL treatment was associated with a 64–66% reduced risk of mortality and 75% reduced risk of relapse.
This is the largest cohort of HIV-VL co-infected patients reported from the Indian subcontinent. Even after initial cure following treatment with AmBisome, these patients appear to have much higher rates of VL relapse and mortality than patients not known to be HIV-positive, although relapse rates appear to stabilize after 2 years. These results extend the earlier findings that co-infected patients are at increased risk of death and require a multidisciplinary approach for long-term management.
| Fifty percent of all visceral leishmaniasis (VL) cases globally occur in India, where up to 90% of cases occur in the state of Bihar. There are also an estimated 300,000 people in Bihar living with HIV/AIDS. Patients with HIV who are treated for VL typically have much worse outcomes than VL patients who are HIV-negative, yet there exists very little evidence suggesting more effective treatments for this group. Between 2007–2012, with support of the Rajendra Memorial Research Institute (RMRI), Médecins Sans Frontières (MSF) treated 8,749 VL patients in Bihar using liposomal amphotericin-B (AmBisome). Here we describe the characteristics and long-term outcomes of a subgroup of 159 HIV-VL co-infected patients treated within this program over the 5-year period. Their estimated mortality risk was 14.3% at six months after treatment, 22.4% at two years and 29.7% at four years. Estimated risk of relapse was 16.1% at one year, 20.4% at two years and 25.9% at four years. We conclude that treatment of HIV-VL co-infected patients with 20–25 mg/kg of liposomal amphotericin-B is well tolerated and relatively effective. However, HIV-VL co-infection is a complex chronic disease with high early mortality and much worse outcomes than VL alone, and requires a multidisciplinary long-term management strategy.
| One third of all HIV patients worldwide live in regions where leishmaniasis is endemic [1]. Visceral leishmaniasis (VL) caused by the parasite L. donovani is endemic to Bihar, a populous state of 110 million people in East India, which carries an estimated 40% of the world's VL burden [2]. Although Bihar has a relatively low prevalence of HIV (between 0.22–0.33%), its high population density means that in absolute numbers an estimated 300,000 people in the state live with HIV/AIDS [3].
Moreover, Bihar is one of the few states in India where the rate of new HIV infections is increasing [4]. This has major implications for VL co-infection: like other opportunistic infections in HIV patients, Leishmania amastigotes have evolved strategies to survive and multiply within macrophages [5], which are enhanced by HIV co-infection [6] and accelerate progression of disease [7]. This may help explain why the risk of developing VL is estimated to be between 100 and 2300 times higher in HIV-infected individuals than in those who are HIV-negative. [8]. Data on the prevalence of HIV-VL co-infection in India is scarce, although estimates range from 2–5.6% [9]–[14]. HIV-VL co-infection therefore appears to be a growing public health issue in India.
Yet the evidence base regarding best treatment practices for co-infected patients is limited, due to a lack of randomized controlled trials and to the fact that most available data comes from observational studies with relatively short follow-up periods and often with high rates of loss to follow-up [15]. Nevertheless, worse outcomes in almost every respect have consistently been reported in this patient group when compared to patients not known to be HIV-positive—for example, in terms of higher relapse rates, mortality, and VL drug toxicity and resistance [15].
Currently the Indian treatment guidelines for VL do not differentiate treatment of HIV-VL co-infected patients from that of other patients presenting with VL. First-line treatment for all VL patients in India is 28 days of oral miltefosine (where not contra-indicated), although the government is currently assessing the use of single-dose AmBisome and lower-dose combinations therapies [16] as recommended by the World Health Organization (WHO) [17]. However, India has not developed a contingency plan for HIV-VL patients.
Since 2007, Médecins Sans Frontières (MSF) has collaborated with the Rajendra Memorial Research Institute (RMRI) and the National Vector Borne Disease Control Program (NVBDCP) to implement a VL treatment program within Ministry of Health (MoH) facilities in Vaishali district, one of the most highly endemic areas for VL in Bihar, The program has treated over 8,500 patients using 20 mg/kg liposomal amphotericin B (AmBisome, Gilead Pharmaceuticals, Foster City, CA, USA). High-dose liposomal amphotericin B is currently recommended by WHO for first-line treatment of HIV-VL co-infection [17]. We have exclusively used AmBisome, a brand name for liposomal amphotericin B, since it is the only preparation of this medication that has received stringent regulatory approval for use in VL [16].
In this retrospective observational cohort study of routinely collected program data, we describe the baseline characteristics of the 159 HIV-VL co-infected patients treated with liposomal amphotericin B in the Bihar program between July 2007 and August 2012. We then describe the outcomes for VL immediately after treatment and in the longer term (up to 5 years), the latter being crucial to monitor given the chronic nature of HIV infection.
In collaboration with RMRI, MSF developed a comprehensive VL program that was fully integrated into the MoH facilities in Vaishali district, Bihar. This program encompassed two main activities: the running of an MSF-led inpatient unit within the district hospital, and the provision of logistic support for VL services to five Primary Health Centers (PHCs) within Vaishali district. From February 2007 until August 2012, 8,749 patients were treated for VL using 20 mg/kg liposomal amphotericin B as first-line therapy, as described in detail elsewhere [18]. Approximately 50% of these patients originated from Vaishali district, while the remainder travelled from adjacent districts. Over time, rising numbers of patients entered the program via referrals from neighbouring district hospitals. In particular, HIV treatment centers outside the district increasingly referred patients who had been diagnosed with HIV and were thought to have VL. During the 5-year study period, the VL program treated 159 HIV co-infected patients, of whom 60 were directly referred by outside HIV treatment centers (Figure 1).
Data collected for all patients diagnosed with VL included general demographic information, clinical history, hemoglobin level, height, weight, and malaria rapid diagnostic test result. The study also recorded information on ‘caste’, a form of social stratification used in India, using the following categories and definitions: scheduled caste and scheduled tribe (terms used for two groups of historically disadvantaged people recognized in the Constitution of India); other backward class (a collective term used by the government of India for castes that are educationally and socially disadvantaged but not specifically mentioned in the Constitution); and general category (not considered to be disadvantaged). The first three groups combined account for approximately 60% of India's population.
All patients with a history of 2 weeks fever and clinical splenomegaly were considered suspect for VL. Diagnosis was confirmed with the rK39 rapid diagnostic test (DiaMed-IT LEISH; DiaMed AG, Cressier, Switzerland). Patients presenting with a history suggestive of relapse or with atypical clinical signs or negative diagnostic tests but a high index of suspicion of VL were referred to the RMRI for parasitological diagnosis through splenic or bone marrow biopsy. Of the 159 co-infected patients, 31 had the diagnosis of VL confirmed solely with rk39 serological testing, with the remainder through parasitological visualization on biopsy using established techniques [19].
At the start of the program, only patients with a history suggestive of possible HIV exposure were offered provider-initiated counseling and testing (PICT) for HIV. Indications for PICT included a history of relapse, a high-risk profession or being a migrant worker, but were otherwise not clearly defined. Patients self-reporting a previous diagnosis of HIV were counseled and re-tested, as were patients referred from external hospitals presenting with a provisional diagnosis of HIV-VL co-infection. However, a few patients treated for VL and discharged from the program subsequently re-presented with confirmed relapse, at which point they were tested and diagnosed with HIV infection. Therefore, the HIV testing policy was changed in March 2011, after which all patients aged ≥14 years were offered PICT for HIV.
HIV testing at the program facility was initially performed using two rapid diagnostic tests in parallel (SD Bioline HIV 1/2 and Determine-HIV 1/2), with patients testing positive referred to MoH testing facilities for further diagnosis as per local protocols using Combaids Advantage, TriLine and TriSpot RDTs. However, use of SD Bioline was stopped in December 2011, in keeping with WHO recommendations [20]. From that point onwards, patients testing positive with Determine-HIV 1/2 tests alone were referred. Discordant tests were confirmed with Western Blot.
Patients with or without HIV who presented to the program for the first time were treated under the same protocol, using 20 mg/kg intravenous liposomal amphotericin B given in 4 doses over 4–10 days depending on the severity of illness. Patients considered to be in good clinical condition were treated on 4 consecutive days, while those requiring a longer period of inpatient observation received the 4 doses over 10 days. All patients diagnosed with a VL relapse and who had previously received the 20 mg/kg liposomal amphotericin B regimen at the program were treated with an increased dose (25 mg/kg) of liposomal amphotericin B following parasitological confirmation of relapse. All patients presenting with VL relapse were offered PICT, Patients were considered ‘initial cures’ once they completed a full course of VL treatment and showed clinical improvement, cessation of fever, reduction of spleen size and return of appetite at the time of discharge following WHO descriptions of treatment response [21].
Test of Cure (ToC) was not routinely performed, due to the risks associated with splenic puncture and in light of a previous study showing a cure rate of >98% at 6 months [22], Instead, splenic or bone marrow aspiration was reserved for confirmation of VL in all patients presenting with relapse, those with suspected initial treatment failure, and initially, for all HIV-VL co-infected patients. However, as neither relapses nor treatment failures occurred in any of the 55 HIV-VL patients treated during the first 6 months, [23], from that point onwards ToC was not routinely performed in this group.
Following completion of VL treatment and improvement of their general condition, patients with HIV–VL co-infection were offered antiretroviral treatment (ART) at either the RMRI or within Vaishali district hospital, using the national program-recommended regimen of stavudine, lamivudine and nevirapine. However, as a growing number of national program centers providing ART opened in Bihar over time, responsibility for initiation and maintenance of ART was transferred to these more local facilities, which helped create a more sustainable, patient-friendly treatment strategy. Communication between the VL treatment program and the ART treatment centers was maintained, and patients with suspected VL relapse were referred back to the MSF program for diagnosis and treatment. In addition, the centers shared information on ART patient adherence and CD4 counts, where available.
After successful completion of VL treatment, patients were asked to return to the MSF program for follow-up at 1, 3, 6, 12 and 24 months from the time of VL treatment initiation, They were also counseled at discharge regarding the high risk of relapse and the importance of adherence to ART. Secondary prophylaxis against VL, which is recommended elsewhere [8], [17], was not offered. The program made considerable effort to maintain long-term contact with patients though active telephone tracing and, in cases where contact through telephone tracing or ART treatment centers failed, through home visits.
All data were entered into a standard Microsoft Excel database; double data-entry was not done at the time of inputting. Regular database cleaning consisted of checks for inconsistencies, with reference to source documents where necessary. Although an epidemiologist ensured the database was well maintained and regularly audited the quality of data transfer, all records of co-infected patients were reviewed again immediately before the final analysis to ensure that data entered into the database was correct.
Nutritional status (Body Mass Index, BMI) was assessed using weight and height data for patients ≥19 years of age, while World Health Organization Anthro and Anthro Plus software (Geneva, Switzerland) was used to calculate a BMI-for-age Z-score for those aged ≥5–19 years and Weight for Height (W/H) Z-score for those 6 months to <5 years of age. For patients ≥19 years of age, severe acute malnutrition (SAM) and moderate acute malnutrition (MAM) were defined as BMI<16 and 16–<17, respectively. For patients aged ≥5–19 years, SAM was defined as BMI-for-age Z-score <−3 and MAM as <−2 but >−3 SD, while for patients aged 6 months–<5 years, SAM and MAM were defined as W/H Z-score <−3 and <−2 but >−3 SD, respectively.
Statistical analysis of data was conducted using STATA version 11 (STATACorp LP, College Station, USA). For the final analysis all data were anonymised. Baseline characteristics for all co-infected patients were compared against VL patients not known to be HIV-positive and treated in the program over the same time period. Primary outcomes were time to death and time to relapse. Person-time at risk was calculated for each patient, starting from the date of VL treatment initiation up to the date of death, date of the last visit (for those lost to follow-up), or 31st August 2013 (for everyone else). With relapse as outcome, follow-up time started with hospital discharge and ended at the date of (first) relapse for those with relapse. The cumulative incidence of mortality or relapse was estimated using Kaplan-Meier methods. A risk factor analysis was performed using multivariate Cox regression modelling. Variables considered for inclusion were age, sex, a history of VL, ART use and the following characteristics (all at the time of VL diagnosis): hemoglobin level, body mass index, spleen size, CD4 cell count, concurrent tuberculosis, and duration of illness (only in relapse analysis).
For those initiating ART, the overall change in CD4 count levels after diagnosis of HIV-VL co-infection was visualized using a nonparametric method called LOWESS smoothing (for locally weighted scatterplot smoothing, ‘lowess’ command in STATA). This provides a representative smooth curve through data using robust local regression.
In the mortality analysis, ART use was categorized as either 1) being on ART at the time of VL diagnosis, or 2) ART initiation after VL diagnosis, included as a time-varying covariate. With relapse as outcome, ART use was categorized as 1) being on ART at the time of VL diagnosis; 2) ART initiation during admission: 3) ART initiation sometime after admission, included as a time-varying covariate. In the main analysis, variables associated with the outcome with a P-value <0.05 in univariate analysis were included in multivariate analysis. The model was reduced by backward stepwise elimination until all variables had a P-value <0.05. In secondary analysis, ART use was forced in the model.
In the main analysis, multiple imputation was used for missing data [24]. In sensitivity analysis 1, we used the missing indicator method (whereby for a specific predictor a separate category is generated for missing data). Continuous co-variables were categorized in the main analysis but included as continuous variables in sensitivity analysis 2, with the functional form determined using the multivariable fractional polynomial (mfp) models command in STATA. Several other secondary and sensitivity analyses were also conducted, including the removal of patients with incomplete information on ART use in models involving this parameter.
In addition, the cumulative incidence of relapse was recalculated to allow for the presence of competing risks (death precluding the occurrence of relapse), since in this case standard survival methods can lead to biased estimates [25], [26]. The proportional hazard assumption was assessed graphically and tested formally using Schoenfeld residuals. Co-linearity was evaluated by calculating the variance inflation factors. The level of significance was set at P<0.05.
This analysis met the Médecins Sans Frontières Institutional Ethics Review Committee criteria for a study involving the analysis of routinely collected program data. Although AmBisome is a new treatment in the Indian setting, it is a recognized treatment for VL; moreover, the program was run in coordination with the State Health Society through a memorandum of understanding, which is the usual procedure for NGOs operating in this context. The HIV-VL clinical treatment guideline had been reviewed and approved by the RMRI Institutional Ethics Committee. All electronic data were analyzed anonymously.
A total of 159 HIV/VL co-infected patients were treated for VL during the 5-year MSF-supported program (July 2007 to August 2012). Of these, 150 (94.3%) were treated with 20 mg/kg liposomal amphotericin B, while 8 (5%) were treated with a 25 mg/kg regimen. All patients completed treatment with no discontinuations or serious adverse events, and no deaths were associated with liposomal amphotericin B treatment. Four patients were lost to follow up within one month after VL diagnosis, one at six months and one at twelve months. The maximum length of follow-up was 5.6 years following completion of VL treatment, with a mean of 2 and median of 1.2 years.
The demographic and baseline clinical characteristics of the 159 co-infected patients compared to the remaining treated cohort of 8,590 patients not known to be HIV-positive are shown in Table 1. The Relative Risk (RR) of being HIV-positive was 3.7 times higher for males than for females, while the RR of being HIV-positive was 25 times lower (0.04, 0.02–0.07) for patients <25 years of age compared to those aged 25–<55 years. The mean ±SD age was 36.6±10.4 years (range 7–70); 145 (91.2%) patients were aged 25–55 years, of whom 64 (44.1%) were aged 35–45 (Table 1). The RR of being co-infected with HIV and of a General Category caste was 2.2 (95% Confidence Interval 1.6–3.1) times higher than compared to being of an Other Backward Class or Scheduled Caste. Compared to patients not known to be HIV-positive, those with HIV-VL co-infection tended to have a lower hemoglobin (<6 g/dl) on admission (RR 1.7, 1.02–2.7, p = 0.04) and a greater degree of splenomegaly >6 cm (RR 2.1, 1.6–2.9, p<0.001).
There was no significant difference between the global nutritional status of patients known to be HIV-positive and the remainder of the cohort (43.3% vs 40.8% globally malnourished respectively, RR (95%CI) = 1.1 (0.8–1.5), p = 0.54), nor was there a significant difference in the prevalence of severe acute malnutrition (SAM) between patients known to be HIV-positive and the remainder of the cohort – 23.9% vs 17.9%, respectively, RR = 1.4 (95%CI 0.95–2.0, p = 0.095).
However, the RR of presenting with a relapse of VL was particularly high – the odds of being HIV-positive and having previously experienced a single or multiple episodes of VL prior to admission was 16.6 times higher than in the overall cohort (95% CI 12.4–22.4; p<0.001).
Of the 159 co-infected patients, 60 (37.7%) had been diagnosed with HIV prior to attending the MSF program, of which less than half (23) were receiving ART at time of diagnosis of VL. The remaining 99 (62.3%) patients were diagnosed with HIV at the time of VL diagnosis. 122 (76.7%) of patients had CD4 counts recorded between 6 months before or one month following treatment for VL, with a mean count of 122 cells/uL, and a median of 111 (IQR 59-193). Nearly half (n = 56, 46%) had counts <100 cells/uL. Patients already on ART at time of treatment for VL had a median CD4 count of 188 cells/ul (IQR 54-164), while patients not on ART had a median CD4 count of 101 cells/ul (IQR 75-234). Patients presenting with primary or a previous episode of VL had a median CD4 counts of 108 cells/uL (IQR 57-163) and 113 cells/uL (IQR 61-216) respectively. A total of 9 (5.7%) patients were suffering from tuberculosis in addition to HIV-VL co-infection.
A total of 36 co-infected patients died, including six who died shortly after admission. Death occurred at a median of 11 weeks (Inter-Quartile Range, IQR 4-51) after starting VL treatment. The estimated mortality risk was 14.3% at six months, 22.4% at two years and 29.7% at four years after diagnosis (Figure 2). In univariate analysis, a low BMI (<16 kg/m2), low hemoglobin (<7 g/dL) at admission, and concurrent tuberculosis were each associated with an increased risk of mortality (Table 2). Use of ART was associated with a decreased risk.
In multivariate analysis, concurrent tuberculosis and low hemoglobin were independent risk factors. This remained true when ART use was included the model. ART use was associated with a 64–66% reduced risk of mortality, but the effect only reached statistical significance when ART was initiated after VL diagnosis (Table 3). Minor changes in the estimates were observed in the sensitivity analyses.
Although all patients completed treatment and showed clinical improvement, six patients died following a period of prolonged admission, likely due to multiple contributing factors. From the remainder, there were no documented cases of treatment failure based on treatment response. Among the 153 patients discharged from the hospital, a total of 26 cases of VL relapse were diagnosed during follow-up, occurring at median of 10 months (IQR 7-14) after discharge. The estimated risk of relapse was 1.6% at six months after discharge, but subsequently increased to 18.5% at one year and 23.8% at two years (Figure 3). Four years following VL treatment the risk was 31.2%.
In univariate analysis, age >40 years was associated with an increased risk of relapse (Table 4). This effect showed borderline significance when ART use was included in the model. The association of ART use with a reduced risk of relapse (75% reduction) was statistically significant only for ART initiation during admission.
We also examined whether CD4 cell count recovery following VL treatment was associated with the risk of relapse. As shown in Figure 4, CD4 recovery was blunted in patients who subsequently relapsed compared to those who remained relapse-free. The median CD4 count of patients who relapse and did not subsequently relapse following treatment was 95 cells/uL, (IQR 63-163) versus 112 cells/uL, (IQR 57-206), respectively. Of the 26 patients who subsequently relapsed, 16 had a CD4 count recorded around the time of relapse, with a median count of 137 cells/uL (IQR 80-255).
Accounting for competing risks, the estimated risk of relapse decreased to 16.1% at one year, 20.4% by two years and 25.9% by four years. The main findings remained unchanged by the other sensitivity analyses (Table 5).
This study describes the admission characteristics and long-term VL treatment outcomes for the largest cohort of HIV-VL co-infected patients from the Indian subcontinent, with a longer follow-up period and lower rate of loss-to-follow-up than any report published to date. As in previous studies, our data show high mortality in these patients, particularly in the early period following diagnosis, and a high VL relapse rate, findings which underscore the crucial importance of early diagnosis and intervention for both diseases.
Early initiation of ART had a clear impact on reducing mortality and relapse, and should therefore be considered a key intervention in the management of these patients. In agreement with our earlier estimates of 2-year outcomes for the first 55 HIV-VL patients in this cohort [23], outcomes for this much larger number of co-infected patients were substantially worse than for VL patients not known to be HIV-positive: after receiving the same VL treatment in the same setting, estimated all-cause mortality and relapse rates at 15 months for patients not known to be HIV positive were 2.8% and 1.2% respectively [18], compared to 18.1% and 16.1% at 12 months respectively for co-infected patients reported in this study.
Concurrent infection with tuberculosis and hemoglobin <7 g/dl were independently associated with mortality. In terms of protective factors, ART initiated immediately following VL treatment was associated with a 64–66% reduced risk of mortality (p<0.05). Similar reductions in mortality risk for co-infected patients adherent to ART have been reported in other studies [8]. Our data suggested that ART use prior to VL diagnosis may also be associated with reduced mortality, but this association did not reach statistical significance. This lack of a demonstrated effect of prior ART could be due simply to the relatively small number of patients, or could reflect the possibility that patients already on ART at time of VL diagnosis may have been experiencing ART treatment failure or more advanced disease – or conversely that those with more favorable responses to ART may be at much lower risk of developing VL and therefore never enroll in the program, leading to an underestimation of the effect of ART. Notably, baseline CD4 counts around the time of VL diagnosis were typically very low in our cohort (mean baseline CD4 count 122 cells/ul).
In terms of VL relapse risk, we did not detect clear demographic associations, in agreement with results from a recent systematic review looking at predictors of VL relapse in HIV-positive patients [27]. However, initiation of ART immediately following VL treatment was associated with a significant reduced risk of relapse (although not if initiated either before or long after VL treatment), while a history of one or more previous VL episodes at time of treatment was not. These results differ from those seen in the meta-analysis, which concluded that ART did not appear to reduce the risk of relapse whereas a previous history of VL was predictive of relapse. However, other than one from Ethiopia, all studies included in this meta-analysis were conducted in Europe, the majority with small sample sizes and limited follow-up periods.
The Ethiopian study alone involved VL caused by L. donovani, as in our Indian setting. Infection with L. donovani has different clinical implications compared to L. infantum, the causal agent in most European and Latin American VL cases, and therefore more relevant to HIV/VL management in India. The study found that ART was partially protective against VL relapse, while a baseline CD4 count of <100 cells/uL and a history of two or more relapses were associated with increased risk of relapse [28], however findings may have been biased by the high proportion of patients not receiving ART who were lost to follow up. In contrast, predicting relapses in India appears more complex as there did not appear to be any effect of either baseline CD4 count or of previous VL history.
This study has a number of limitations. Primarily, although admission and VL treatment data had relatively few missing values, data from the HIV management perspective was incomplete; as follow-up periods extended past 3 years, the number of available CD4 counts decreased, which prevented further accurate modeling. Second, a larger sample size may have yielded more precise estimates for both risk factors and measures of outcomes. Third, the prevalence of HIV-VL co-infection cannot be estimated from this study, since all patients were not systematically screened for HIV, and it is likely that a substantial number of co-infected patients were missed in the overall treated cohort.
Another limitation was that we considered all-cause mortality in the analysis, therefore excluding the possibility that death may have occurred due to other causes unrelated to HIV-VL. A further weakness is that the analysis included 5% (n = 8) of the patients who received a 5 mg/kg higher dose of AmBisome than the remainder. Lastly, although no initial treatment failures were seen in patients discharged from the program, it is likely that the routine use of ToC would have identified treatment failures that were missed clinically. It is unclear what the value of partial response patterns (eg partial but not complete regression of splenomegaly) is in determining true treatment response, particularly in co-infected patients. An Ethiopian study with systematic ToC found 32% parasitological failure in co-infected patients after treatment with 30 mg/kg liposomal amphotericin B despite good clinical response [29].
Co-infected patients show decreased cellular and humoral response to Leishmania parasites and are considered difficult to achieve a definitive cure from VL. As such, suspicion of relapse is more challenging in co-infected patients, since these patients often have persistent haematological abnormalities and residual hepato-splenomegaly at the end of treatment. Indeed, worsening of these abnormalities in the absence of fever may itself represent a new episode of VL, and as such it is plausible that there was under-reporting of relapse cases due to the importance given to fever in the routine diagnosis of symptomatic VL.
The findings from this cohort analysis have a number of implications for improving the outcome of HIV-VL co-infected in India. Recent studies in the Indian subcontinent have recommended increasing the routine follow-up period after VL treatment from 6 months to 1 year [30]–[32]. However, for HIV-VL co-infected patients, it appears that the risk of relapse is greatest within 18 months following treatment, suggesting that routine follow-up should be extended even further for co-infected patients. Furthermore, if secondary prophylaxis is to be initiated, this period might be the most effective phase for its use.
Achieving longer follow-up without loss of many patients will/would require some changes to current practice, since maintaining long-term contact with patients who complete treatment is not integrated into existing VL programs, and without an existing framework, requires significant effort and resources. Similarly, there are no established mechanisms for sharing information about co-infected patients between the vertical VL and HIV programs in India. If such mechanisms were developed, they could facilitate more robust longer-term patient management, as has been seen in other co-infections, such as HIV/TB.
Although routine PICT for VL patients is recommended by WHO in areas where HIV counseling and access to ART are available [17], this service is lacking across the majority of endemic areas in India and is not included in the national VL program guidelines. Conversely, screening for VL in HIV-infected patients who have spent a significant amount of time in VL endemic areas is not mentioned by existing National AIDS Control Organisation (NACO) guidelines. We suggest that directives encouraging early diagnosis of co-infection are crucial as a means of reducing the high early levels of mortality observed in this study.
NACO guidelines recommend initiation of ART in all patients with clinical stage IV disease irrespective of CD4 count. However this recommendation refers to WHO guidelines, which identify ‘atypical disseminated visceral leishmaniasis’ as a stage IV defining opportunistic infection [33], rather than simply ‘visceral leishmaniasis’. This leads to confusion in the field when making decisions to start ART in co-infected patients, considering the WHO expert committee on VL clearly identifies HIV-VL co-infection as an AIDS defining illness [1]. Simultaneously, in the absence of national guidelines, maintaining consistent health messaging between parallel programs for HIV and VL is challenging. Reported non-adherence to ART regimens in India varies considerably, from 14%–86% [34]; in this study, 23 (14.9%) of patients either chose to discontinue, died prior to starting or did not start ART despite being referred to appropriate care providers. The provision of field-based guidance and training for the management of HIV-VL co-infection, as already exists for HIV-TB co-infection [35], [36], could be of great benefit in raising health provider awareness and improving management of these patients.
This study suggests that 20–25 mg/kg liposomal amphotericin B is a well-tolerated and relatively effective treatment for HIV-VL co-infection in the Indian setting. However, these patients have a high risk of relapse, and clearly, repeated treatment with mono-therapy in cases of relapse may not be ideal as it may contribute to decreased drug susceptibility in the parasite [37]. Mechanisms for resistance to amphotericin B in clinical isolates of L. donovani have already been described [38], and decreased efficacy observed in co-infected patients after several treatment cycles [39], [40]. Additionally, unresponsiveness to liposomal amphotericin B seemed to develop rapidly in co-infected patients In Ethiopia where parasitological failure rates were 16% in primary HIV-VL, and 57% in relapse HIV-VL previously treated with AmBisome [29]. However, to date no parasite strains resistant to liposomal amphotericin B have been found, suggesting host-related factors may play a more important role in treatment unresponsiveness than parasite resistance. Although higher dose combination therapy has been recommended in cases of multiple VL relapse in co-infected patients [8] and has been successfully used in India [41], the use of such combinations needs to be further evaluated in the Indian subcontinent for all HIV-VL co-infected patients.
Within the field of TB-HIV co-infection, over the last 15 years there have been a number of observational studies conducted to understand the effect of ART on TB mortality, and the effect of timing of ART initiation. Combined, these allowed a clearer picture to emerge, which in turn contributed to the design of several clinical trials on the subject. In the absence of other studies from the Indian subcontinent within the field of HIV-VL co-infection, the data from this program constitutes a clear step forward, however highlights the need for additional studies to consolidate the evidence base and allow triangulation of different study findings.
Like patients with Post Kala-Azar Dermal Leishmaniasis (PKDL), HIV-VL patients harbor chronic infection, often have very high parasite loads and are therefore potential long-standing reservoirs for VL transmission. The role of asymptomatic VL infection has not yet been definitively established [42], however it is likely that an increase in HIV prevalence in endemic areas will lead to an associated increase in symptomatic VL infections. As such the importance of early identification, appropriate treatment, multidisciplinary management and follow-up of HIV-VL co-infected patients should be considered a public health priority if the goal of VL elimination is to be realized [43].
|
10.1371/journal.pgen.1007832 | Mek1 coordinates meiotic progression with DNA break repair by directly phosphorylating and inhibiting the yeast pachytene exit regulator Ndt80 | Meiotic recombination plays a critical role in sexual reproduction by creating crossovers between homologous chromosomes. These crossovers, along with sister chromatid cohesion, connect homologs to enable proper segregation at Meiosis I. Recombination is initiated by programmed double strand breaks (DSBs) at particular regions of the genome. The meiotic recombination checkpoint uses meiosis-specific modifications to the DSB-induced DNA damage response to provide time to convert these breaks into interhomolog crossovers by delaying entry into Meiosis I until the DSBs have been repaired. The meiosis-specific kinase, Mek1, is a key regulator of meiotic recombination pathway choice, as well as being required for the meiotic recombination checkpoint. The major target of this checkpoint is the meiosis-specific transcription factor, Ndt80, which is essential to express genes necessary for completion of recombination and meiotic progression. The molecular mechanism by which cells monitor meiotic DSB repair to allow entry into Meiosis I with unbroken chromosomes was unknown. Using genetic and biochemical approaches, this work demonstrates that in the presence of DSBs, activated Mek1 binds to Ndt80 and phosphorylates the transcription factor, thus inhibiting DNA binding and preventing Ndt80’s function as a transcriptional activator. Repair of DSBs by recombination reduces Mek1 activity, resulting in removal of the inhibitory Mek1 phosphates. Phosphorylation of Ndt80 by the meiosis-specific kinase, Ime2, then results in fully activated Ndt80. Ndt80 upregulates transcription of its own gene, as well as target genes, resulting in prophase exit and progression through meiosis.
| Sexual reproduction requires that cells deliberately introduce large numbers of double strand breaks into their chromosomes. Repair of these breaks creates physical connections between homologs that promote proper segregation during meiosis. It is critical that segregation not proceed until all the breaks have been fixed. How does the cell determine when sufficient double strand break repair has occurred? Our work provides a mechanistic explanation to this question. The meiosis-specific Mek1 kinase is activated by double strand breaks. High numbers of breaks result in high Mek1 activity, resulting in phosphorylation of the meiosis-specific Ndt80 transcription factor. Negative charges conferred by phosphorylation prevent Ndt80 from binding the promoters of its target genes, including genes necessary for completing recombination and meiotic progression, thereby preventing their transcription. As breaks are repaired, Mek1 kinase activity decreases and the inhibitory phosphorylation on Ndt80 is lost, allowing Ndt80 to activate transcription of its target genes. As a result, crossover formation is completed and intact chromosomes proceed properly through the meiotic divisions.
| One of the most dangerous things for a cell is the occurrence of DNA double strand breaks (DSBs) in its chromosomes. Failure to repair a DSB may result in a loss of genetic material and lethality. DSBs arise due to exogenous damage such as radiation, or endogenous errors such as stalled replication forks. Repair of DSBs by non-homologous end joining may lead to deletions, translocations or inversions, which can have adverse consequences such as cancer [1]. The most conservative way to repair a DSB is by homologous recombination, using the sister chromatid as the template. Indeed, in mitotically dividing cells, homologous recombination mediated by the evolutionarily conserved recombinase, Rad51, is biased towards using sister chromatids [2, 3].
DSBs trigger an evolutionarily conserved DNA damage checkpoint, which delays or arrests cell cycle progression to provide time for repair [4]. The DNA damage checkpoint is mediated by two kinases, Tel1 (ATM in mammals), which responds to blunt ends, and Mec1 (ATR in mammals) which is activated by single stranded DNA generated by resection of the 5’ ends of the breaks. In yeast, these kinases phosphorylate the adaptor protein, Rad9, which in turn recruits the Forkhead-associated (FHA)-domain containing effector kinase, Rad53, (related to Chk2 in mammals), resulting in Rad53 autophosphorylation and activation. Rad53 phosphorylation of various proteins then prevents cohesin destruction and mitotic exit.
While the purpose of mitosis is to produce genetically identical daughter cells, the specialized cell division of meiosis divides the chromosome number in half to produce gametes for sexual reproduction. After premeiotic chromosome duplication, meiotic chromosomes segregate twice without an intervening round of DNA synthesis: homologous pairs of sister chromatids go to opposite poles at Meiosis I (MI), while sister chromatids separate at Meiosis II (MII). Proper alignment at Metaphase I requires tension that is generated when sister kinetochores from one homolog attach to the spindle pole opposite that of the other homolog. This tension occurs because homologs are physically connected by a combination of crossovers (COs) and sister chromatid cohesion [5]. COs are initiated by programmed DSBs generated by Spo11, a meiosis-specific, evolutionarily conserved topoisomerase-like protein that cuts in preferred regions of the genome called “hotspots” [6]. Unlike mitotic cells, meiotic DSB repair is biased to use the homolog as the repair template [7].
It is key that every pair of homologs contains at least one crossover. Towards this end, many more DSBs are generated during meiotic prophase than the number of necessary COs (e.g.,yeast, 160 DSBs/16 homolog pairs; mouse, 250–300 DSBs for 20 homolog pairs) [6]. The repair of these breaks must be carefully regulated to ensure not only the requisite number of COs, but also that no breaks remain when Anaphase I begins. The meiotic recombination checkpoint delays meiotic prophase while interhomolog recombination is occurring. This checkpoint uses meiosis-specific modifications to the DNA damage checkpoint and is dependent upon protein components of a specialized chromosomal structure called the synaptonemal complex (SC) [8–10].
After chromosome duplication in yeast, cohesin complexes containing the meiosis-specific Rec8 kleisin subunit hold sister chromatids together [11]. Sister chromatids condense along protein cores containing Rec8, as well as the meiosis-specific Hop1 and Red1 proteins, to form axial elements (AEs). Hop1 contains the evolutionarily conserved HORMA domain which mediates homo-oligomerization, as well as interaction with Red1 [12–15]. Chromosome condensation occurs by the formation of chromatin loops, with axis proteins at their bases [16, 17]. Spo11 is indirectly recruited to the axes by phosphorylation of the DSB protein, Mer2 [17–19]. In addition, Mer2 interacts with Spp1, which binds to trimethylated histones flanking hotspot sequences to bring the hotspots to the axis [17, 20, 21]. DSB formation on the loops therefore occurs in the vicinity of Hop1 and Red1 on the axis.
COs created by DSB repair are primarily generated using a functionally diverse set of proteins collectively called the ZMM proteins (Zip1-3, Zip4/Spo22, Msh4, Mer3, Msh5, and Spo16) [22, 23]. Holliday junctions formed by the ZMM pathway exhibit biased resolution to form COs that are distributed throughout the genome [22, 24, 25]. The ZMM pathway is also necessary to form stable associations between homologs, leading to the insertion of the transverse filament protein, Zip1, between the AEs to create the tripartite SC [22, 26, 27]. At the pachytene stage of meiotic prophase, all the homolog pairs are fully synapsed.
Similar to vegetative cells, meiotic DSBs result in the recruitment and activation of the Tel1 and Mec1 checkpoint kinases. These kinases phosphorylate Hop1, which replaces Rad9 as the adaptor [28]. Phosphorylated Hop1 is bound by the FHA domain of the meiosis-specific paralog of Rad53 and Chk2, Mek1 (also known as Mre4), resulting in Mek1 oligomerization and activation by autophosphorylation in trans [28–31]. Chromatin-immunoprecipitation experiments using phosphorylation of Histone H3-T11 as a marker for Mek1 activity revealed that this activity is highest at axis sites that correlate with the presence of Hop1 and Red1 and can spread for several kilobasepairs (kb) surrounding a DSB [32].
Mek1 is a key regulator of meiotic DSB repair. It promotes interhomolog bias by inhibiting Rad51 from interacting with its accessory factor, Rad54 in two ways: (1) phosphorylating and stabilizing Hed1, a meiosis-specific protein that binds to Rad51, thereby excluding Rad54 and (2) phosphorylating Rad54 which reduces its affinity for Rad51 [33–36]. These mechanisms prevent Rad51 from competing with the meiosis-specific recombinase, Dmc1, which mediates the bulk of meiotic recombination [37, 38]. Mek1 antagonizes sister chromatid cohesion locally at DSBs to facilitate strand invasion of homologs and regulates whether interhomolog recombination intermediates are repaired as either COs or noncrossovers by enabling phosphorylation of Zip1 by the Cdc7-Dbf4 (DDK) cell cycle kinase [39, 40]. Finally, MEK1 is required for the meiotic recombination checkpoint delay that prevents cells from entering into the meiotic divisions with unrepaired DSBs [41–44].
Checkpoint delay is part of the normal meiotic program, but this delay can be exacerbated in mutants that initiate, but fail to complete, DSB repair. An extreme case occurs in dmc1Δ diploids in the SK1 strain background, where strand invasion does not occur because Dmc1 is absent and Rad51 activity is inhibited by Mek1 [34, 45–47]. The high number of DSBs generates high levels of activated Mek1, resulting in meiotic prophase arrest due to a lack of Cdc28-Clb1 (CDK-Clb1) activity [8, 41, 45, 48, 49]. The checkpoint inhibits CDK-Clb1 by two separate mechanisms: (1) activation and stabilization of the Swe1 kinase which places an inhibitory phosphate on tyrosine 19 of Cdc28 [50] and (2) inactivation of the meiosis-specific transcription factor, Ndt80, thereby preventing CLB1 transcription [51–53]. During early meiotic prophase, a meiosis-specific E3 ligase targets mitotic regulators such as polo-like kinase (Cdc5) and Clb1 cyclin for degradation [54]. As a result, their production is dependent upon the transcriptional activity of Ndt80.
Ndt80 is a sequence-specific DNA binding protein that recognizes a nine-base pair sequence called the middle sporulation element (MSE) in the promoters of >300 target genes (called “middle” and “late genes”) [51, 55, 56]. NDT80 transcription occurs in two stages [57]. In the first stage, expression of NDT80 requires the transcriptional regulator Ime1, which is also responsible for transcribing early genes such as HOP1, MEK1, SPO11 and DMC1 [58]. NDT80 transcription is delayed relative to the early genes, however, because of the Sum1 repressor, which binds to MSEs in the NDT80 promoter and the promoters of Ndt80 target genes [59]. Sum1 removal requires phosphorylation by the meiosis-specific Ime2 kinase, in combination with CDK and DDK [60–62]. Since IME2 is an early gene, it must be transcribed and translated before Sum1 repression can be relieved, hence the delay in Ime1-mediated NDT80 transcription. The relatively low level of Ndt80 protein generated by Ime1 is inhibited by the meiotic recombination checkpoint until sufficient DSB repair has occurred to lower Mek1 kinase levels below the amount necessary to inactivate Ndt80 [41, 51–53]. The second stage of NDT80 transcription is marked by phosphorylation of Ndt80 by Ime2 that facilitates Ndt80’s ability to activate transcription [48, 63, 64]. Ndt80 then activates transcription of its own gene to initiate a positive feedback loop, as well as promoting transcription of target genes such as CDC5. Expression of CDC5 triggers to resolution of Holliday junction intermediates into COs and degradation of Red1 to dissemble the SC [41, 54, 65]. Removal of Red1 leads to inactivation of the remaining Mek1, allowing residual DSBs to be repaired prior to CLB1-promoted entry into Meiosis I [41].
Exit from pachynema and entry into Meiosis I has been proposed to be controlled by a switch between two stable states [54]. In the first state, CDK-Clb1 levels are low due to the meiotic recombination checkpoint, thereby preventing meiotic progression. In the second state, CDK-Clb1 levels are high because DSBs have been repaired, leading to a decrease in the checkpoint signal and activation of Ndt80, thereby allowing CLB1 transcription and progression into the meiotic divisions. What was unknown was how this switch is controlled. This work shows that Mek1, after being activated by DSBs, directly binds and phosphorylates Ndt80, thereby inhibiting Ndt80 from activating transcription. As DSBs are repaired, Mek1 activity decreases, and inhibitory Mek1 phosphosites are removed. Ime2 phosphorylation then promotes Ndt80 activity, resulting in expression of genes necessary for completing recombination and exiting prophase. Mek1 phosphorylation of Ndt80 therefore provides an elegant way for cells to know when it is safe to enter the first meiotic division.
A two-hybrid screen using lexA-MEK1 revealed an interaction with a fragment of NDT80 (amino acids 287–627) fused to the Gal4 activation domain (GAD). This fusion is hereafter referred to as GAD-NDT80. The strain contained HIS3 and lacZ reporter genes under the control of promoters containing lexA operator sites [66]. Two-hybrid interactions were therefore manifested either by growth on medium lacking histidine or production of ß -galactosidase. The GAD-NDT80 fragment begins near the end of the Ndt80 DNA binding domain (DBD) and goes to the end of the protein. In addition to the activation domain in the C terminus, this fragment includes a 57 amino acid sequence in the middle of Ndt80 that is required for meiotic recombination checkpoint arrest (Fig 1A and 1B, row 2)[64, 67–69]. The NDT80-bc allele, which encodes an Ndt80 protein deleted for this 57 amino acid sequence, no longer responds to the checkpoint triggered by unrepaired breaks in both zip1Δ and dmc1Δ mutants (bc stands for “bypass checkpoint”)[69]. We have therefore named this 57 amino acid sequence the “bc” domain. Disruption of the Mek1 FHA domain using the R51A mutation had no effect on the Ndt80 interaction, indicating that the FHA domain does not mediate binding (Fig 1C)[47, 70]. In contrast, deletion of the bc domain from GAD-NDT80 eliminated interaction with lexA-MEK1, even though the GAD-Ndt80-Δbc protein was more abundant than GAD-Ndt80, ruling out protein instability as the reason for the loss of the two-hybrid signal (Fig 1B, row 3 and 1D, lanes 2 and 3). In addition to being necessary for lexA-MEK1 interaction, the bc domain was also sufficient, as the GAD-bc fusion produced a positive two-hybrid signal in combination with lexA-MEK1 (Fig 1B, row 4).
A 60 amino acid sequence containing the bc domain was used to probe Ndt80 proteins from other fungi for homology. A small region containing amino acids 371–375, RPSKR, is conserved in several yeast species (Fig 1E). A consensus motif generated from these alignments showed that lysine (K) 374 and arginine (R) 375 from S. cerevisiae Ndt80 are completely conserved (Fig 1F). Deletion of the sequence encoding RPSKR from GAD-NDT80 abolished the two-hybrid interaction with lexA-MEK1, as did substituting the KR sequence with aspartic acids (DD) (Fig 1B, rows 5 and 7). The KR to alanine (AA) mutant still interacted with lexA-MEK1, although not quite as well as GAD-NDT80 (Fig 1B, compare rows 6 and 2). The RPSKR sequence is therefore required for interaction between lexA-Mek1 and Ndt80.
The Ndt80-Mek1 interaction has thus far not been confirmed by co-immunoprecipitation experiments from meiotic extracts due to technical problems obtaining soluble Ndt80. A functional genetic approach was therefore used to test the importance of this interaction in vivo. Overexpression of NDT80 can partially bypass the meiotic recombination checkpoint arrest triggered by the unrepaired DSBs that accumulate when the DMC1 recombinase is absent [53, 69, 72]. One explanation for this result is that during meiosis in wild-type (WT) cells there is sufficient Mek1 to bind and inactivate all of the Ime1-dependent Ndt80 protein. However, when Ndt80 is in excess of Mek1, some Ndt80 escapes phosphorylation, resulting in transcription of the NDT80 gene to start the positive feedback loop leading to meiotic progression. If this model is correct, and if the bc domain recruits Mek1 to Ndt80 in dmc1Δ-arrested cells, then over-expressing the bc domain by itself could titrate Mek1 away from endogenous Ndt80, resulting in activation of the transcription factor and sporulation. To limit expression of the GAD-bc fusion to meiotic cells, the hybrid gene was placed under the control of the MEK1 promoter. A dmc1Δ diploid transformed with a DMC1 CEN ARS plasmid only partially complemented the sporulation defect, perhaps due to plasmid loss during growth on the Spo plate (Fig 1G). The GAD-bc transformants partially bypassed the dmc1Δ checkpoint arrest, exhibiting increased sporulation after three days on Spo medium compared to GAD alone (Fig 1G). Further support for the titration model is that this partial checkpoint bypass was decreased when counteracted by overexpression of GST-MEK1 from a high copy number plasmid (Fig 1G).
Previous work has shown that the presence of Cdc5 is sufficient to trigger SC disassembly and Red1 degradation [41, 54, 65, 75]. These experiments showed that induction of CDC5 in the ndt80Δ background (where Mek1 levels are low) resulted in the disappearance of Red1 and elimination of the SC [41, 65]. Interfering with the Mek1-Ndt80 interaction in the dmc1Δ background allowed activation of Ndt80 (indicated by increased phosphorylation) and production of Cdc5 in the presence of high levels of Mek1 activity (Fig 3C, dmc1Δ ndt80-ΔRPSKR and dmc1Δ ndt80-KR>DD). In these cases, Red1 persisted for at least two hours after Cdc5 was first detected. In contrast, Red1 was eliminated within two hours after the appearance of Cdc5 in the dmc1Δ mek1Δ diploid (Fig 3C). These results suggest that Cdc5 is not as efficient in targeting the degradation of Red1 in the presence of high levels of Mek1 activity (Fig 3B). Red1 disappears more rapidly in the dmc1Δ ndt80-ΔRPSKR mutant than dmc1Δ ndt80-KR>DD and persisted for the length of the time course in the dmc1Δ ndt80-KR>AA strain (Fig 3C). These differences reflect the larger defect in Mek1 interaction resulting from the deletion of the RPSKR sequence compared to the aspartic acid substitution mutations. Ndt80 was activated more quickly in ndt80-ΔRPSKR (increased phosphorylation at 8 hours) (Fig 3C) so Cdc5 was produced earlier as well. While Mek1 kinase activity delayed Red1 degradation in the presence of Cdc5, it did not prevent it completely. As a result, Mek1 was gradually decreased due to loss of Red1, allowing DSB repair by Rad51 (indicated by loss of Hop1 phosphorylation), leading to a further reduction in Mek1 activity and more efficient Cdc5-dependent degradation of Red1 (Fig 3C). These results have therefore uncovered yet another mechanism to ensure that cells do not enter MI prematurely, i.e., the prevention of Red1 degradation and therefore, SC disassembly, when Mek1 levels are high. The mechanism by which Mek1 inhibits Red1 degradation remains to be determined.
Meiotic time courses were performed with NDT80, NDT80-6A, ndt80-6D and ndt80-R177A diploids. The ndt80-R177A diploid was used as a negative control as the Ndt80-R177A protein is defective in binding to MSEs and therefore is unable to activate transcription either of itself or other NDT80 targets [61, 67, 68, 77]. The ndt80-6D diploid was phenotypically identical to ndt80-R177A, indicating that it also is defective in activating transcription. Both mutants arrested in meiotic prophase, while NDT80 and NDT80-6A exhibited similar kinetics for meiotic progression (Fig 4A). The R177A and 6D diploids entered the meiotic program efficiently, as evidenced by similar levels of phosphorylated Hed1 protein compared to NDT80 and NDT80-6A at the 4-hour time point (Fig 4B)[34]. Whereas the Ndt80 and Ndt80-6A protein levels peaked at six hours and then decreased until they were nearly gone by 10 hours, the R177A and 6D proteins exhibited reduced levels that slowly accumulated throughout the length of the time course (Fig 4B and 4C). This result is consistent with the occurrence of Ime1-driven transcription of the ndt80-R177A and ndt80-6D genes, followed by a failure of the mutant proteins to activate transcription of their own genes. In addition, the R177A and 6D mutants failed to express CLB1 and CDC5, although both proteins were observed for NDT80 and NDT80-6A (Fig 4B).
An alternative explanation for the ndt80-6D phenotypes is that Ndt80-6D is transcriptionally active, but the aspartic acid substitutions destabilize the protein so that there is insufficient Ndt80-6D protein to promote transcription of CDC5, CLB1, etc. This hypothesis was tested using ndt80-6D under control of the GAL1 promoter in a strain containing a GAL4-estrogen receptor fusion (GAL4-ER). The resulting allele (indicated as ndt80-6D-IN) can be induced ectopically by addition of estradiol to the Spo medium [41, 48, 78]. If the aspartic acid residues destabilize Ndt80, then the induced Ndt80-6D levels should be lower compared to Ndt80 and Ndt80-6A. In contrast, if the reduced level of the endogenous Ndt80-6D protein is due to a failure in Ndt80-activated transcription of the ndt80-6D gene, Ndt80-6D levels should be equivalent to Ndt80 and Ndt80-6A, since transcription is now under the control of a heterologous promoter.
Induction of the NDT80-IN alleles after five hours in Spo medium resulted in meiotic progression of the NDT80 and NDT80-6A diploids, while ndt80-6D remained arrested in prophase (Fig 4D). Both the Ndt80-6A and Ndt80-6D proteins were present in greater abundance than Ndt80 throughout the timecourse. Importantly, the Ndt80-6A and Ndt80-6D proteins exhibited similar kinetics of induction and peaked at the same level. However, while the Ndt80-6A protein was nearly gone by 9 hours, the Ndt80-6D protein persisted and exhibited reduced phosphorylation (Fig 4E and 4F). Cdc5 and Clb1 were generated in the WT and NDT80-6A strains but not in ndt80-6D, confirming that constitutive negative charges at Mek1 consensus sites in the DBD impede the ability of Ndt80 to activate transcription (Fig 4E).
Phosphorylation of Ndt80 by Ime2 results in multiple mobility shifts that enhance Ndt80 transcriptional activity [48, 63, 64]. One report found that inactive Ndt80 derived from checkpoint arrested cells was not phosphorylated [53], and suggested that Ndt80 phosphorylation is solely used for activation of the transcription factor. A different group detected a phosphorylation-dependent mobility shift in a dmc1Δ diploid which was not as slow as the Ndt80 mobility shifts observed from WT cells [63], consistent with our hypothesis that Mek1 phosphorylation of Ndt80 is inhibitory. One difficulty with interpreting these experiments is that the checkpoint prevents Ndt80 from activating transcription of itself, and therefore Ndt80 protein levels are low, making the protein more difficult to detect [51–53]. The estradiol-inducible NDT80 system was therefore used to determine whether inactive Ndt80 is phosphorylated.
A dmc1Δ mek1-as NDT80-IN diploid was incubated in Spo medium for 5 hours to arrest cells with unrepaired DSBs. The mek1-as allele encodes an analog-sensitive (as) kinase with an enlarged ATP binding pocket that allows for inhibition of the kinase by addition of the 1-NA-PP1 inhibitor to the Spo medium [47]. NDT80 transcription was induced by addition of estradiol in the presence or absence of Mek1-as inhibitor. Inactivation of Mek1 resulted in loss of phosphorylated Hop1 at the 6-hour time point, consistent with repair of DSBs, loss of Hed1 phosphorylation and efficient meiotic progression (Fig 5A and 5B, +1-NA-PP1). Ndt80 was highly phosphorylated, resulting in production of Cdc5 and destruction of Red1 (Fig 5B, +1-NA-PP1). That this high level of phosphorylation occurs only after Mek1 inactivation suggests that Mek1 activity somehow inhibits phosphorylation of Ndt80 by Ime2. In the absence of inhibitor, Ndt80 was inactive. Only a small fraction of cells entered the meiotic divisions (Fig 5A), phospho-Hop1, phospho-Hed1 and Red1 persisted, and Cdc5 was not detected three hours after induction (Fig 5B, -1-NA-PP1). The activation state of Ndt80 at the 6-hour timepoint was therefore determined by whether Mek1 was active (inactive Ndt80) or inhibited (active Ndt80).
Phosphatase treatment of inactive Ndt80 resulted in the loss of the slower migrating species, producing two predominant bands (Fig 5C). The band indicated as “pr-Ndt80” has previously been interpreted to be unphosphorylated Ndt80, while the fastest migrating band (“Ndt80” in Fig 5C) was said to be a “degradation fragment” [53, 63]. Instead, the latter band more likely represents completely unphosphorylated Ndt80 because (1) it runs close to the molecular weight for unmodified Ndt80 (69 kD); (2) the extracts used for these experiments were fixed with trichloroacetic acid prior to lysis and protease inhibitors were included during lysis, making proteolysis unlikely; and (3) this band was not observed when phosphatase inhibitors were included in the reactions (Fig 5C) [53, 63]. We propose that “pr-Ndt80” represents phosphorylated Ndt80 that is more refractile to phosphatase treatment than the phosphorylated forms exhibiting slower mobility. The critical point is that unphosphorylated Ndt80 appeared when inactive Ndt80 was treated with phosphatase, indicating the presence of phosphates.
Determining whether phosphorylation of inactive Ndt80 is dependent upon MEK1 is difficult because inhibition of Mek1 eliminates the checkpoint by allowing intersister DSB repair [79], resulting in activated Ndt80 and the Ime2-dependent shift [53]. Instead we tested whether phosphorylation of inactive Ndt80 by Ime2 could be ruled out. This goal was accomplished by phosphatase treatment of extracts from a dmc1Δ NDT80-IN diploid containing an analog sensitive version of IME2, IME2ΔC241-as. This allele encodes a truncation of the C-terminus of Ime2 that results in stable, constitutively active kinase [80]. Ime2ΔC241-as activity can be abolished using the 3-MB-PPI inhibitor [48, 81].
Induction of NDT80 in the dmc1Δ IME2ΔC241-as diploid resulted in significantly more meiotic progression than the dmc1Δ mek1-as NDT80-IN diploid, although it was much slower than the isogenic DMC1 diploid, indicating that the checkpoint was active (Fig 5A and 5D). The IME2ΔC241-as allele makes hyperactive Ime2 due to the removal of a C-terminal negative regulatory domain [80]. Constitutively high levels of Ime2 activity may be able to counteract the inhibitory phosphorylation of the induced Ndt80 protein better than the endogenous Ime2 activity in the mek1-as dmc1Δ diploid, resulting in more progression. Addition of the Ime2ΔC241-as inhibitor decreased the amount of meiotic progression, indicating that the inhibitor was working (Fig 5D). In addition, the slowest migrating Ndt80 species disappeared, Hed1 and Hop1 phosphorylation were stabilized and only a very low level of Cdc5 was detected at the 8-hour time point (Fig 5D and 5E). Therefore, the 7-hour time point in the presence of inhibitor represents inactive Ndt80 that lacks Ime2 phosphorylation. Phosphatase treatment resulted in faster migrating bands, demonstrating that checkpoint inactivated Ndt80 contains Ime2-independent phosphates which we propose are mediated by Mek1 (Fig 5F).
Crystal structures of the Ndt80 DBD (amino acids 1–340 or 59–330) bound to an MSE show that the Mek1 consensus sites at S205, T211, S327 and S329 are juxtaposed to the sugar-phosphate backbone of the DNA (Fig 6A) [67, 68]. (No structural information is available for S24). Phosphorylation of these sites therefore places negatively charged phosphates in positions where they could repel the negatively charged DNA, thereby preventing DNA binding. This idea was tested using electrophoretic mobility shift assays (EMSA) with recombinant Ndt80 DBD (aa 1–340) and 29-mer duplexes containing either the MSE from SPS4 or a non-specific sequence designated as Scr (Fig 6D). Note that these DBDs contain the first five Mek1 consensus sites, but not S343. The SPS4 MSE was previously used for in vitro DNA binding assays and structural studies [67].
The Ndt80 WT, 5A and 5D DBDs were fused to a six-histidine tag and purified following the protocol of [82]. All three purified proteins exhibited similar yields and elution profiles (Fig 6B)(S1 Fig). Differential scanning fluorimetry (DSF) was used to determine whether any of the proteins were unfolded. This assay involves incubating proteins with a fluorescent dye and then slowly increasing the temperature to denature the proteins. As the proteins unfold, hydrophobic regions bind the dye, resulting in an increase in fluorescence [83]. The fluorescence values were then normalized to generate melting curves (Fig 6C). Melting temperatures (Tm) were calculated as temperatures with fluorescence values midway between the two extremes. The Ndt80 WT, 5A and 5D DBDs exhibited similar melting curves with Tms of 45.2°C, 45.5°C, and 47°C, respectively. These values indicate that all three proteins were similarly folded, while the 5D protein was even more stable than the WT or 5A protein (Fig 6C).
DNA binding was assayed by incubating Ndt80 WT DBD with a Cy3 fluorescently labeled DNA duplex containing an MSE. All of the duplex was bound, resulting in decreased mobility of the fluorescent DNA (Fig 6D, lane 2). More than 90% of the binding was specific for the MSE sequence, since unlabeled MSE duplex was an effective competitor, decreasing the amount of shifted fluorescent duplex to ~5% (Fig 6D, lanes 3–5). In contrast, equivalent molar amounts of unlabeled Scr duplex did not compete for binding (Fig 6D, lanes 7–9). The Ndt80 5A DBD also bound specifically to the MSE duplex, although less efficiently than WT, while no binding was observed for the 5D protein (Fig 6E, lanes 2–4). Non-specific DNA binding was assessed in a separate reaction using fluorescently labeled Scr duplex. A similar pattern was observed: the WT DBD exhibited the highest level of non-specific binding, followed by the 5A DBD and no binding for the 5D protein (Fig 6E, lanes, 6–8). We conclude that negative charges on the DBD inhibit Ndt80’s ability to interact even non-specifically with DNA.
Having recombinant WT and 5A DBDs in hand allowed us to test whether Mek1 directly phosphorylates the Ndt80 DBD in vitro. Phosphorylation was detected using the semi-synthetic epitope system [84, 85]. Kinase assays contained active GST-Mek1-as isolated from meiotic yeast cells and the ATP analog, 6-Fufuryl-ATPγS. This ATP analog can fit into the enlarged ATP pocket present in the GST-Mek1-as kinase, but not in the ATP binding pockets of other kinases that may have co-purified with GST-Mek1-as. Phosphorylation by GST-Mek1-as transfers a thio-phosphate onto its substrates which are then chemically alkylated to generate an epitope that is recognized by a commercially available thio-ester antibody. GST-Mek1-as auto-phosphorylation was used as an internal control to show that the kinase reaction worked (Fig 6F, lane 1) [35, 85]. The Ndt80 WT DBD was phosphorylated by GST-Mek1-as (Fig 6F, lane 3). Both GST-Mek1-as autophosphorylation and DBD phosphorylation were eliminated by addition of 1-NA-PP1, confirming that GST-Mek1-as kinase activity was responsible for the signal (Fig 6F, lane 2). The 5A DBD was also phosphorylated by GST-Mek1-as, but less efficiently (Fig 6F, lane 9 and 6G). Decreasing the amount of kinase reduced 5A phosphorylation more rapidly than WT DBD phosphorylation. We conclude that: (1) Mek1 phosphorylates at least one of the Ndt80 DBD consensus sites in vitro and (2) Mek1 can also phosphorylate non-consensus sites within the DBD.
It has been known for several years that the meiotic recombination checkpoint in yeast requires MEK1 and that a key target of the checkpoint was Ndt80, but how the two were connected was unclear. The simplest idea, that Mek1 inactivates Ndt80 by directly phosphorylating it, was not considered for two reasons. First, Ndt80 phosphorylation was proposed to promote, not inhibit, Ndt80 activity [48, 63, 64]. Second, deletion of MEK1 has no effect on the mobility shift of Ndt80, leading to the conclusion that Ndt80 is not a substrate of Mek1 [53]. The latter result is misleading, however, because absence of MEK1 results in efficient DSB repair using sister chromatids and therefore removes the signal to the checkpoint. As a result, Ndt80 is activated and phosphorylated by Ime2. Therefore, it is impossible to determine whether Ndt80 is phosphorylated by Mek1 under checkpoint arrested conditions simply by comparing Ndt80 mobility shifts in diploids with or without Mek1 activity.
Using a combination of different approaches, we have demonstrated that Mek1 phosphorylation of Ndt80 is responsible for the meiotic recombination checkpoint delay/arrest. First, Ndt80 is phosphorylated when it is inactivated by the meiotic recombination checkpoint and this phosphorylation is independent of IME2. Second, Ndt80 contains ten Mek1 consensus phosphorylation sites, eight of which are located either within the DNA binding domain or the “middle region” that is required to inhibit Ndt80 in response to the checkpoint. Preventing phosphorylation at these sites using alanine mutations results in partial bypass of the checkpoint triggered by unrepaired DSBs in the dmc1Δ background. Third, Ndt80 contains a conserved five amino acid sequence within the middle region that is required both for checkpoint arrest and for interaction with Mek1. Mutation of this site results in checkpoint bypass without directly affecting Mek1 kinase activity. Fourth, substitution of negatively charged amino acids at Mek1 consensus sites within the Ndt80 DBD constitutively inactivates the transcription factor. Several of these putative Mek1 sites are located immediately adjacent to the negatively charged DNA sugar-phosphate backbone of the MSE. Recombinant Ndt80 DBD containing negative charges at these sites does not bind DNA, even non-specifically. These observations suggest that phosphorylation of the DBD by Mek1 prevents Ndt80 from binding to MSEs and explains how Mek1 phosphorylation can inhibit Ndt80 activity. Finally, Mek1 directly phosphorylates at least one of the Mek1 consensus sites in the Ndt80 DBD in vitro, with less efficient phosphorylation of at least one non-consensus amino acid that has not yet been identified.
While disrupting DNA binding may be the major mechanism by which Ndt80 is inactivated by Mek1, it is unlikely to be the only one. Mek1 phosphorylation of Ndt80 at multiple sites appears to inhibit Ndt80 activity in additional ways because preventing phosphorylation of all of the Mek1 consensus sites in the Ndt80 DBD only weakly bypassed the dmc1Δ checkpoint arrest. This bypass was increased when the DBD alanine substitutions were combined with alanine mutations within the bc domain (ndt80-10AMS). Furthermore, the ndt80-10AMS checkpoint bypass was less efficient than that observed for the deletion of the RPSKR sequence within the bc domain. Since RPSKR is necessary for Mek1 interaction in the two-hybrid system, we propose that deletion of RPSKR disrupts the Mek1-Ndt80 interaction in meiotic cells, preventing any Mek1 phosphorylation of Ndt80 from occurring. In contrast, Mek1 can bind to the Ndt80-10AMS protein via the RPSKR motif and may then phosphorylate non-consensus sites within the DBD and/or the middle region.
Ndt80 that is inactivated by the meiotic recombination checkpoint preferentially localizes to the cytoplasm [69]. It has been proposed that this localization is due to a checkpoint activated cytoplasmic tether but how this tether would work is not clear. An alternative possibility is that Ndt80 constantly shuttles in and out of the nucleus and only when it binds to DNA does Ndt80 remain stably inside the nucleus. Ndt80 is larger than 40 kD, meaning that it is too big to diffuse freely through nuclear pores and must be actively transported [86]. Therefore cytoplasmic localization of inactive Ndt80 could also occur if phosphorylation either promotes nuclear export or inhibits nuclear import. Finally, Mek1 phosphorylation of Ndt80 could inhibit Ime2 phosphorylation at different sites on the transcription factor.
The following model describes how meiotic gene transcription is integrated with meiotic chromosome structure and DSB repair to promote entry into the meiotic divisions only after DSB repair is complete. In vegetative cells, homologs are not associated and transcription of NDT80 and its target genes are repressed by Sum1 bound to MSEs [59] (Fig 7A, 7D and 7G). Early meiotic gene expression is prevented by the Ume6 repressor complex bound to a specific Upstream Repression Sequence called URS1 [90, 91] (Fig 7D). Transfer to Spo medium results in the removal of Ume6 and binding of the Ime1 transcriptional activator at URS1 sites, resulting in expression of early genes such as REC8, HOP1, RED1, MEK1 and SPO11 [92, 93] (Fig 7E). These gene products (and others) function to assemble AEs, make DSBs and activate Mek1 (Fig 7B). Early in meiosis, when DSBs are first occurring, they are repaired primarily using sister chromatids, indicating that Mek1 activity has not reached the threshold necessary to impose interhomolog bias [94]. Similarly, a threshold amount of Mek1 is necessary to inactivate Ndt80. The cell provides time for Mek1 activation by delaying Ime1-driven NDT80 transcription through the additional step of removing the Sum1 repressor. This removal requires that Sum1 be phosphorylated by Ime2, along with CDK and DDK (Fig 7E)[57, 60–62]. Since IME2 is an early gene, it must be transcribed and translated after induction of meiosis. By the time that Ime1-driven transcription of NDT80 occurs, there is sufficient activated Mek1 (Fig 7B, yellow stars) to phosphorylate Ndt80 (Fig 7H, red stars), thereby preventing Ndt80 from binding to DNA (Fig 7H). Mek1 phosphorylation somehow interferes with Ime2 phosphorylation of Ndt80, which also contributes to keeping the transcription factor inactive. Deletion of SUM1 from dmc1Δ diploids results in bypass of the meiotic recombination checkpoint [72]. We propose that when Ndt80 is prematurely expressed in the sum1Δ, there is not enough time to make DSBs and activate Mek1, thereby allowing Ndt80 to activate transcription of its own gene and start the positive feedback loop. A stable interaction between Mek1 and Ndt80 may be necessary to ensure that the kinase is able to counteract removal of phosphates by a phosphatase such as Glc7, which has a role in promoting pachytene exit [95]. As DSBs are processed into double Holliday junctions, their repair promotes chromosome synapsis, resulting in the elimination of most of the Mek1 from chromosomes and a reduction in overall Mek1 kinase activity (Fig 7C) [41, 96]. Without sufficient Mek1 activity, the phosphatase wins out and removes the Mek1-dependent phosphorylation. As a result, Ndt80 becomes activated (which is enhanced by phosphorylation due to Ime2), binds to an MSE in its own promoter to become stably localized to the nucleus and activates a second wave of NDT80 transcription in a positive feedback loop (Fig 7F and 7I) [48, 57, 63, 64]. In addition, Ndt80 target genes are expressed, including CDC5 and CLB1 (Fig 7F). Cdc5 promotes Holliday junction resolution into COs, degradation of Red1 and SC disassembly, thereby eliminating any remaining Mek1 activity [41, 54, 65, 75]. As a result, Rad51 can bind to Rad54 and repair any remaining DSBs prior to entry into MI [41]. Finally, prophase exit resulting from Ndt80-mediated transcription shuts down Spo11 so that no further DSBs are made [97, 98].
While the Ndt80 protein is not conserved outside of fungi, the structure of the DBD is conserved. Ndt80 is a member of the Ig-fold family of transcription factors that includes p53, RUNX, NFAT and NF-kb from mammals [67, 68, 77, 99]. This domain contains a series of loops extending out from several β-sheets that contact DNA to mediate site-specific binding [99]. Interestingly, Ndt80 has more extensive contacts with DNA than other Ig-fold transcription factors, perhaps because Ndt80 binds DNA as a monomer, in contrast to other the proteins which bind as dimers [67, 77].
A common feature of Ig-fold transcription factors is the inhibition of DNA binding by phosphorylation, leading to cytoplasmic localization. For example, NFAT is required for the transcription of cytokine genes involved in T cell activation. In unstimulated cells, phosphorylation of the NFAT nuclear localization signal (NLS) blocks import of the protein into the nucleus [100]. Stimulation of a human T cell lymphoma cell line with phorbol ester activates a phosphatase that removes the phosphorylation at the NLS, allowing translocation into the nucleus where NFAT binds to a specific DNA sequence in its target genes [101]. This binding is inhibited by Cyclosporin A, which results in phosphorylation of the NFAT DBD and cytoplasmic localization of the protein [101]. In another example, phosphorylation of threonine 173 in the DBD of Runx3 by Aurora kinase prevents DNA binding [102]. Similar to the Ndt80 phosphosites, T173 is present at the Runx3-DNA interface. Dissociation from the DNA results in relocalization of Runx3 to the cytoplasm and centrosome during early mitosis. Finally, the p53 protein is a transcription factor that functions in tumor suppression by transcribing genes that promote cell cycle arrest and apoptosis in response to DNA damage [103]. Aurora-A phosphorylates serine 215 in the p53 DBD in vivo. A p53-S215D, but not S215A, mutant prevents DNA binding, resulting in down regulation of target genes necessary for tumor suppression [104]. Although these transcription factors and Ndt80 regulate vastly different processes, there is clearly conservation of the regulatory mechanism that controls them.
Many components of the meiotic recombination checkpoint are conserved between yeast and mammals, even though mammalian meiosis is more complicated than yeast, due to the presence of the X and Y sex chromosomes in males and the dictyate arrest that occurs in oocytes after pachytene exit [10]. DSB-dependent checkpoints have been observed in both mouse oocytes and spermatocytes [105–108]. To study the role of the meiotic recombination checkpoint in mice, a hypomorphic allele of the Trip13 gene called Trip13mod has been used. This mutant has the advantage that chromosomes synapse but many DSBs remain unrepaired, thereby eliminating signals that might arise from a synapsis checkpoint. Trip13mod triggers a DSB-dependent arrest in early pachynema that can be distinguished from later arrest points by the absence of a testis-specific histone variant called H1t. Using this assay, the DSB-dependent arrest has been shown to be dependent on Atm, Chk2, and HORMAD1/2, similar to the requirements for the orthologous yeast genes, TEL1, MEK1 and HOP1, respectively in the meiotic recombination checkpoint [10]. A key target of the mammalian checkpoint is p53 and its paralog, TAp63. Deletion of p53 or TAp63 in Trip13mod/mod mice allows both oocytes and spermatocytes to progress beyond the early pachytene arrest [105, 106]. Using radiation induced DSBs in oocytes, Bolcun-Filas et al (2104)[105] showed that TAp63 is phosphorylated in a Chk2-dependent manner that requires the Chk2 consensus phosphorylation site (LXRXXS) [109]. The mammalian checkpoint response therefore resembles that of yeast: DSBs indirectly activate an FHA-domain containing effector kinase, Chk2 or Mek1, in the context of the AE to regulate an Ig-fold transcription factor, p53/TAp53 or Ndt80, thereby creating an arrest. The main difference is that in mice the checkpoint activates the p53/TAp63 transcription factors while in yeast phosphorylation of Ndt80 prevents transcription.
S1 Table contains a list of plasmids used in this work with the relevant yeast genotypes. S2 Table lists oligonucleotides and their sequences that were used to construct plasmids. Relevant genes in all of the plasmids were sequenced in their entirety by the Stony Brook University DNA Sequencing Facility to ensure that no unexpected mutations were present. Site directed mutagenesis of NDT80 was carried out using the URA3 NDT80 integrating plasmid, pHL8 [61] and the protocol in the Quikchange kit (Stratagene) to generate the 2A, 4AMS, 5AMS, 7AMS, 9AMS, 10AMS 10DMS, 6N, S24D, S343D, S205D T211D, S327D S329D, K374A R375A and K374D R375D mutations. The ndt80-6A, 6D and 8D alleles, in pNH400, pNH401 and pNH405, respectively, were constructed using three fragment Gibson Assembly (GA) reactions (New England BioLabs). One fragment was pRS306 digested with EcoRI and ClaI [110]. The second fragment was 3.3 kb, with overlapping homology with the EcoRI side of the vector and the NDT80 gene between codons 379 and 385. It was amplified using the polymerase chain reaction (PCR) with the primers NDT80-WT-EcoRI-F1 and NDT80-R-385. The third fragment was 1.1 kb and contained NDT80 sequence between codons 379 and 385 and overlapping homology with the ClaI side of the vector. Amplification of this fragment used primers NDT80-WT-ClaI-R1 and NDT80-F-379. For pNH400 and pNH405, pHL8-10AMS was used as the template for the 3.3 kb fragment containing the S24A, S205A, T211A, S327A, S329A and S343A mutations. For pNH401, pHL8-10DMS was the template for the 3.3 kb fragment containing the S24D, S205D, T211D, S327D, S329D and S343D mutations. The 1.1 kb fragment for pNH400 and pNH401 was amplified from the NDT80 gene in pHL8, while pHL8-2A (T399A T420A) was the template for 1.1kb fragment used to make pNH405. The three fragment GA reaction used to make NDT80-Δbc (pHL8-Δbc) used pHL8 to generate two fragments, one using primers NDT80-WT-EcoRI-F1/NDT80-bc-Cla-R1 and the other using NDT80-bc-Cla-F2/NDT80-WT-ClaI-R1, which were then assembled into EcoRI/ClaI digested pRS306.
Estradiol inducible alleles of NDT80 (NDT80-IN) were created using three fragment GA reactions. One fragment was EcoRI/ClaI digested pRS306. The second fragment, containing the GAL1 promoter with homology on one end to sequences flanking the EcoRI site of pRS306, was amplified using pFA6a-HIS3MX6-PGAL1-GFP as the template and PGAL1-EcoRI-F1 and PGAL1-R1 as primers. The third fragment contained the NDT80 open reading frame (ORF) and 3’ flanking sequence. One end had homology to the 3’ end of the GAL1 promoter and the other end to sequences flanking the ClaI site in pRS306. For pBG4, this fragment was amplified using pHL8 as template and the primers, NDT80-ORF-GAL1-F1 and NDT80-WT-ClaI-R1. For pXC11 and pXC12, the template for Fragment 3 was pNH400 and pNH401, respectively.
The E. coli expression plasmids, pNH407-WT, -5A, and -5D were also constructed using GA. The plasmids contain the NDT80 DBD (codons 1–340) followed by a stop codon, fused in frame to six histidines in the pET-28a vector (Novagen). 1.1 kb fragments containing the DBD with overlapping homology flanking the Nde1 and XhoI sites of pET-28a were amplified using either pHL8 (WT), pHL8-10AMS (5A) or pHL8-10DMS (5D) and the primers pET28a-NDT80-F and pET28-NDT80-340-R. These fragments were then incubated with pET-28a digested with NdeI and XhoI and the GA reagent.
The lexA-MEK1 plasmid, pTS3, was constructed using PCR and the primers MEK1-lexA-5/MEK1-lexA-3 to amplify a 1.5 kb fragment containing MEK1 with BamHI sites engineered onto either end. This fragment was ligated into BamHI-digested pSTT91, resulting in an in-frame fusion of the MEK1 ORF with lexA. The R51A mutation in the FHA domain was introduced into pTS3 by site-directed mutagenesis to make pTS3-R51A [47].
The GAD-ndt80284-627 fusion (plasmid A32) was isolated from a two-hybrid screen using lexA-MEK1 as bait. This allele was then re-created de novo in pXC13 using GA so that direct comparison to various deletion alleles could be made. All ndt80 sequences were fused in-frame with GAD and had the same transcriptional terminator and NDT80 3’ untranslated region (UTR). For pXC13, PCR was used to amplify a fragment containing the GAD-ndt80 fusion from A32 using the primers, NDT80-GAD-F/NDT80-GAD-R. The resulting 1.8 kb fragment was then cloned into pACTII digested with NcoI and XhoI. The K374A R375A and K374D R375D mutations were separately introduced into pXC13 by site-directed mutagenesis to make pXC13-KR>AA and pXC13-KR>DD, respectively. The GAD-ndt80-Δbc allele contains an internal, in-frame deletion of the 57 codons of the “bc” domain and was created using a three fragment GA reaction. The first fragment contained GAD fused to NDT80 codons 284–345 and was generated using the primers, NDT80-GAD-F/NDT80-bc-Cla-R1. The second fragment contained NDT80 codons 403–627 along with overlap with the 3’ end of fragment 1. In this case the primers were NDT80-bc-Cla-F2/NDT80-GAD-R. These fragments were then cloned into NcoI/XhoI digested pACTII to generate pXC14. The pXC18 plasmid contains the 57 amino acid “bc” domain directly fused to GAD. Fragment 1 was generated using NDT80-GAD-bc-F3 and NDT80-N1-R1. This fragment was reacted with the NDT80 3’UTR fragment and NcoI/Xho1-digested pACTII to make GAD-bc. The RPSKR sequence was deleted from GAD-ndt80 to make GAD-ndt80-ΔRPSKR in the following way. Fragment 1 was created using NDT80-GAD-F and NDT80-370-R as primers and pXC13 as template to generate a fragment with homology on one end to the 3’ end of GAD and on the other end to NDT80 ending at codon 370. The second fragment was amplified using NDT80-RPSKRΔ-F and NDT80-GAD-R with the pXC13 template. This fragment overlaps on one end with NDT80 codons 360–370, then deletes codons 371–375 and continues to end of NDT80 and homology to the XhoI digested end of pACTII. These two fragments were joined with NcoI/XhoI-digested pACTII by GA to make pNH318. A similar strategy was used to delete the RPSKR codons from the NDT80 ORF to make pNH317. The three fragment GA reaction consisted of (1) EcoRI/ClaI digested pRS306; (2) a fragment amplified from pHL8 using NDT80-WT-EcoRI-F1 and NDT80-370-R and (3) a fragment amplified from pHL8 using NDT80-RPSKRΔ-F and NDT80-WT-Cla-RI. To put GAD-bc under control of the MEK1 promoter, a 1.2 kb fragment containing GAD-bc was amplifed using YEp-GADbc-F and YEp-GADbc-R with pXC18 as the template. This fragment has one end homologous to the MEK1 promoter and the other end homologous to sequences downstream of the NdeI site in pDW14. GA was used to introduce the GAD-bc fragment into Nde1-digested pDW14 to make pLB1.
All strains were derived from the SK1 background unless otherwise noted and their genotypes are listed in S3 Table. Liquid and solid media used for growing cells vegetatively or for sporulation are described in [85]. PCR-based methods were used to delete genes with the drug resistance markers, kanMX6, natMX4 and hphMX4 [111–113]. All deletions were confirmed by PCR. The presence of the deletion allele was confirmed using a forward primer upstream of the ORF and a reverse primer in the drug resistance gene. The absence of the WT allele was also tested using the same forward primer with a reverse primer internal to the gene’s ORF.
To make NH2081, NDT80 was deleted from the haploid parents of NH144, which were then mated. The second exon of DMC1 was then deleted from the ndt80Δ::hphMX4 haploids and mated to make NH2402. The NH144 diploid is heteroallelic for leu2, making it impractical to transform with LEU2 plasmids since transformants cannot be distinguished from mitotic recombinants. The LEU2 gene was therefore deleted in one of the NH144 parents and then mated to the other to make the leu2ΔhisG/leu2Δ::kanMX6 diploid, NH2444.
The NH2426:pEP1052::pX2 (where the “2” indicates a homozygous plasmid) series of diploids was constructed by first deleting NDT80 with hphMX4 from SKY370 and SKY371. The TRP1 GAL4-ER integrating plasmid, pEP105, was digested with Nhe1 and integrated at the trp1::hisG locus in the resulting haploids [41]. URA3-integrating plasmids containing different alleles of NDT80 were then digested with NsiI to target integration to ura3. Integration of the plasmids was confirmed by PCR. The haploids were mated to form homozygous diploids.
The phosphatase experiments were performed NH2437::pEP1052::pBG42, which was derived from the haploid parents of NH2092 [114]. These haploids, NH2091-2-4::pJR2 and NH2091-8-2::pJR2 contain pJR2, a mek1-as URA3 plasmid integrated just upstream of mek1Δ::kanMX6 [114, 115]. Cells that lost the pJR2 plasmid were selected for using 5-fluororotic acid (5-FOA) [116]. To determine whether the mek1-as or mek1Δ::kanMX6 allele remained in the chromosome, FOAR colonies were screened for sensitivity to G418. The first 222 bp of the TRP1 gene were then deleted from the resulting mek1-as haploids using natMX4 [39] and NDT80 was deleted using kanMX6. The GAL4-ER fusion was integrated into the 3’end of TRP1 using NheI-digested pEP105. The PGAL1-NDT80 plasmid, pBG4, was integrated at the ura3 locus using NsiI. The haploids were mated to make NH2437::pEP1052::pBG42. NH2451 was created by deleting the second exon of DMC1 with kanMX6 from the haploid parents of yLJ92 and mating to make the diploid.
Two-hybrid experiments were carried out using the L40 strain that contains lexA operator sequences upstream of the HIS3 and lacZ genes [66]. HIS3 expression was assayed on selective medium (SD-leu–trp–his), while lacZ was assessed using a colorimetric enzyme assay that produces blue color when ß-galactosidase is present [117]. For the two-hybrid screen, L40 containing pTS3 (2μ lexA-MEK1 TRP1) was transformed with a genomic 2μ LEU2 GAD fusion library [118] and 1.1 X 106 transformants were screened for growth on SD -leu, -trp, -his medium. Fifteen His+ transformants also expressed lacZ. The GAD plasmids were isolated from the transformants and the fusion junctions sequenced using GAD-AD-5’. One of these transformants contained GAD fused in-frame to codons specifying amino acids 284–627 of NDT80.
Transformants containing different GAD plasmids and lexA-MEK1 were grown overnight at 30°C on a roller in SD-Leu-Trp. The cells were diluted 1:10 in water and the optical density at 660 nm (OD660) was determined using a spectrophotometer. Culture volumes equivalent to two ODs were pelleted in microfuge tubes and resuspended in 100 μl sterile water. The cells were transferred to a 96-well plate and ten-fold serial dilutions were made. Ten μl cells were spotted onto SD-Leu-Trp and SD-Leu-Trp-His plates. In addition, four μl of each dilution were plated on a paper filter placed onto an SD-Leu-Trp plate. After growth overnight at 30°C, ß-galactosidase assays were performed. The remaining plates were incubated for three days prior to being photographed.
Both the 6xHis-Ndt80-WT-DBD (called WT DBD), 6xHis-Ndt80-DBD-5D (called 5D DBD) and 6xHis-Ndt80-DBD-5A (called 5A DBD) proteins were purified from two different 250 ml cell pellets and used for DNA binding assays. Similar results were obtained with both protein preparations. The protein purification protocol was based on the one described in [82].
The E. coli expression plasmids, pNH407-WT, -5D and 5A were each transformed into BL21(DE3) Codon Plus RIL bacterial cells (Agilent Genomics). Transformants were selected on LB + kanamycin (50 μg/ml) and chloramphenicol (30 μg/ml) (LB +KC) plates. For each plasmid, multiple transformants from a single plate were scraped together and used to inoculate 10 ml LB +KC liquid medium. The cultures were incubated with shaking at 37°C overnight. The next day each culture was diluted to an OD600 of 0.02 in 1 L LB + KC in a 4 L flask and the cultures were grown with shaking at 37°C to an OD600 of 0.4. Imidazole was added to a final concentration of 1 mM to induce transcription of the tagged 6xHis-ndt80 DBD alleles and the cells remained shaking at 37°C for 4 hours. Each liter of culture was divided into 250 ml aliquots and the cells were pelleted by centrifugation, resuspended in wash buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl and 1 mM EDTA), transferred to 50 ml conical tubes and pelleted again. The supernatants were discarded and the cell pellets stored at -20°C.
To lyse the cells, 250 ml cell pellets were thawed on ice and resuspended in 17.5 ml Vershon Lysis Buffer (VLB) (50 mM NaH2PO4/Na2HPO4, pH 7.8, 1 M NaCl) containing 5 mM imidazole and 0.2 mM phenylmethylsulfonyl fluoride (PMSF). The imidazole was made fresh and the PMSF added immediately before sonication. Cell suspensions were transferred to pre-chilled 50 mL glass beakers on ice and the cells lysed by sonication using a Qsonica Q500 ultrasonic processor with a 12.7 mm probe (6 pulses of 15 sec, with 30 sec rests) at 70% power. The lysates were then transferred to pre-chilled polyallomer Beckman centrifuge tubes (25 X 80 mm) and centrifuged in a JA-25.50 rotor at 19,647 X g for 30 min at 4°C. The cleared lysates were then transferred to 15 ml conical tubes, flash frozen in liquid nitrogen and stored at -80°C.
To purify the recombinant proteins, the lysates were thawed on ice, distributed between microfuge tubes and spun at 13,000 X g in a microfuge for 10 min to remove any precipitated material. Lysates were pooled and loaded onto a column containing 0.5 ml bed volume of Ni-NTA Superflow agarose beads (Qiagen) equilibrated in VLB+ 5 mM imidazole. Protein bound beads were washed twice with 2.5 ml VLB+ 5 mM imidazole and then subjected to increasing concentrations of imidazole in the following steps: 10 mM, 50 mM, 100 mM, 200 mM and 250 mM. All of the DBDs eluted with the 50 mM and 100 mM steps (S1 Fig). The second 2.5 ml 50 mM imidazole fraction was mixed with the first 2.5 ml 100 mM imidazole fraction and the proteins were concentrated by centrifugation using Amicon Ultra filters (UFC50124). The molar concentrations of the proteins were determined based on the OD280 absorbance measured by a NanoDrop spectrophotometer (Thermo Scientific) and the calculated molecular weight of the DBD which is 40,169 g/mole. To visualize the proteins, an appropriate volume of 5 X protein sample buffer was added to each sample and the samples were heated at 95°C for 5 min. Proteins were then fractionated on 12.0% SDS-polyacrylamide gels (1.0 mm spacers), using 250 volts for 25 minutes, and then stained with GelCode Blue (Thermo Scientific).
DSF was performed on an Applied Biosystems 7500 Fast Real-Time PCR System using the protocol outlined in the “Protein Thermal Shift Studies” User Guide (Applied Biosystems) with minor modifications. DSF experiments used purified Nt80 DBDs at a final concentration of 5 μM in 96-well PCR plates for fast thermocyclers (VWR, Cat. No. 892180296). Each well (50 μL) contained SyproOrange Dye (Sigma Aldrich) diluted to a final concentration of 5x. The plate temperature was ramped from 25°C to 95°C with a linear gradient (1% ramp rate). The fluorescence of the SyproOrange Dye was detected by selecting ROX as the reporter (filter 4, emission range between 600–625 nm). The fluorescence values were normalized to a range between 0.0 and 1.0 using the equation Ynormalized = (Yraw data−Ymin)/(Ymax−Ymin), where Ymin and Ymax refer to the minimum and maximum values of fluorescence, respectively. Tm values were calculated using the Boltzmann sigmoidal equation in the program GraphPad Prism 4. The Boltzmann sigmoid equation is Yfluorescence = Bottom + (Top—Bottom) / [1 + exp ((V50—Xtemp) / Slope)]. V50 refers to the temperature at which fluorescence is halfway between the bottom and top fluorescence values. Tm values are equal to the calculated V50 value.
Fluorescently labeled 29-mer oligonucleotides (oligos) containing either the SPS4 MSE sequence (5’ Cy3-ATTGACGCGCGCCACAAAAACGTATCATT) or the Scr sequence (5’ Cy5-ATTGACGCGGCTTCATCTCACGTATCATT)(indicated in bold, respectively) were synthesized by Integrated DNA Technologies with high performance liquid chromatography (HPLC)-grade purity. Unlabeled complementary strands were ordered through the Stony Brook Oligonucleotide Facility. Oligos were resuspended at a concentration of 100 μM in water. Complementary strands were annealed by combining equal amounts of each oligo, adding NaCl to a final concentration of 100 mM, incubating the oligos at 95°C for 5 min and then turning off the hot block to allow the strands to slowly anneal overnight. The resulting Cy3- and Cy5-labeled duplex molecules were purified by size exclusion chromatography using a Superdex 200 Increase 10/300 GL column in eluent A (10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 0.01% NP-40 substitute, 50 mM NaCl) at a flow rate of 0.5 mL/min. The molar concentration of DNA in the peak fractions was quantified based on ultraviolet absorbance at 260 nm using a NanoDrop spectrophotometer.
To visualize the DNA, the appropriate volume of 6 X sucrose buffer (7.2 g sucrose in 10.2 mL 1x TE pH 7.8) was added to each sample, and the DNA was resolved on a 6% polyacrylamide gel (1.5 mm thickness) in 0.5 X TBE at 110V for 45 minutes. In-gel Cy3 or Cy5 fluorescence was detected by a Typhoon 9500 scanner (GE Healthcare). Unlabeled duplexes were constructed similarly, except the Cy3 and Cy5 oligos were replaced with unlabeled oligos and the DNA was visualized by in-gel Sybr Gold staining (Thermo Scientific).
Protein dilution buffer (20 mM Tris-HCl, pH 8, 50 mM NaCl, 1 mM EDTA, 1 mg/ml bovine serum albumin (BSA), 5 mM 2-mercaptoethanol) and EMSA reaction buffer (10 mM Tris-HCl pH 7.5, 40 mM NaCl, 4 mM MgCl2, 6% (w/v) glycerol, 10 mg/ml BSA, 10 μg/ml sonicated salmon sperm DNA) were taken from [82]. DNA binding assays were carried out in 20 μl reactions. Reactions were started by addition of the DBD and were incubated at room temperature for 30 min. Reactions and gels were covered with aluminum foil to minimize exposure of the fluorescently labeled to DNA to light. Competition experiments contained 10 nM (1X) Cy3-MSE and 50 nM WT DBD. Unlabeled MSE or Scr duplex 29-mers were added at 2.5X, 10X or 40X the amount of the labeled duplex. Four μl 6 X sucrose buffer were added to each reaction, which were then immediately loaded onto a 6% 0.5 X TBE gel. The DNA was visualized as described above. Specific and. non-specific DNA binding for the different DBDs were compared using 50 nM of each DBD and 50 nM Cy3-MSE or Cy5-Scr.
Protein extracts were generated using the tri-chloroacetic acid method described in [119]. A list of primary and secondary antibodies, sources and dilutions can be found in S4 Table.
Calf intestinal alkaline phosphatase (AP) treatment of TCA extracts was based on a protocol described in [120] with the following modifications. Sixty μL of extract in 100 mM Tris-HCl, pH 6.8, 4% SDS, 200 mM dithiothreitol (DTT) and 20% glycerol were diluted with 408 μL PMP buffer (50 mM HEPES, pH 7.5, 100 mM NaCL, 2 mM DTT and 0.01% Brij 35). One PhosSTOP tablet (Sigma, Cat. #4906845001) containing phosphatase inhibitors was dissolved in 0.5 ml PMP buffer. For each extract, 4 μL AP (80 units)(Sigma, 11097075001) were added to 40 μL PMP buffer (AP alone), as well as 40 μL PMP buffer plus PhosSTOP inhibitors (AP + Inhibitors) and incubated at room temperature for 30 min. This preincubation step was necessary to get more complete inhibition of the AP. Equal amounts of the diluted extracts (156 μL) were aliquoted into separate microfuge tubes: (1) no AP, (2) AP, and (3) AP plus phosphatase inhibitors. To the “no AP” tube 1, 40 μL PMP and 4 μL AP buffer [25 mM Tris-HCl, pH 7.5, 1 mM MgCl2, 0.1 mM ZnCl2, 50% glycerol (v/v)] were added; 44 μL AP in PMP buffer was added to tube 2 and 44 μL AP in PMP buffer plus inhibitors was added to tube 3. The final reactions therefore contained 10-fold less protein than the TCA extracts. The reactions were incubated at 30°C for two hours and then stopped by the addition of 5 X Protein sample buffer. The proteins were fractionated on a 7.5% SDS-polyacrylamide gel, transferred to a filter and probed with α-Ndt80 antibodies.
|
10.1371/journal.pbio.1000373 | Directing Astroglia from the Cerebral Cortex into Subtype Specific Functional Neurons | Astroglia from the postnatal cerebral cortex can be reprogrammed in vitro to generate neurons following forced expression of neurogenic transcription factors, thus opening new avenues towards a potential use of endogenous astroglia for brain repair. However, in previous attempts astroglia-derived neurons failed to establish functional synapses, a severe limitation towards functional neurogenesis. It remained therefore also unknown whether neurons derived from reprogrammed astroglia could be directed towards distinct neuronal subtype identities by selective expression of distinct neurogenic fate determinants. Here we show that strong and persistent expression of neurogenic fate determinants driven by silencing-resistant retroviral vectors instructs astroglia from the postnatal cortex in vitro to mature into fully functional, synapse-forming neurons. Importantly, the neurotransmitter fate choice of astroglia-derived neurons can be controlled by selective expression of distinct neurogenic transcription factors: forced expression of the dorsal telencephalic fate determinant neurogenin-2 (Neurog2) directs cortical astroglia to generate synapse-forming glutamatergic neurons; in contrast, the ventral telencephalic fate determinant Dlx2 induces a GABAergic identity, although the overall efficiency of Dlx2-mediated neuronal reprogramming is much lower compared to Neurog2, suggesting that cortical astroglia possess a higher competence to respond to the dorsal telencephalic fate determinant. Interestingly, however, reprogramming of astroglia towards the generation of GABAergic neurons was greatly facilitated when the astroglial cells were first expanded as neurosphere cells prior to transduction with Dlx2. Importantly, this approach of expansion under neurosphere conditions and subsequent reprogramming with distinct neurogenic transcription factors can also be extended to reactive astroglia isolated from the adult injured cerebral cortex, allowing for the selective generation of glutamatergic or GABAergic neurons. These data provide evidence that cortical astroglia can undergo a conversion across cell lineages by forced expression of a single neurogenic transcription factor, stably generating fully differentiated neurons. Moreover, neuronal reprogramming of astroglia is not restricted to postnatal stages but can also be achieved from terminally differentiated astroglia of the adult cerebral cortex following injury-induced reactivation.
| The brain consists of two major cell types: neurons, which transmit information, and glial cells, which support and protect neurons. Interestingly, evidence suggests that some glial cells, including astroglia, can be directly converted into neurons by specific proteins, a transformation that may aid in the functional repair of damaged brain tissue. However, in order for the repaired brain areas to function properly, it is important that astroglia be directed into appropriate neuronal subclasses. In this study, we show that non-neurogenic astroglia from the cerebral cortex can be reprogrammed in vitro using just a single transcription factor to yield fully functional excitatory or inhibitory neurons. We achieved this result through forced expression of the same transcription factors that instruct the genesis of these distinct neuronal subtypes during embryonic forebrain development. Moreover we demonstrate that reactive astroglia isolated from the adult cortex after local injury can be reprogrammed into synapse-forming excitatory or inhibitory neurons following a similar strategy. Our findings provide evidence that endogenous glial cells may prove a promising strategy for replacing neurons that have degenerated due to trauma or disease.
| While exerting diverse functions within the brain parenchyma [1], astroglia are remarkable in that they also function as neural stem or progenitor cells in specific regions of the postnatal and adult brain [2], such as the ventricular subependymal zone [3] and the subgranular zone of the hippocampus [4],[5]. This raises the possibility that even astroglia from non-neurogenic regions such as the cerebral cortex may be reprogrammed towards neurogenesis when provided with the appropriate transcriptional cues. Indeed, we could previously show that astroglia from the early postnatal cerebral cortex can be reprogrammed in vitro towards the generation of neurons capable of action potential (AP) firing by a single transcription factor, such as Pax6 or its target, the pro-neural transcription factor neurogenin-2 (Neurog2) [6],[7]. These findings may open interesting avenues towards the potential activation of endogenous astroglia for neuronal repair of injured brain tissue.
However, several major obstacles remained to be overcome to fully exploit the potential of reprogrammed astroglia as an endogenous cellular source for neuronal repair. Firstly, reprogramming of astroglia towards neurons remained incomplete as the astroglia-derived neurons failed to establish a functional presynaptic output [7], an obvious hurdle towards functional repair that requires participation in a neural network. Secondly, given the lack of functional presynaptic output, we could not determine the neuronal subtype generated by the reprogrammed astroglial cells [7]. This raises the conceptual concern of whether neurons derived from astroglial cells in a given brain region may be restricted towards the generation of a respective neuronal subtype. During development of the forebrain in rodents, stem/progenitor cells in the dorsal telencephalon generate exclusively excitatory glutamatergic neurons, directed by Pax6 and Neurog1/2 [8]–[10], while stem/progenitor cells in the ventral telencephalon give rise primarily to inhibitory GABAergic neurons, governed by the fate determinants mammalian achaete-schute homolog 1 (Mash1) [11],[12] and Dlx1/2 [13]. Region-specific fate restriction also seems to apply for adult neural stem cells that are intrinsically specified towards the generation of distinct neuronal subtypes [14]. This implies that despite their multipotent nature in regard to generating different glial cell types and neurons, the subtype identity of the neurons generated from these stem cells is predetermined (see also [15]) and is not altered following transplantation [14]. This raises the important question of whether neuronal reprogramming of astroglia derived from the cerebral cortex, a region derived from the dorsal telencephalon, may be restricted towards the generation of glutamatergic neurons, or whether this region-specific bias could be overcome by forced expression of the appropriate neurogenic fate determinants. Such an ability to generate both glutamatergic and GABAergic neurons from astroglia may be a crucial step towards restoring a damaged or imbalanced neuronal network.
Towards this end, we first aimed at a more potent neuronal reprogramming by inducing higher and more persistent expression of neurogenic fate determinants in astroglial cells. This allowed us not only to obtain fully functional neurons that also establish synapses from astroglial cells in vitro but also to demonstrate that distinct neurogenic transcription factors, such as on the one hand Neurog2 and on the other Dlx2 alone or in combination with Mash1, can indeed instruct the selective generation of different neuronal subtypes, such as glutamatergic and GABAergic neurons, respectively. Moreover, we found that the reprogramming efficiency of postnatal cortical astroglia towards GABAergic neurons by Dlx2 could be enhanced by first expanding the astroglial cells under neurosphere conditions prior to forced expression of Dlx2. Given that following brain injury reactive astroglia from the adult cerebral cortex de-differentiate, resume proliferation, and can give rise to self-renewing neurospheres in vitro [16], we finally show that neuronal reprogramming and subtype specification are not restricted to postnatal stages but can also be achieved from adult cortical astroglia responding to injury.
Failure to establish a functional presynaptic compartment by astroglia-derived neurons may be due to an incomplete reprogramming [7]. Here, we hypothesized that stronger and more persistent expression of neurogenic fate determinants may be a prerequisite for a more complete reprogramming of astroglia towards synapse-forming neurons. We have previously shown that expression levels from LTR (long terminal repeat)-driven MMLV (Moloney Murine Leukemia Virus)-derived retroviral constructs, which we employed in previous studies, are only about 2–3-fold of the endogenous expression [6],[17]. Moreover, these viral vectors are prone to silencing [18] and we observed a severe decrease in Neurog2 or green fluorescent protein (GFP) reporter expression already 7–14 d after transduction [7],[19]. Thus, in order to overcome the limitations in synaptogenesis of neurons derived from reprogrammed astroglia, we examined the effect of stronger and more persistent expression of Neurog2 on neuronal reprogramming of astroglia from the cerebral cortex. We therefore subcloned Neurog2 into a self-inactivating retroviral vector driving gene expression under the control of a chicken beta-actin promoter (pCAG) optimized for long-term expression over months in the adult mouse brain [20]. Astroglia cultures were prepared from postnatal day 5–7 (P5–P7) cerebral cortex as described previously [7] and 1 wk later cells were passaged and subsequently transduced with a retroviral vector encoding Neurog2 and DsRed (pCAG-Neurog2-IRES-DsRed) or with a control virus encoding DsRed only (pCAG-IRES-DsRed). Consistent with a stronger and more persistent expression driven by the pCAG promoter, high levels of Neurog2 and DsRed protein were still detected at 5–6 wk following retroviral transduction of cortical astroglia (unpublished data).
In agreement with our previous observation on the high efficiency of neurogenesis from astroglia following forced Neurog2 expression, the vast majority of Neurog2-transduced astroglia had differentiated into βIII tubulin-positive, GFAP-negative neurons after 10 d in culture (Figure S1A and S1A'; 70.2%±6.3% at 9.8±3.1 days post-infection (DPI), 5 independent experiments, n = 1,022 DsRed-positive cells counted), in contrast to control retrovirus transduced cells (1.8%±1.8% of βIII tubulin-positive cells at 7.3±1.0 DPI, 3 independent experiments, n = 3,235 DsRed-positive cells counted). Time-lapse video microscopy revealed that the initial conversion of astroglia into neurons requires approximately 4 d, confirming previous results [7], and can occur at high efficiency (Video S1). By 2–3 wk post-transduction, neurons derived from Neurog2-transduced astroglia had acquired MAP2 immunoreactivity, indicative for dendritic maturation (Figure 1A and 1B). Most strikingly, immunostaining for the vesicular glutamate transporter 1 (vGluT1), present in synaptic vesicles within presynaptic terminals of glutamatergic neurons, revealed that the vast majority of astroglia-derived neurons exhibited a dense labelling with vGluT1-positive puncta outlining their soma and their MAP2-positive processes 4 wk post-infection with Neurog2 (Figure 1A and 1B, 85.4%±5.0% of DsRed-positive neurons at 26.3±2.2 DPI, n = 3 independent experiments, n = 170 DsRed-positive neurons counted). This was in pronounced contrast to the virtual absence of such staining upon transduction with the LTR-driven construct (pCLIG-Neurog2) as described previously [7] and also no vGluT1 immunoreactivity could be detected in astroglial cultures transduced with the control vector (unpublished data). Thus, these data suggest that astroglia reprogrammed with the pCAG-Neurog2-containing retroviral vector acquire a glutamatergic phenotype forming presynaptic specializations.
As vGluT1 immunoreactivity does however not allow to ascertain the neurotransmitter identity of an individual labelled neuron, as the vGluT1-positive puncta may arise from other neurons in the next set of experiments we assessed with single and pair electrophysiological recordings whether astroglial cells reprogrammed by Neurog2 indeed give rise to functional glutamatergic autapses or synapses after a period of 14–32 DPI. As shown in Figure 1D, suprathreshold step-depolarisation of a DsRed-positive neuron (i.e. presynaptic neuron, black asterisk, Figure 1C) resulted in an autaptic response in the stimulated neuron and an inward current in a nearby DsRed-positive neuron with a short delay typical of a monosynaptic connection (i.e. postsynaptic neuron, red asterisk, Figure 1C). In addition, the AMPA/kainate glutamate receptor antagonist CNQX completely abolished both the autaptic and the synaptic current, demonstrating the glutamatergic nature of the presynaptic neuron (Figure 1D). Among all the Neurog2-transduced astroglia-derived neurons recorded (n = 36, average age of cells: 24.6±0.9 DPI), 58.3% exhibited either glutamatergic autaptic connections onto themselves or glutamatergic synapses onto nearby neurons (Figure S2A). In none of the recordings from neurons derived from Neurog2-transduced astroglia was a GABAergic connection observed (Figure S2A). In accordance, cultures transduced with Neurog2 encoding retrovirus were devoid of any vesicular GABA transporter (vGaT) immunoreactivity (unpublished data). Thus, these data provide evidence that Neurog2 does not only induce a generic neuronal fate in postnatal astroglia but selectively elicits differentiation along the glutamatergic lineage, in exclusion of GABAergic neurogenesis. Consistent with the specification of postnatal astroglia towards a glutamatergic identity, forced expression of Neurog2 resulted in the induction of the T-box transcription factors Tbr2 (Figure S1B and S1B') in 20.7%±1.9% of the DsRed-positive cells at 4 DPI (n = 4 coverslips, n = 634 DsRed-positive cells counted) and Tbr1 (48.2% of DsRed-positive neurons at 7 DPI, n = 112 DsRed/βIII tubulin-double positive cells counted; Figure S1C and S1C') as shown previously [7], hence of two well characterised hallmarks of glutamatergic neurogenesis [21]. Moreover, by 4 wk of forced Neurog2 expression, astroglia-derived neurons expressed high levels of the forebrain glutamatergic neuron specific Ca2+/Calmodulin dependent kinase subunit IIα [22], accumulating at dendritic spine-like structures which were typically in opposition of vGluT1-positive presynaptic terminals (Figure 1E–1F').
Consistent with the development of excitatory networks in Neurog2-reprogrammed astroglia cultures, we also observed the emergence of self-driven synaptic activity, resulting eventually in the occurrence of barrages of synaptic currents (Figure 2A). To monitor such self-driven activity, we performed calcium imaging experiments of neurons derived from Neurog2-reprogrammed astroglia. Figure 2B illustrates two neurons that exhibited spontaneous, recurrent, and synchronous Ca2+ transients (Figure 2B–2B”). These Ca2+ transients were completely abolished in the presence of CNQX (Figure 2B”). The majority of the DsRed-positive neurons that we analysed (63.8%, n = 47 imaged neurons, 3 independent experiments) exhibited Ca2+ transients at 14–43 d after transduction with Neurog2, thus indicating the high degree of incorporation of Neurog2-transduced astroglia into excitatory neuronal networks. These data clearly demonstrate that forced expression of Neurog2 driven by the pCAG retroviral vector is sufficient to instruct postnatal cortical astroglia to generate fully functional synapse-forming glutamatergic neurons.
In order to ascertain the astroglial nature of the cells that gave rise to functional glutamatergic synapses following reprogramming by Neurog2, we took advantage of a transgenic mouse line in which GFP expression can be induced in astroglia and is maintained in their progeny. Heterozygous mice in which the expression of a tamoxifen-inducible Cre recombinase is driven by the astroglia specific L-glutamate/L-aspartate transporter promoter (GLAST::CreERT2) [23] were crossed to a reporter mouse line (Z/EG) [24] to generate double heterozygous mutants (GLAST::CreERT2/Z/EG) that were used in the present study. Cre-mediated recombination of the reporter locus was induced via tamoxifen administration from postnatal day 2 (P2) until sacrifice (P5–P7). Astroglia cultures were prepared as described above and, 1 wk later, cells were passaged onto glass coverslips. The vast majority of GFP reporter-positive cells were immunoreactive for GFAP (98.7%±0.7%) at 1 d after plating, with few cells being positive for the oligodendroglial markers NG2/O4 (1.2%±0.7%) and none (0.1%±0.1%) for the neuronal marker βIII tubulin (Figure 3A; n = 3 independent experiments, n = 1,560 GFP-positive cells counted). These data indicate that, under our culture conditions, most reporter-positive cells at the time of transduction possess an astroglial identity. These cells largely remain within their astroglial lineage (86.9%±12.7% of GFP-positive cells expressing GFAP) when analysed at later stages (Figure 3A'; n = 4 independent experiments, n = 1,363 GFP-positive cells counted; 9–21 d following plating). We noted, however, a slight increase in the number of NG2/O4-positive cells (13.0%±12.8%), likely due to the expansion of few reporter-positive clones of oligodendrocyte precursors. Also at later stages reporter-positive cells did not give rise to βIII tubulin-positive neurons (0.1%±0.1%; Figure 3A').
To determine the identity of fate-mapped astroglial cells following retroviral transduction, we performed immunostaining for GFP (identifying cells of astroglial origin), DsRed (identifying transduced cells), and either βIII tubulin, MAP2, or GFAP (identifying neuronal and astroglial cells, respectively). Notably, the stochastic infection of the subset of genetically recombined cells results in a limited number of double-targeted cells. When cultures of adherent astroglia were transduced with the control retrovirus encoding DsRed only, fate-mapped astroglial cells co-expressing GFP and DsRed remained in the glial lineage, as revealed by their astroglial morphology and GFAP expression 1 mo after transduction (Figure 3B–3B”). In sharp contrast, when cultures of tamoxifen-induced astroglia were transduced with the new retrovirus encoding Neurog2 and DsRed, most GFP/DsRed-double-positive fate-mapped astroglia were reprogrammed into neurons expressing the neuronal markers βIII tubulin and MAP2 (67.3%±12.7% among GFP/DsRed-double positive cells at 8.0±1.0 DPI, n = 3 independent experiments, n = 217 double-positive cells counted; Figure 3C–3C”). Single cell tracking of GFP-reporter positive cells following Neurog2-transduction allowed the direct visualisation of the glia-to-neuron conversion of fate-mapped cells over the time course of 5 d (Figure 3D and 3D'; Video S1 and Video S2).
Perforated patch clamp recordings of these fate-mapped astroglia-derived cells reprogrammed by Neurog2 revealed their functional neuronal identity as these cells fired APs following step-current injection in current clamp (n = 8; Figure 4A–4C). In the next set of experiments, we assessed whether neurons derived from fate-mapped astroglia could give rise to functional glutamatergic autapses (Figure 4D–4I). Step-depolarisation of GFP/DsRed-double-positive neurons at 0.05 Hz evoked a sequence of both autaptic and polysynaptic components (2 out of 8 cells recorded) consistent with the excitatory nature of the recorded neurons (average age of the cells: 18.1±2.2 DPI; Figure 4D–4I, insets), while at higher stimulation frequency (1 Hz) the autaptic component with a short decay time typical of glutamatergic synaptic transmission [25] could be observed in isolation (Figure 4F and 4I). Consistent with their glutamatergic nature, fate-mapped astroglia reprogrammed by forced expression of Neurog2 also exhibited a dense labelling of vGluT1-positive puncta (Figure 4J and 4K). These data clearly demonstrate that Neurog2 instructs fate-mapped astroglia from the postnatal cerebral cortex to acquire a glutamatergic identity.
Given that our reprogramming strategy is based on retrovirally mediated expression of neurogenic fate determinants, only cells undergoing cell division will be targeted. In order to examine whether cell division is required for fate conversion to occur, we assessed whether neuronal reprogramming can be also achieved when the Neurog2 and DsRed encoding plasmid is delivered to the postnatal astroglia by transfection, i.e. a gene transfer strategy which does not select for dividing cells, and tracked single transfected cells by time-lapse video microscopy. Transfection with the Neurog2 encoding plasmid resulted in a similar degree of reprogramming after 4 d (14 cells out of 17, Figure 5) as obtained after retroviral transduction. Of note, in four cases neurons were generated directly from single astrocytes without a prior cell division (Figure 5 and Video S3). Thus direct lineage conversion can occur in the absence of cell division.
Based on our finding that forced expression of Neurog2 can selectively drive cortical astroglia towards the generation of functional and synaptically integrated glutamatergic neurons, we next asked whether cortical astroglia may also be directed towards distinct neuronal subtypes. In particular, we asked whether neuronal fate determinants known to instruct the genesis of GABAergic neurons during embryonic development may be sufficient to exert a similar effect on postnatal astroglia. As the homeobox transcription factor Dlx2 is one of the key factors involved in GABAergic neuron specification in the developing ventral telencephalon [13] and in adult neurogenesis [26], we examined whether forced expression of Dlx2 is also sufficient to induce a neuronal and possibly a GABAergic fate in cortical astroglia. To test this, astroglia cultures from P5–P7 cortex of C57BL/6J or GLAST::CreERT2/Z/EG mice were transduced with the same high-expressing retrovirus encoding in this case Dlx2 and DsRed (pCAG-Dlx2-IRES-DsRed), and cells were immunostained for GFP (to identify cells of astroglial origin), DsRed (to identify Dlx2-transduced cells), and the neuronal markers βIII tubulin or MAP2 after various differentiation time periods in culture.
Upon forced expression of Dlx2, a substantial number of postnatal cortical astroglia were redirected towards a neuronal identity as revealed by βIII tubulin or MAP2 expression (35.9%±13.0% at 10.7±2.0 DPI, n = 3 independent experiments, n = 392 DsRed-positive cells counted). Notably, however, the efficacy of neurogenesis elicited by Dlx2 was significantly lower than the one elicited by Neurog2 (see above). Fate-mapping analysis confirmed the astroglial nature of the cells reprogrammed by forced expression of Dlx2 as observed 22 DPI (Figure 6A–6A”). Next, to confirm the neuronal identity of the astroglia-derived cells, we performed patch clamp recordings. All the cells expressing Dlx2 and exhibiting a neuronal morphology, that we recorded, were capable of AP firing in response to step current injection (n = 33). In particular, this also held true for GFP-positive neurons originating from fate-mapped astroglia that had been reprogrammed by Dlx2 (n = 9; Figure 7A–7C”). Notably, neurons derived from Dlx2-transduced astroglia exhibited distinct firing patterns, with most of them revealing immature characteristics (single to few spikes, 22 out of 30 cells recorded) (Figure 7A” and Figure 8). The eight remainder cells exhibited firing patterns which could be classified into three categories [27],[28], namely regular, stuttering, and low-threshold burst spiking (Figure 7B”–7C” and Figure 8), suggestive of the maturation into distinct types of non-fast spiking interneurons [29]. Similarly, the majority of fate-mapped astroglia reprogrammed by Dlx2 (7 out of 9 cells recorded) exhibited immature firing patterns (Figure 7A–7A”), while 2 out of 9 fate-mapped cells developed more mature interneuron-like firing (Figure 7B–7C”). Consistent with the generation of regular- and burst-spiking interneurons [29], we observed calretinin immunoreactivity in a small subset of the Dlx2-expressing cells (Figure 6C), while no parvalbumin immunoreactivity could be detected.
The predominant appearance of immature firing patterns, however, suggests an overall hampered maturation of Dlx2-reprogrammed astroglia. Accordingly, astroglia-derived neurons reprogrammed by Dlx2 exhibited much higher input resistance values than Neurog2-derived neurons after the same time in culture (Figure S2B and S2C; 2,319.2±187.9 MΩ at 26.0±1.4 DPI (n = 26) versus 1,111.8±211.1 MΩ at 23.7±1.5 DPI (n = 20), Dlx2 versus Neurog2, respectively). Surprisingly, the high input resistance of Dlx2-expressing neurons did not decrease but even slightly increased with time in culture (Figure S2B; 2,786.4±440.3 MΩ at 35.9±1.2 DPI (n = 7)), while in the case of Neurog2-reprogrammed astroglia input resistance decreased over time (608.6±125.0 MΩ at 26.8±1.2 DPI; n = 14; Figure S2C). Taken together, these data show that some postnatal cortical astroglia can be redirected by forced expression of Dlx2 towards a neuronal identity; however, in sharp contrast to the progressive maturation of Neurog2-transduced cells, most of the astroglia-derived neurons reprogrammed by Dlx2 remain in a rather immature state, suggesting a comparatively less efficient reprogramming by Dlx2.
Next we assessed whether some of these relatively immature neurons derived from Dlx2-reprogrammed astroglia may nevertheless establish functional autaptic or synaptic connections. We first performed immunocytochemistry for vGluT1 and for vGaT, the latter known to be expressed in synaptic vesicles located in presynaptic terminals of GABAergic neurons. In sharp contrast to reprogramming by Neurog2, astroglia-derived neurons reprogrammed by Dlx2 were devoid of vGluT1 immunoreactivity (unpublished data), but some of them (33.7%±3.6% at 22.0±0.6 DPI, n = 339 DsRed-positive neurons counted; n = 3 independent experiments) were found to exhibit labelling of vGaT-positive puncta outlining both their soma and their processes (Figure 6D). In addition, a small subset of DsRed-positive neurons exhibited GAD67 immunoreactivity (Figure 6B and 6B'). These findings therefore suggest that Dlx2 induces a GABAergic identity in the reprogrammed astroglia. Consistent with an interneuron phenotype, we could also record in 9 out of 33 neurons spontaneous synaptic currents exhibiting a slow decay time, characteristic of GABAergic synaptic events (Figure 7D and 7E). Finally, in few cases, step-depolarisation in voltage clamp evoked an autaptic response of the stimulated neuron (6.1% of the DsRed-positive neurons recorded, n = 33, age of the cells: 26.9±1.4 DPI; Figures 7F and S2A). In accordance with the above data, these autaptic responses exhibited slow decay time kinetics and were abolished by the GABAA receptor antagonist bicuculline (Figure 7F), thus demonstrating the GABAergic nature of these autapses.
Taken together, these data strongly indicate that forced expression of Dlx2, in sharp contrast to Neurog2, can induce the reprogramming of astroglia from the postnatal cortex towards a GABAergic phenotype. However, whereas Neurog2 redirected the majority of astroglia towards functional glutamatergic neurons, only few astroglial cells reprogrammed by Dlx2 differentiated into fully functional, GABAergic neurons (58% versus 6%, respectively), thus indicating that Dlx2-induced reprogramming remains partial in most of the cells.
Since we have previously shown that Mash1, a transcription factor located up-stream of Dlx2 in the interneuron fate specification [30], the direct targets of which overlap only partially with that of Mash1 [31], can also reprogram postnatal astroglia towards neurogenesis [7], we tested whether co-expression of these two transcription factors further promote neurogenesis and subsequent interneuron differentiation of reprogrammed astroglia [32]. Consistent with previous data [7], 33.5%±17.8% of astroglia expressing Mash1 alone developed into βIII tubulin-positive neurons (Figure 9B, n = 3 independent experiments, n = 226 DsRed-positive cells counted at 10.7±2.0 DPI). In contrast, co-expression of Mash1 and Dlx2 significantly augmented neurogenesis from postnatal astroglia (93.0%±3.1% of βIII tubulin-positive neurons amongst DsRed-positive cells, n = 3 independent experiments, n = 548 DsRed-positive cells counted at 10.7±2.0 DPI; Figure 9B), indicating that these two factors indeed act synergistically. Moreover, compared to cells expressing Dlx2 alone, Mash1/Dlx2 co-expressing neurons exhibited lower input resistance values (1,237.5±278.8 MΩ at 18.4±1.0 DPI; n = 15; Figure 9C). Consistent with a more mature status, a higher proportion of Mash1/Dlx2 co-expressing neurons exhibited specific interneuronal firing patterns (6 out of 15 cells recorded, Figure 8 and Figure 9A–9A”) compared to Dlx2 (8 out of 30 cells recorded, Figure 8). Despite this enhanced degree of differentiation, none of the recorded cells co-expressing Mash1 and Dlx2 showed an autaptic response (Figure S2A, n = 15).
Taken together, our data provide evidence that postnatal astroglia from the cerebral cortex can be driven towards the generation of interneurons with distinct functional properties by forced expression of Dlx2 or Dlx2 in combination with Mash1.
Next, we examined whether a complete and more efficient reprogramming of astroglia towards synapse-forming functional GABAergic neurons may be achieved by first expanding astroglial cells as neurospheres in presence of mitogens that on the one hand promote de-differentiation of astroglia and on the other hand up-regulate fate determinants normally involved in the generation of GABAergic neurons in the telencephalon, such as Mash1 [33]–[35]. We therefore cultured postnatal astroglial cells as neurospheres before transducing the astroglia-derived neurosphere cells with Dlx2. It has been previously shown that early postnatal cortical astroglial cells can give rise to neurospheres until P11 [36]. Cortical tissue from P5–P7 C57BL/6J, GLAST::CreERT2/Z/EG, or hGFAP-GFP mice was cultured as neurospheres under non-adherent conditions in serum-free medium and in the presence of EGF/FGF2, and 1 wk later, astroglia-derived neurosphere cells were passaged and subsequently transduced with retrovirus encoding either Dlx2-DsRed or Neurog2-DsRed, and were allowed for differentiation. The astroglial origin of the neurosphere founder cells was confirmed by culturing single GFP-positive cells derived from the postnatal cortex of hGFAP-GFP mice which gave rise to neurospheres (Figure S3A–S3D). Moreover, quantitative RT-PCR demonstrated that during expansion in EGF/FGF2 neurosphere cells expressed mRNAs for different astroglial markers, such as the specific pan-astrocyte marker Aldh1L1 [37], at similarly high levels as adherent astroglia after 1 wk in culture, and the mRNA encoding GFAP at even higher levels (Figure 10A–10E). In contrast, no βIII tubulin mRNA could be detected (Figure 10F). Likewise, similar to adherent astroglia neurosphere cells did not express detectable levels of endogenous Neurog2 mRNA (Figure 10G). These data support the notion that during the expansion in EGF/FGF2, neurosphere cells have an astroglial character.
In contrast to adherent astroglia cultures, 94.7%±0.3% (n = 3 independent experiments, n = 644 DsRed-positive cells counted) of the astroglia-derived neurosphere cells transduced with Dlx2 differentiated into MAP2-positive neurons (Figure S4A and S4C). Strikingly, even at younger stages in culture (20.1±1.9 DPI), Dlx2-expressing neurons derived from neurosphere cells exhibited substantially lower input resistances (1,266.4±294.3 MΩ, n = 7) compared to Dlx2-reprogrammed adherent astroglia recorded at 4 wk in culture (2,319.2±187.9 MΩ at 26.0±1.4 DPI, n = 26; Figure S2B), indicative of a more advanced neuronal maturation. Immunostaining for vGaT revealed a dense labelling of vGaT-positive puncta (Figure 11A and 11B), thus suggesting that astroglia-derived neurosphere cells transduced with Dlx2 had acquired a GABAergic identity. Finally, electrophysiological recordings demonstrated the GABAergic phenotype of the fate-mapped astroglia expanded as neurospheres and reprogrammed by Dlx2 (n = 9; Figure 11C–11E). Consistent with the widespread vGaT expression, Dlx2-expressing neurons received spontaneous synaptic activity displaying slow decay time kinetics characteristic of GABAergic currents (9 out of 10 cells recorded, unpublished data). As shown in Figure 11E, step-depolarisation of a GFP-positive, Dlx2-expressing neuron (Figure 11C and 11D, black asterisk) evoked an autaptic response of the stimulated neuron that was blocked by bicuculline. Four out of 10 Dlx2-transduced neurosphere-derived neurons recorded were found to form GABAergic autapses, despite being analysed at a younger age compared to the adherent astroglia (n = 10, age of the cells: 20.1±1.6 DPI; Figure S2A). Consistent with the development of GABAergic networks in Dlx2-transduced cultures, calcium imaging experiments performed at 27 DPI did not reveal any spontaneous Ca2+ transients in the analysed DsRed-positive neurons (n = 70 neurons recorded, n = 2 independent experiments; Figure S5). Thus, culturing astroglia under neurosphere conditions clearly eases the reprogramming towards functional GABAergic neurons by Dlx2 transduction compared to the effects obtained in adherent astroglia (40% versus 6%, neurosphere cells versus adherent astroglia, respectively).
Interestingly, astroglia cultured as neurosphere cells could still be reprogrammed by Neurog2 towards a glutamatergic neuronal phenotype. Virtually all astroglia-derived neurosphere cells transduced with Neurog2 differentiated into MAP2-positive neurons (91.4%±2.2%, n = 2,272 DsRed-positive cells counted, n = 3 independent experiments, in agreement with [7]) that exhibited a large soma size and extended several MAP2-positive processes (Figure S4B and S4D). Again, fate-mapping analysis corroborated the astroglial origin of the reprogrammed cells (Figure 12D–12F). These neurons derived from astroglia-derived neurosphere cells exhibited quite low input resistances similarly to the adherent astroglial cells reprogrammed by Neurog2, therefore suggesting that they had reached a mature neuronal state (Figure S2C; 751.9±118.8 MΩ at 15.0±1.4 DPI (n = 10) versus 608.6±125.0 MΩ at 26.8±1.2 DPI (n = 14), Neurog2-reprogrammed neurosphere cells versus Neurog2-reprogrammed adherent astroglia, respectively). Immunostaining for vGluT1 revealed a massive labelling of vGluT1-positive puncta indicating that the Neurog2-transduced cells had acquired a glutamatergic identity (Figure 12A–12C). Indeed, electrophysiological pair recordings unambiguously demonstrated the glutamatergic phenotype of the fate-mapped cortical astroglia expanded as neurospheres and reprogrammed by Neurog2 (n = 5) (Figure 12E–12G). In addition, step-depolarization of Neurog2-expressing neurons also evoked in several cases a sequence of polysynaptic components consistent with the development of excitatory networks in these cultures (Figure S3G and S3H). Nine out of 21 DsRed-positive neurons recorded exhibited glutamatergic autaptic or synaptic connections (age of the cells: 14.2±0.7 DPI; Figure S2A). In addition, calcium imaging experiments performed 3–4 wk post-infection in these Neurog2-reprogrammed cultures revealed a high degree of self-driven, synchronous excitatory network activity, which was blocked by CNQX and AP5 treatment (97 out of 98 DsRed-positive cells imaged; n = 3 independent experiments; Figure 13). As can be appreciated from Figure 13 many of the neurons recruited to the self-driven networks were derived from fate-mapped GFP-positive neurosphere cells indicating their astroglial origin (28 out of 29 fate-mapped GFP/DsRed-double-positive cells imaged).
These data suggest that astroglia initially expanded as neurosphere cells are more plastic in regard to their differentiation into various neuronal subtypes. What may be the molecular changes underlying the increased plasticity of neurosphere cells compared to adherent astroglia given their common astroglial origin? The striking increase in the efficiency of reprogramming by Dlx2 could be accounted for by a loss in the expression of molecular cues associated with a glutamatergic bias. However, quantitative RT-PCR showed that there was no difference in the expression of Emx1 or Emx2 mRNAs [38] between astroglial cells cultured adherently or under neurosphere conditions (Figure 10H and 10I). In contrast, astroglial cells cultured under neurosphere conditions expressed drastically higher levels of Sox2 mRNA compared to adherent astroglia (Figure 10J). Given the high level of expression of Sox2 in neural stem cells [39], these data are in agreement with the observation that culturing postnatal astroglia from the cerebral cortex under neurosphere conditions leads to their de-differentiation towards a more stem cell-like state [33],[34].
The above finding that Neurog2 or Dlx2 overexpression can induce with high efficiency the generation of functional neurons from postnatal astroglia-derived neurosphere cells prompted us to examine whether reactive astroglia from the adult cortex following injury can also be reprogrammed to generate functional neurons after prior expansion as neurospheres. Indeed, previous work of our laboratory has shown that following a local injury such as a stab wound lesion, reactive astroglia isolated from the adult cerebral cortex de-differentiate, resume proliferation, and have the capacity to give rise to self-renewing neurospheres in vitro, in contrast to the intact contralateral cortex [16]. To examine the reprogramming potential of reactive astroglia isolated from the adult injured cerebral cortex, we performed local stab wound lesions in the right cortical hemisphere of adult C57BL/6J mice and dissociated both the control and injured cortical hemispheres 3 d later for subsequent neurosphere cultures. While control cortical tissue did not generate neurospheres, the injured hemisphere gave rise to neurospheres as reported [16]. Confirming previous results [16], neurospheres could also be obtained from single GFP reporter-positive cells derived from either GLAST::CreERT2/Z/EG mice (unpublished data) or hGFAP-GFP mice (Figure 14A–14A”). After 1–2 wk, single neurospheres were plated, subsequently transduced with retrovirus encoding Neurog2 and DsRed or Dlx2 and DsRed, and then allowed for differentiation.
When astroglia-derived neurosphere cells obtained from lesioned cortex of wild type mice were transduced with Neurog2, virtually all the DsRed-positive cells had developed into MAP2-positive neurons at 15 DPI (>50 DsRed-positive cells per sphere; 25 spheres analysed, Figure 14C and 14D). Importantly, when GFP-positive neurospheres originating from the injured cortex of hGFAP-GFP mice (Figure 14A–14B) were transduced with Neurog2, we observed the generation of numerous DsRed-positive neurons (total number of neurons >50, 10 spheres analysed). In contrast to untransduced lesion cortex neurosphere cells, GFP expression was lost following neuronal differentiation (Figure 14B), consistent with the astroglia-to-neuron fate change.
Following step-current injection, Neurog2-expressing neurosphere-derived cells responded with a train of repetitive APs demonstrating their neuronal nature (n = 30; Figure 15A and 15A'). In addition, these neurons derived from adult glia exhibited relatively low input resistance values similar to Neurog2-transduced neurons derived from the postnatal cortical neurosphere cells (1,009.2±177.4 MΩ at 21.9±1.2 DPI (n = 19) versus 751.9±118.8 MΩ at 15.0±1.4 DPI (n = 10), respectively). We next assessed whether these neurons could establish functional synaptic connections. Single adult cortex-derived neurospheres showed a massive immunostaining of vGluT1-positive puncta as shown 28 DPI, that outlined the dense network of intermingled MAP2-positive processes (Figure 15B) and the soma of Neurog2-transduced cells (Figure 15C). Consistent with vGluT1 expression, immunostaining for the presynaptic protein synapsin also revealed a dense labelling of synapsin-positive puncta, thus suggesting the development of synaptic contacts between adult lesioned cortex-derived neurosphere cells (Figure 15D). Furthermore electrophysiological recordings revealed the emergence of CNQX-sensitive spontaneous synaptic currents in these neurons in accordance with vGluT1 and synapsin expression (8 out of 30 cells recorded at 22.5±0.9 DPI; Figure 15E–15F”). These data indicate that adult astroglia-derived neurosphere cells transduced with Neurog2 mature into functional glutamatergic neurons.
To examine the extent of plasticity of glial cells derived from the adult lesioned cortex we also tested as a proof-of-principle experiment whether adult lesioned cortex-derived neurosphere cells could also be directed by forced expression of Dlx2 towards MAP2-expressing neurons (Figure S6A and S6A'). However, Dlx2 reprogrammed neurons were rather few and fragile due to their rather small soma size, thus hampering extensive electrophysiological analysis. Nevertheless we could record from one cell shown in Figure S6, where step-depolarisation evoked an autaptic response that was blocked by bicuculline, indicating the development of functional GABAergic connections (Figure S6B and S6B'). These data show that even adult cells isolated from the injured cortex and expanded as neurospheres can be instructed by forced expression of Neurog2 or Dlx2 to generate mature neurons able to establish functional glutamatergic or GABAergic connections, respectively.
The present study provides four major findings: firstly, it provides new independent experimental evidence based on genetic fate-mapping that astroglia from the postnatal cerebral cortex can be reprogrammed by a single transcription factor into functional neurons; secondly, we have succeeded in overcoming the previous limitations in synaptogenesis of neurons derived from postnatal cortical astroglia by the use of retroviral vectors conveying higher and more persistent levels of neurogenic fate determinants' expression [7]; thirdly, based on the transcription factor used for reprogramming, postnatal astroglia can be directed towards the generation of glutamatergic and GABAergic neurons, providing proof-of-principle evidence that selective subtypes can be generated from the same cells of origin; fourthly, reprogramming efficiency is further enhanced by prior de-differentiation of the astroglia as provided by the expansion under neurosphere conditions. Importantly, using the latter procedure we even succeeded in reprogramming reactive astroglia after adult brain injury. Thus, reprogramming of astroglia towards functional neurons with a single transcription factor is not restricted to postnatal stages but can also be achieved from astroglia of the adult cerebral cortex following injury-induced reactivation.
We have previously shown that postnatal astroglia can be reprogrammed into neurons [6],[7]. However, as the astroglial origin of the reprogrammed cells is a very important issue, here we sought to provide new experimental evidence via genetic fate-mapping of astroglia by using the GLAST::CreERT2/Z/EG mouse line developed in our laboratory [23]. To ensure the specificity of this mouse model we showed that virtually all fate-mapped, i.e. GFP-reporter positive, cells remain in the glial lineage under our culture conditions, with the vast majority being identified as astrocytes based on GFAP expression as well as GLAST (unpublished data) and a minor population as oligodendroglial cells, while none of the fate-mapped cells spontaneously gave rise to neurons (Figure 3A and 3A'). These data support the notion that the cells cultured under these conditions do not possess an intrinsic neurogenic potential. This is consistent with the finding of virtually absent endogenous Neurog2 expression compared to cortical precursors isolated at the embryonic stage (Figure 10G) and in agreement with the epigenetic silencing of the neurogenin-1 and -2 loci at the transition between neurogenic and astrogliogenic precursors [40]. These data do not rule out the possibility that in vivo a small subset of astroglial cells can still give rise to neurons as suggested by hGFAP::CreERT2-mediated fate mapping showing that neurons can be generated at early postnatal stages from genetically marked cells [41]. However, so far it could not be experimentally distinguished whether the postnatal generated neurons indeed had been derived from astroglia local to the cerebral cortex or would be derived from astroglial stem cells in the subependymal zone that had subsequently immigrated into the cortex [42]. In any case our cultures do not sustain conditions for the genesis of neurons from astroglia (even in the presence of EGF/FGF2) without forced expression of neurogenic fate determinants.
Yet despite generating only glia when adherently grown in serum containing medium, postnatal astroglia exhibit a remarkable degree of plasticity as indicated by the fact that at least some can give rise to self-renewing, multipotent neurospheres ([36] and present study). The latter fact could be taken as evidence that early postnatal astroglia possess stem cell character, particularly in the light of the notion that cells from the postnatal cerebral cortex may contribute to the pool of radial/astroglial stem cells within the dorsal adult subependymal zone [14] (for review see [43]). However, several lines of evidence argue against a stem cell character of the astroglial cells studied here. Firstly, neither wild-type nor genetically fate-mapped astroglia spontaneously give rise to neurons, which is inconsistent with the stem cell defining hallmark of multipotency (Figure 3A–3B”). Moreover, the large number of neurons generated following forced expression of Neurog2 argues against the possibility that the successfully reprogrammed cells derive from rare stem cells within this culture (Video S1). Secondly, while Dlx2 very efficiently directs adult neural stem cells in vitro towards neurogenesis [26], the responsiveness of adherent astroglia is much more limited, suggesting a reduced susceptibility to Dlx2 transcriptional activity. The third line of evidence is based on the striking difference in Sox2 mRNA levels following expansion of the postnatal astroglia as neurosphere cells. Sox2 is a transcription factor well known to play a key role in neural stem cell self-renewal [39]; thus the massive up-regulation of Sox2 following exposure of astroglia to neurosphere culture conditions suggests that these cells undergo de-differentiation eventually acquiring indeed stem cell properties, while the comparatively lower levels of Sox2 in the adherent cultures would be in agreement with their non-stem cell character. Consistent with the higher degree of plasticity characterizing neural stem cells, the efficiency of reprogramming of astroglia-derived neurosphere cells by Dlx2 was found to reach levels comparable to bona fide neural stem cells (Figure S4; [26]). Intriguingly, exposure of non-stem cell astroglia to EGF/FGF2 in the absence of serum factors may thus mimic similar extrinsic signals encountered by astroglial cells during postnatal development that later give rise to the stem cell compartment in the adult subependymal zone [44]. In fact the conversion of quiescent astrocytes from the adult cerebral cortex into stem cell-like cells following injury ([16] and present study) clearly demonstrates that cells apparently devoid of stem cell properties can acquire stem cell hallmarks such as self-renewal and multipotency when exposed to the appropriate environment.
While our previous findings of eliciting neurogenesis by a single transcription factor from postnatal astroglial cells demonstrated the potency of these neurogenic fate determinants [6],[7], a major obstacle towards reprogramming into fully functional neurons was encountered in the failure of the astroglia-derived neurons to provide functional presynaptic output to other neurons [7]. In our previous study neuronal reprogramming was achieved by transcription factor-encoding retroviral vectors that exhibit relatively low levels of overexpression and are subject to substantial silencing in neurons [18]. Of note, long-term expression of an exogenous transcription factor appears to be required for maintaining a new phenotype following cellular reprogramming [45]. For instance, it has been shown that reprogramming of fibroblasts into macrophage-like cells remains unstable, resulting in the loss of macrophage markers, following silencing of retrovirally expressed PU.1 and C/EBPα [46]. Here we demonstrated that stronger and more prolonged expression of neurogenic fate determinants from retroviral constructs more resistant to silencing [20] indeed permits more complete reprogramming of postnatal cortical astroglia towards synapse-forming neurons. In support of our hypothesis, we found that Dlx2 expression driven from a weaker and silencing-prone retroviral vector (pMXIG) [26] resulted in nearly negligible neurogenesis in postnatal astroglia cultures (unpublished data), while the same fate determinant encoded by the pCAG vector induced substantial neurogenesis. Consistent with a more efficient reprogramming via a strong and silencing-resistant retroviral expression system, we found that forced expression of Neurog2 or Dlx2 endowed astroglia-derived neurons not only with the ability to receive synaptic input but also to form functional presynaptic output onto other astroglia-derived neurons to such degree as generating networks of spontaneously active neurons. Thus, one of the major obstacles that may impair incorporation of astroglia-derived neurons into a neuronal network, namely the inability to give rise to functional presynaptic output, can be overcome by appropriate expression of neurogenic fate determinants. Interestingly, neuronal reprogramming of astroglia by Neurog2 towards mature neurons appears to involve similar developmental steps as in newborn neurons during embryonic and adult neurogenesis. For instance, while GABA acts as an inhibitory neurotransmitter onto mature astroglia-derived neurons reprogrammed by Neurog2, at a more immature stage these neurons respond with a rise in free intracellular calcium (Blum et al., submitted), very similar to immature embryonic- or adult-generated neurons [47],[48] and such a differential response may be required for proper maturation to proceed [49].
Of note, despite its continued expression, Neurog2, a transcription factor normally expressed in progenitors and only transiently maintained in postmitotic neurons [50], does not seem to interfere with a surprisingly normal maturation of the reprogrammed cells. Firstly, input resistances of Neurog2-expressing neurons reach levels that correspond well to values observed for neurons recorded in slices from postnatal mice of roughly matching age [51]. Secondly, as a sign of proper morphological maturation dendrites of Neurog2-expressing neurons were covered with spines after 1 mo in culture, in agreement with progressive formation of excitatory synapses. Finally, astroglia-derived neurons reprogrammed by Neurog2 exhibit prominent expression of the Ca2+/Calmodulin-dependent kinase IIα subunit, which in vivo is exclusively expressed in excitatory neurons starting from the first postnatal week [22]. Moreover, expression of Ca2+/Calmodulin-dependent kinase IIα is a pre-requisite for the occurrence of changes in synaptic efficacy [52], suggesting that Neurog2-reprogrammed cells potentially also acquire the ability to undergo synaptic modification.
The fact that cells committed to the astroglial lineage can be reprogrammed into fully functional synapse-forming neurons also sheds some light on the more general question of whether the conversion across cell lineages induced by a single transcription factor can generate fully differentiated and stable cell fates that closely mirror cell types found in vivo [45]. Our study provides definitive positive evidence that such a cell lineage conversion is indeed possible and does not require passing through a pluripotent ground state. Interestingly, a recent study has shown that even mouse embryonic or perinatal fibroblasts, i.e. cells of the mesodermal lineage, can be converted by combined forced expression of three defined factors (Mash1, Brn2, Myt1l) into functional neurons [53]. Unexpectedly this combination seems to favour the generation of glutamatergic neurons, while the possibility to generate neurons of other phenotypes remains to be explored. Our study indeed reveals for the first time the feasibility of direct conversion of a somatic cell type into distinct neuronal subtypes by selective expression of transcription factors. Moreover, the efficiency of fibroblast conversion into neurons remains markedly lower (∼20%) despite the joint use of three transcription factors. Conversely, we demonstrate here that cells of closer lineage-relationship to ectoderm-derived neurons, namely the astroglia, require only a single transcription factor (Neurog2) to be converted towards fully functional neurons with 60% efficiency. This does not only confirm the notion that lineage reprogramming is achieved with best results by using the closest related cells but is also of profound relevance in regard to the eventual translational potential towards regenerative medicine. Activation of endogenous brain cells towards neuronal repair may be feasible if only one factor needs to be activated, e.g. by transvascular delivery of small molecules and RNAs [54]. In addition, the regional specification of astroglial cells characterized by distinct transcription factor profiles [55],[56] may create a specific bias towards the generation of the type of neurons normally residing within the respective brain region and hence favour the fate conversion into the appropriate neuronal subtypes. Our results therefore further pave the way towards neuronal reprogramming from endogenous cells residing within the brain, circumventing the complications associated with transplantation.
Intriguingly, although the majority of astroglial cells undergoing reprogramming by Neurog2 also undergo cell cycle division, single cell tracking demonstrated that astroglia can give rise to neurons without dividing. These data show that cell division is not a sine qua non condition for successful reprogramming, providing additional evidence for a direct fate conversion by-passing a proliferative state. Future studies will have to reveal whether even adult quiescent astrocytes could be reprogrammed into neurons without the requirement of entering the cell division cycle.
Notably, given that expression of neurogenic fate determinants allows for the generation of synapse-forming neurons, we could assess whether distinct transcription factors promote neuronal subtype specification of the reprogrammed cells. Indeed, our immunocytochemical and electrophysiological analysis demonstrated that astroglia derived from the cerebral cortex can be reprogrammed not only towards the generation of glutamatergic neurons by Neurog2, a fate determinant that regulates glutamatergic neuron generation in the developing dorsal telencephalon [8], but also towards the generation of synapse-forming GABAergic neurons by Dlx2. Of note, glutamatergic neurogenesis from cortical astroglia following reprogramming by Neurog2 was accompanied by the (transient) up-regulation of Tbr2 and Tbr1, T-box transcription factors that characterize the genesis of glutamatergic neurons throughout the forebrain [21]. Moreover, Neurog2 induced the expression of the Ca2+/Calmodulin-dependent kinase IIα, which is selectively expressed in glutamatergic neurons of the forebrain [22]. The fact that Neurog2-reprogrammed astroglia derived from the cerebral cortex generate glutamatergic neurons rather than other neuronal subtypes which also require Neurog2 expression for their specification, such as spinal cord cholinergic motoneurons, is consistent with the notion that astroglial cells retain region-specific identity characterized by distinct transcription factor profiles [55],[56], in this case dorsal telencephalic cues [57].
In contrast to Neurog2, Dlx2 is normally expressed in progenitor cells derived from the ventral telencephalon and has been shown to play a crucial role in the genesis of GABAergic neurons during development [13] and in adulthood [26] in this region. Yet the fact that astroglial cells of dorsal telencephalic origin can be forced to adopt a “ventral” fate is consistent with previous findings that ventral transcription factors can instruct a GABAergic fate in dorsally derived progenitors including induction of endogenous Dlx gene expression [12],[58]. Of note, however, this cellular competence to respond appropriately to “wrong” regional transcriptional cues is not restricted to relatively unspecified precursors but can be even observed in cells committed to the astroglial lineage.
However, the ability of Dlx2 to induce neuronal reprogramming from cortical astroglia is limited. In contrast to reprogramming induced by Neurog2 that occurred with very high efficiency, only a third of all Dlx2-transduced astroglia differentiated into neurons and most of these exhibited high input resistances typical of immature neurons even after prolonged periods of culturing. Moreover, the apparent impairment of maturation resulted in an even lower number of Dlx2-expressing neurons forming functional GABAergic synapses. Also only a minority of the neurons obtained from Dlx2-reprogrammed astroglia displayed more mature interneuronal firing patterns, with the majority of neurons often responding with one or few spikes to prolonged current injection. Intriguingly, however, among those neurons acquiring mature firing properties we could clearly discern distinct types of patterns which have been classified as regular spiking, stuttering, and low-threshold burst spiking. These data indicate that Dlx2-reprogrammed astroglia can eventually mature into specific interneuron subtypes. Consistent with this we found that a small subset of the reprogrammed cells expressed the calcium binding protein calretinin, which is normally expressed within the cortex in subpopulations of regular or low-threshold burst spiking interneurons [29].
The apparent limitations of Dlx2-mediated astroglia-to-neuron fate conversion could be due to inaccessibility of some downstream targets of Dlx2 in cortex-derived astroglia or to an overall lower potency of Dlx2 compared to Neurog2 in reprogramming and/or the need for additional co-factors missing in cortical astroglia. Along these lines we examined whether co-expression of Mash1 with Dlx2 may further promote the maturation and specification of astroglia into synapse-forming interneurons. Mash1 is a ventral telencephalic transcription factor upstream of Dlx2 in the cascade of interneuron specification during development [11],[12]. Importantly, while being upstream of Dlx2, which is a direct transcriptional target of Mash1 [30], the latter factor is also known to activate targets that are not shared with Dlx2, suggesting that complete interneuron specification may require the activity of both factors [31],[32]. In agreement with this, we found that co-expression of Mash1 and Dlx2 promoted neurogenesis from astroglia in a synergistic manner to levels similar to Neurog2 (Figure 9). Moreover, input resistances were lower compared to those of Dlx2 only expressing neurons. Finally, the relative proportion of neurons exhibiting more mature interneuronal firing patterns was increased (Figure 8). However, there was no apparent enhancement of synapse formation following co-expression. This may point to the possibility that full interneuronal maturation requires extrinsic signals provided by their target cells, i.e. excitatory neurons, which are absent in cultures from Dlx2-reprogrammed astroglia. Alternatively, additional transcription factors such as Dlx5/Dlx6 may be required for complete maturation to occur.
Given that different subtypes of GABAergic neurons are generated in the telencephalon ranging from medium spiny projection neurons of the striatum to various types of aspiny interneurons throughout the telencephalon [29], it will be of great interest to develop strategies to further refine the subtype specification of GABAergic neurons generated upon forced Dlx2 or Mash1/Dlx2 expression, by using additional transcriptional cues in order to generate the full spectrum of GABAergic interneurons. Such directing of astroglia towards specific interneuron phenotypes may allow for the development of alternative approaches for the treatment of pharmaco-resistant epileptic disorders, particularly during early childhood [59],[60].
Interestingly, culturing astroglial cells prior to transduction under neurosphere conditions improved the efficiency of reprogramming towards GABAergic neurons quite dramatically. Indeed, the synapse-forming capacity of Dlx2-expressing astroglia-derived neurosphere cells exceeded that of adherent astroglia-derived neurons expressing Dlx2 by a factor of seven. In addition, neurons derived from Dlx2-expressing neurosphere cells exhibited significantly lower input resistances consistent with a more mature state.
These findings are not only of great importance in regard to eliciting the generation of neurons of different subtypes but may also support the concept that culturing cells under neurosphere conditions (in serum-free medium containing high levels of EGF and FGF2) induces the erasure of region-specific transcription [33],[34]. Indeed culturing neural cells of different origins under neurosphere conditions has been shown to induce a partial loss of region-specific transcription factor expression while resulting in the up-regulation of Olig2 and Mash1 [34], providing a more permissive transcriptional environment that favours reprogramming towards distinct neuronal subtypes. However, we found no striking differences in the expression of the mRNAs encoding the cortical patterning factors Emx1 and Emx2 between cortical astroglia cultured adherently or expanded under neurosphere condition (Figure 10H and 10I). In contrast, the two populations differed drastically in their expression levels of Sox2 mRNA (Figure 10J), which may indicate a more stem cell-like status of the astroglia under neurosphere conditions and which may be the molecular correlate for their higher degree of plasticity.
The successful reprogramming of astroglia following expansion under neurosphere conditions also encouraged us to investigate whether neurosphere cells derived from the adult cerebral cortex after injury can be similarly directed towards neurogenesis. Thus, we assessed here whether adult cortex-derived neurosphere cells also can be driven towards fully functional neurons following forced expression of Neurog2 or Dlx2. Indeed, virtually all the cells expressing Neurog2 acquired a neuronal identity as revealed by MAP2 staining and their ability to fire APs. Moreover, we also provide evidence by vGluT1 immunoreactivity that the neurons derived from these lesion cortex-derived neurosphere cells undergo a subtype specification similar to reprogrammed postnatal astroglia and differentiate into glutamatergic neurons. Consistent with the development of functional synapses by Neurog2-reprogrammed adult cortex-derived neurosphere cells, we could record spontaneous glutamatergic events in these cultures. Conversely, following forced expression of Dlx2, we could observe the generation of functional GABAergic neurons from adult cortex-derived neurosphere cells, indicating that the same dichotomy of subtype specification observed in postnatal astroglia also holds true for adult cortex-derived neurosphere cells following injury. While this first demonstration that functional neurons of different subtypes even undergoing synaptic connectivity can be derived in vitro from adult glial cells isolated from the injured cortex is an exciting step forward to utilize endogenous glial cells for repair of neurons [61]–[63], it will still be a major challenge to translate these in vitro findings into the context of the injured brain.
Experiments were conducted either on C57BL/6J mice or transgenic GLAST::CreERT2/Z/EG double heterozygous mice. Briefly, heterozygous GLAST::CreERT2 mice [23] were crossed with the Z/EG reporter mouse line [24] to generate double heterozygous mutants. The activation of the tamoxifen-inducible form of the Cre recombinase was done as follows: From postnatal day 2, i.e. at the peak of cortical astrogliogenesis [64], until sacrifice at postnatal day 7, tamoxifen (20 mg/mL, dissolved in corn oil, Sigma-Aldrich, Munich, Germany) was administered as described by Mori et al. [23] to mothers and mice pups thus received tamoxifen via the milk from their mother during lactation. Using the same mouse line as in this study, Mori et al. found that by recombination already at E18 virtually all fate-mapped cells give rise to glia [23], indicating that at that stage GLAST expressing cells have lost their intrinsic neurogenic potential.
In some experiments, hGFAP-GFP transgenic mice were also used [65]. All animal procedures were carried out in accordance with the policies of the use of Animals and Humans in Neuroscience Research, revised and approved by the Society of Neuroscience and the state of Bavaria under licence number 55.2-1-54-2531-144/07. All efforts were made to minimize animal suffering and to reduce the number of animals used.
Retroviral transduction of astroglia cultured as adherent astroglia or expanded as neurosphere cells was performed 2–3 h after plating on coverslips, using VSV-G (vesicular stomatitis virus glycoprotein)-pseudotyped retroviruses encoding neurogenic fate determinants. Neurog2 or Dlx2 were expressed under control of an internal chicken β-actin promoter with cytomegalovirus enhancer (pCAG) together with DsRed located behind an internal ribosomal entry site (IRES) [20]. The Neurog2 coding cDNA was subcloned from the pCLIG-Neurog2 construct [67] into the EcoRI site of the pSKSP shuttle vector, from where it was then subcloned between the 5′SfiI and 3′PmeI restriction sites of the pCAG retroviral vector to generate pCAG-Neurog2-IRES-DsRed. The Dlx2 coding cDNA was subcloned from the pMXIG-Dlx2 construct [26] and inserted into the pCAG retroviral vector following the same cloning strategy to generate pCAG-Dlx2-IRES-DsRed. For control, cultures were transduced with a virus encoding only DsRed behind an IRES under control of the chicken β-actin promoter (pCAG-IRES-DsRed). Viral particles were produced using gpg helperfree packaging cells to generate VSV-G (vesicular stomatitis virus glycoprotein)-pseudotyped viral particles [68]. Viral particles were titered by clonal analysis after transduction of E14 cortical cultures.
Twenty-four hours after transduction, the medium of astroglia cultured as adherent astroglia was completely replaced by a differentiation medium consisting of DMEM/F12, 3.5 mM glucose, penicillin/streptomycin, and B27 supplement, and the cells were allowed for differentiation for different time periods. Similarly, the medium of astroglia-derived neurosphere cells was replaced by a differentiation medium consisting of DMEM/F12, 3.5 mM glucose, penicillin/streptomycin, supplemented with B27, and buffered with HEPES. As it has been found that Brain-Derived Neurotrophic Factor (BDNF) is required for robust synapse formation of neurons derived from neural stem cells [69], 20 ng/mL of BDNF (Calbiochem) were added to the cultures every fourth day during the differentiation period. Cells were cultured at a CO2 concentration of 9%, resulting in a pH of the differentiation medium of ∼7.2. After 7–41 d following transduction, cells were used either for immunocytochemistry, electrophysiology, or calcium imaging experiments. The number of days after retroviral transduction is indicated as DPI.
Transfection via DNA-liposome complexes was performed 2 h after plating of passaged cortical astroglia on poly-D-lysine coated 24-well tissue plates. DNA-liposome complexes were prepared in Optimem medium (Invitrogen) using the retroviral plasmid pCAG-Neurog2-IRES-DsRed or the control plasmid pCAG-IRES-DsRed and Lipofectamine 2000 (Invitrogen) as cationic liposome formulation. Astrocyte cultures were exposed to DNA-liposome complexes at a concentration of 0.5 µg DNA per 400 µL of Optimem medium for 4 h. Subsequently the medium was replaced by differentiation medium consisting of DMEM/F12, 3.5 mM glucose, penicillin/streptomycin, and B27 supplement, and the 24-well tissue plates were placed into the time-lapse incubating chamber.
Time-lapse video microscopy [70],[71] of P7 cortical astrocyte cultures was performed with a cell observer (Zeiss) at a constant temperature of 37°C and 8% CO2. Phase contrast images were acquired every 4 min and fluorescence images every 6–12 h for 6–8 d using a 20× phase contrast objective (Zeiss) and an AxioCamHRm camera with a self-written VBA module remote controlling Zeiss AxioVision 4.7 software [72]. Single-cell tracking was performed using a self-written computer program (TTT) [72]. Videos were assembled using Image J 1.42q (National Institute of Health, USA) software and are played at speed of 4 frames per second.
Adult C57BL/6J mice of 8–10 wk of age (20–25 g) were injured in the neocortex as described previously [61]. Briefly, mice received Rimadyl (4 mg/kg, s.c., Carprofen) as analgesic treatment and were anesthetized with ketamine (100 mg/kg, i.p., Ketavet, GE Healthcare, Germany) and xylazine (5 mg/kg, i.p., Rompun, Bayer, Germany) and placed in a stereotaxic frame in a flat skull position. After trepanation, a stab wound was made in the right cerebral sensorimotor cortex by using a sharp and thin scalpel (Ophthalmic Corneal V-lance knife, Alcon, Germany) at the following coordinates: anteroposterior (AP) = from −1.6 to −2.4, mediolateral (ML) = −1.5, dorsoventral (DV) = −0.6 mm with Bregma as reference. After surgery, mice were housed in individual Plexiglas cages with food and water ad libitum and kept in a 12 h light-dark cycle (room temperature = 22±1°C).
Three days after stab wound lesion, mice were killed by cervical dislocation following euthanasia in rising CO2 concentrations. After removal of the meninges grey matter tissue surrounding the stab wound injury of the neocortex and a similar piece at the same rostrocaudal level in the contralateral hemisphere were dissected and neurospheres were generated as described above.
After 7–14 d, neurospheres generated from the adult injured cerebral cortex were collected and plated as single neurosphere on poly-D-lysine (Sigma-Aldrich) coated coverslips in 24-well plates (BD Biosciences) in a medium consisting of DMEM/F12 supplemented with B27, EGF, FGF2, penicillin/streptomycin, and buffered with HEPES. Two to 3 h after plating on coverslips, neurosphere cells were transduced as described above with retroviral vectors encoding Neurog2 (pCAG-Neurog2-IRES-DsRed), Dlx2 (pCAG-Dlx2-IRES-DsRed), or the control retrovirus (pCAG-IRES-DsRed) and were then allowed for differentiation.
For immunocytochemistry, cultures were fixed in 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS) for 15 min at room temperature. Cells were first pretreated in 0.5% Triton X-100 in PBS for 30 min, followed by incubation in 2% BSA and 0.5% Triton X-100 in PBS for 30 min. Primary antibodies were incubated on specimen overnight at 4°C in 2% BSA, 0.5% Triton X-100 in PBS. The following primary antibodies were used: anti-GFP (GFP, chicken, 1∶1000, Aves Labs, GFP-1020), polyclonal anti-Glial Fibrillary Acidic Protein (GFAP, rabbit, 1∶4000, DakoCytomation, Z0334), polyclonal anti-Red Fluorescent Protein (RFP, rabbit, 1∶500, Chemicon, AB3216), polyclonal anti-RFP (rabbit, Rockland, 1∶2000, 600-401-379), polyclonal anti-vesicular glutamate transporter 1 (vGluT11, rabbit, 1∶1000, Synaptic Systems, 135302), monoclonal anti-Microtubule Associated Protein 2 (MAP2, mouse IgG1, 1∶200, Sigma-Aldrich, M4403), monoclonal anti-synapsin 1 (mouse IgG2, 1∶2000, Synaptic Systems, 106001), polyclonal anti-vGaT (guinea pig, 1∶200, Synaptic Systems, 131004), polyclonal anti-Tbr1 (rabbit, 1∶1000, Millipore, AB9616), polyclonal anti-Tbr2 (rabbit, 1∶500, Millipore, AB9618), monoclonal anti-βIII tubulin (mouse IgG2b, 1∶500, Sigma, T8660), polyclonal anti-GAD1(GAD67) (rabbit, 1∶500, Synaptic Systems, 198013), monoclonal anti-calretinin (mouse IgG1, 1∶200, Millipore, MAB1568), and monoclonal anti-CaM kinase IIα (mouse IgG1, 1∶200, Abcam, ab2725). After extensive washing in PBS, cells were incubated with appropriate species- or subclass-specific secondary antibodies conjugated to Cy™2, Cy™3, Cy™5 (1∶500, Jackson ImmunoResearch), Alexa Fluor 488 (1∶500, Invitrogen), FITC (fluorescein isothiocyanate, 1∶500, Jackson ImmunoResearch), TRITC (tetramethyl rhodamine isothiocyanate, 1∶500, Jackson ImmunoResearch), or biotin (1∶500, Jackson ImmunoResearch or Vector Laboratories) for 2 h in the dark at room temperature, followed by extensive washing in PBS. Following treatment with secondary antibodies conjugated to biotin, cells were subsequently incubated for 2 h at room temperature with AMCA streptavidin (1∶200, Vector Laboratories) or Alexa Fluor 647 streptavidin (1∶500, Invitrogen). Coverslips were finally mounted onto a glass slide with an anti-fading mounting medium (Aqua Poly/Mount; Polysciences, Warrington, PA).
Stainings were first examined with an epifluorescence microscope (BX61, Olympus, Hamburg, Germany) equipped with the appropriate filter sets. Stainings were further analyzed with laser-scanning confocal microscopes (SP5, Leica, Wetzlar, Germany or LSM710, Carl Zeiss, Göttingen, Germany). Z-stacks of digital images were captured using the LAS AF (Leica) or ZEN software (Carl Zeiss). Single confocal images were then extracted from the Z-stacks. Alternatively, the Z-stacks were collapsed in one resulting picture using the maximum intensity projection function provided by the above mentioned softwares.
Cell counts were performed by taking pictures of several randomly selected views per coverslip analysed by means of a Zeiss LSM 710 confocal microscope using a 25× objective. Subsequently, pictures were analysed for cell quantification using Image J 1.42q (National Institute of Health, USA) software. For each quantification, values are given as mean ± SEM. Cell counting data from reprogramming induced by Dlx2, Mash2, and Mash1 in combination with Dlx2 were subjected to a two-tailed Student's t test for statistical significance. Differences were considered statistically significant when the probability value was <0.05.
Electrophysiological properties of neurons derived from reprogrammed astroglial cells were analyzed 11–41 d following retroviral transduction. Single or dual perforated patch-clamp recordings [73],[74] were performed at room temperature with amphotericin-B (Calbiochem) for perforation. Micropipettes were made from borosilicate glass capillaries (Garner, Claremont, CA, USA). Pipettes were tip-filled with internal solution and back-filled with internal solution containing 200 µg/mL amphotericin-B. The electrodes had resistances of 2–2.5 MΩ. The internal solution contained 136.5 mM K-gluconate, 17.5 mM KCl, 9 mM NaCl, 1 mM MgCl2, 10 mM HEPES, and 0.2 mM EGTA (pH 7.4) at an osmolarity of 300 mOsm. The external solution contained 150 mM NaCl, 3 mM KCl, 3 mM CaCl2, 2 mM MgCl2, 10 mM HEPES, and 5 mM glucose (pH 7.4) at an osmolarity of 310 mOsm. The recording chamber was continuously perfused with external solution at a rate of 0.5 mL/min.
Cells were visualized with an epifluorescence microscope (Axioskop2, Carl Zeiss) equipped with the appropriate filter sets. For patch clamp recordings, virally transduced cells were selected on the basis of their DsRed immunoreactivity. In addition, to ascertain the astroglial origin of the recorded neurons, DsRed- and GFP-expressing cells from GLAST::CreERT2/Z/EG animals were also selected for patch clamp recordings. Digital pictures of the recorded cells were acquired using a digital camera (AxioCam, Carl Zeiss).
Signals were sampled at 10 kHz with Axopatch 200B patch-clamp amplifiers (Axon Instruments, Foster City, CA, USA), filtered at 5 kHz and analyzed with Clampfit 9.2 software (Axon Instruments). For assessing a cell's ability to fire APs, cells received depolarizing step-current injections. AP amplitudes were measured by subtracting the threshold voltage of the AP from the AP maximum amplitude. For determining input resistance, hyperpolarizing currents of small amplitudes were injected into the cells under current clamp condition at a holding potential of −70 mV and input resistances were calculated from the corresponding voltage deviation.
To examine spontaneous synaptic input into a given neuron, cells were kept in voltage clamp at a holding potential of −70 mV and synaptic events were recorded throughout a period of 1 to 5 min. In order to assess autaptic connections, single cells were step-depolarized in voltage clamp for 1 ms from −70 to +30 mV at a frequency of 0.05 Hz and responses were recorded in the same cell. Responses were considered to be autaptic when they occurred within 3 ms after the step-depolarization [75]. Synaptic connectivity was investigated by means of pair recordings in voltage clamp mode. One neuron was stimulated at low frequency (0.05–0.1 Hz) by a 1 ms step-depolarization from −70 to +30 mV and the response was recorded from the other neuron, and vice versa. To determine the nature of the autaptic or synaptic responses, neurons were step-depolarized as described above and we assessed whether responses could be abolished in the presence of either the GABAA receptor antagonist bicuculline (10 µM) or the AMPA/kainate receptor antagonist CNQX (5 µM). Finally, the recovery of the autaptic or synaptic response was assessed following washout of the pharmacological drugs.
Neurons derived from reprogrammed astroglial cells were further analyzed 22, 27, and 29 d after retroviral transduction with calcium imaging experiments. A 5 mM stock solution of Oregon-Green BAPTA1, AM (KD: 170 nM, Invitrogen, O6807) was prepared in 8.9 µL 20% Pluronic F-127 (Invitrogen) in dimethylsulfoxide (DMSO) by means of a sonifier bath (Bandelin, Berlin, Germany) for 3 min. Reprogrammed astroglial cells were incubated with 5 µM Oregon-BAPTA1, AM in artificial cerebrospinal fluid solution (ACSF: in mM: 127 NaCl, 4.5 KCl, 2.5 NaH2PO4, 2 CaCl2, 2 MgCl2, 23 NaHCO3, and 25 D-glucose., bubbled with 95% O2/5% CO2.). The incubation of neurons with the dye-ACSF was performed in a cell culture incubator (37°C, 9% CO2) for a loading time of 10–15 min. Cells were washed with ACSF in a perfusion chamber (Volume: 150–200 µL) for at least 10–20 min with a flow rate of approximately 2–3 mL/min. Cells were imaged under continuous perfusion with ACSF solution at 26–30°C. Ligands (Ascent scientific) were bath-applied and used at the following concentrations: CNQX (10 µM), D-AP5 (10 µM), and tetrodotoxin (TTX, 500 nM). For confocal Ca2+ imaging (256×256 pixels, 2.16 Hz), an inverted confocal microscope (Olympus IX70, equipped with a Fluoview 300 laser scanning system) was used in combination with an Olympus, UPlanApo 20×/0.7 objective. Transduced cells were identified by means of DsRed expression. Oregon Green-derived fluorescence was excited with a 488 nm laser line (emission filter: band pass 510/540 nm). DsRed was excited at 543 nm (emission filter: band pass: 580 nm±40 nm).
In some experiments, astroglia from GLAST::CreERT2/Z/EG mice, reprogrammed by forced expression of neurogenic fate determinants, were used for calcium imaging experiments. To ascertain the astroglial origin of the DsRed-positive reprogrammed cells that will be analyzed, GFP expression was first pictured at the confocal microscope and was carefully bleached by using the 488-laser afterwards. The DsRed-positive reprogrammed astroglial cells were subsequently incubated with Oregon Green and calcium imaging experiments were processed as described above.
Images were analyzed using IMAGEJ software (WS Rasband, IMAGEJ, US National Institutes of Health, Bethesda, Maryland, USA, http://rsb.info.nih.gov/ij/, 1997–2006). XY-time Calcium imaging results were analyzed by a region-of-interest analysis (pixel intensity) in the extended TIFF format as described before [76].
Total RNA was extracted with RNeasy Plus MicroKit (Qiagen), according to the manufacturer's instructions. 1–1.5 µg of total RNA was retro-transcribed using Super-ScriptIII Reverse Transcriptase (Invitrogen) and random primers. Each cDNA was diluted one to ten, and 2 µl was used for each real-time reaction. mRNA quantitation was performed on a DNA Engine Opticon 2 System (Bio-Rad) following the manufacturer's protocol using the IQ SYBR Green SuperMix (Bio-Rad). The following oligonucleotide primers were used for the qPCR: Gapdh, Emx1, and Emx2 [34]; Sox2 [77], Ngn2 [78], βIIItub [79], Glt1 [80], Glu1 (Glutamine Synthetase) (CCTGGACCCCAAGGCCCGTA; TGGCAGCCTGCACCATTCCAG), Aldh1l1 (TGTTTGGCCAGGAGGTTTAC; AGGTCACCAGTGTCCAGACC), S100β (GATGTCTTCCACCAGTACTCC; CTCATGTTCAAAGAACTCAT), and Gfap [81]. The amount of each gene was analyzed in triplicate, and the analysis was repeated on at least three independent samples (n = 3 for NPC and postnatal neurospheres, n = 5 for postnatal adherent astrocytes). Data analysis was performed with the ΔΔCt method [82].
|
10.1371/journal.pbio.1001143 | Pancreatic Mesenchyme Regulates Epithelial Organogenesis throughout Development | The developing pancreatic epithelium gives rise to all endocrine and exocrine cells of the mature organ. During organogenesis, the epithelial cells receive essential signals from the overlying mesenchyme. Previous studies, focusing on ex vivo tissue explants or complete knockout mice, have identified an important role for the mesenchyme in regulating the expansion of progenitor cells in the early pancreas epithelium. However, due to the lack of genetic tools directing expression specifically to the mesenchyme, the potential roles of this supporting tissue in vivo, especially in guiding later stages of pancreas organogenesis, have not been elucidated. We employed transgenic tools and fetal surgical techniques to ablate mesenchyme via Cre-mediated mesenchymal expression of Diphtheria Toxin (DT) at the onset of pancreas formation, and at later developmental stages via in utero injection of DT into transgenic mice expressing the Diphtheria Toxin receptor (DTR) in this tissue. Our results demonstrate that mesenchymal cells regulate pancreatic growth and branching at both early and late developmental stages by supporting proliferation of precursors and differentiated cells, respectively. Interestingly, while cell differentiation was not affected, the expansion of both the endocrine and exocrine compartments was equally impaired. To further elucidate signals required for mesenchymal cell function, we eliminated β-catenin signaling and determined that it is a critical pathway in regulating mesenchyme survival and growth. Our study presents the first in vivo evidence that the embryonic mesenchyme provides critical signals to the epithelium throughout pancreas organogenesis. The findings are novel and relevant as they indicate a critical role for the mesenchyme during late expansion of endocrine and exocrine compartments. In addition, our results provide a molecular mechanism for mesenchymal expansion and survival by identifying β-catenin signaling as an essential mediator of this process. These results have implications for developing strategies to expand pancreas progenitors and β-cells for clinical transplantation.
| Embryonic development is a highly complex process that requires tight orchestration of cellular proliferation, differentiation, and migration as cells grow within loosely aggregated mesenchyme and more organized epithelial sheets to form organs and tissues. In addition to intrinsic cell-autonomous signals, these events are further regulated by environmental cues provided by neighboring cells. Prior work demonstrated a critical role for the surrounding mesenchyme in guiding epithelial growth during the early stages of pancreas development. However, it remained unclear whether the mesenchyme also guided the later stages of pancreas organogenesis when the functional exocrine and endocrine cells are formed. Here, we show that specific genetic ablation of the mesenchyme at distinct developmental stages in vivo results in the formation of a smaller, misshapen pancreas. Loss of the mesenchyme profoundly impairs the expansion of both endocrine and exocrine pancreatic progenitors, as well as the proliferative capacity of maturing cells, including insulin-producing beta-cells. Thus, our studies reveal unappreciated roles for the mesenchyme in guiding the formation of the epithelial pancreas throughout development. The results suggest that identifying the specific mesenchymal signals might help to optimize cell culture protocols that aim to achieve the differentiation of stem cells into insulin-producing beta cells.
| Organogenesis is a complex and dynamic process that requires tight spatial and temporal regulation of differentiation, proliferation, and morphogenesis. The pancreas serves as an interesting model for the study of these processes as its epithelium gives rise to functionally distinct cells: endocrine cells, including insulin-producing β-cells that release hormones into the blood stream to regulate glucose homeostasis, and exocrine cells that produce, secrete, and transport digestive enzymes. These diverse cell types derive from common progenitors residing in the embryonic pancreatic epithelium through a well-orchestrated multi-step process. While numerous studies have delineated the cascades of transcription factors within the epithelium that guide epithelial cell development (reviewed in [1],[2]), the role of the surrounding mesenchyme in governing pancreas organogenesis at different stages remains largely unknown.
Mesenchymal cells start to coalesce around the nascent gut tube shortly before pancreas epithelial cells evaginate around mouse embryonic day 9.5 (e9.5) to form the dorsal and ventral buds [1]. At e13.5–e14.5 Pdx1+ epithelial precursor cells become committed to either the endocrine or the exocrine lineage, and from e15.5 until the end of gestation, pancreatic cells undergo final differentiation to give rise to all pancreatic cell types found in the adult organ. The first evidence that mesenchymal cells were required for pancreatic epithelial growth was provided in the 1960s by seminal work by Golosow and Grobstein [3], in which it was shown that e11 mouse pancreatic epithelium rudiments stripped of their overlying mesenchyme failed to grow in culture. However, further studies addressing the role of the mesenchyme at later stages have been difficult as the expanding pancreas epithelium quickly branches into the surrounding mesenchyme, thus preventing clean physical separation of these two layers after ∼e12 in the mouse. Additionally, while improved culture conditions for organ rudiments mimic embryonic development during early stages quite well [4], full replication of all in vivo aspects of later pancreas organogenesis have not been achieved ex vivo [5]. As a consequence, studying the role of the mesenchyme at advanced stages of pancreas development using explant systems resulted in controversial findings. A number of studies have shown that while mesenchymal cells have a positive effect on exocrine differentiation and growth in culture, they impair endocrine cell development [6]–[10]. Other studies have observed that close proximity between mesenchyme and epithelium promotes exocrine differentiation, while secreted mesenchymal factors enhance endocrine differentiation over a distance [5]. More recently, a study by Attali and colleagues showed that co-culture of epithelium with mesenchyme promotes the production of insulin-expressing cells, an effect largely due to the expansion of Pdx1+ precursor cells rather than maturation or proliferation of insulin-positive cells [11]. Importantly, endocrine development was highly variable and dependent on the culture conditions such as oxygen levels [11], further indicating that in vivo manipulation of mesenchymal gene expression is necessary to fully uncover all mesenchymal functions throughout pancreas development.
Starting in the 1970s, extensive efforts were made to identify mesenchymal factors responsible for these effects on the epithelial compartment [12],[13]. A decade ago, Bhushan and colleagues demonstrated that fibroblast growth factor 10 (Fgf10), expressed by mesenchymal cells from e9.5 until e11.5, is essential for pancreas growth and differentiation as it stimulates proliferation of Pdx1-expressing precursor cells [14]. Since then, germ-line knock out mouse lines, genetically manipulated zebra fish, and transfected chick embryos have been used to study a limited number of additional mesenchymal signaling pathways for their role in guiding pancreas formation (summarized in [1]). These studies provided evidence for Retinoic Acid (RA), Wnt, FGF, BMP, TGFβ, and EGF signaling pathways as important regulators of pancreas formation [1],[10],[14]–[17]. However, detailed studies of the requirement for individual mesenchymal factors in pancreas development have been hampered by the lack of transgenic tools that permit manipulation of gene expression specifically in the pancreatic mesenchyme.
Here, we present experiments that take advantage of Nkx3.2 (Bapx1)-Cre transgenic mice in which Cre-expression is directed to the embryonic pancreatic mesenchyme, but not the epithelium. Using this Cre line in conjunction with mouse lines allowing Diphtheria Toxin (DT) induced apoptosis, we depleted mesenchymal cell during various stages of in vivo pancreas development. As expected, elimination of mesenchymal cells at the onset of pancreas development completely blocked pancreas organogenesis. Surprisingly, mesenchymal requirement was not restricted to this early stage, as ablation at later developmental stages also led to severe epithelial hypoplasia, reduced branching, and impaired β-cell and exocrine cell expansion. To elucidate the signaling pathways essential for mesenchyme function, we eliminated canonical Wnt signaling from the tissue. Loss of Wnt signaling within the mesenchyme resulted in mesenchymal cell ablation—subsequently leading to reduction in both exocrine and endocrine cell mass. Summarily, our results demonstrate that the pancreatic epithelium depends on mesenchymal signals for proper expansion and morphogenesis throughout development.
In order to manipulate gene expression in pancreatic mesenchyme, but not epithelium, we looked for genes whose expression matches this pattern. Previous studies pointed to the homeobox gene Nkx3.2 (also known as Bapx1), whose expression was found in the forming somites as well as in the mesenchyme of developing pancreas, stomach, and gut [18]–[23]. In contrast, Nkx3.2 expression was not detected in endodermally derived cells in these tissues [18],[20],[23]. In the pancreatic mesenchyme Nkx3.2 is expressed as early as e9.5, and by e12.5 its expression becomes restricted to the mesenchymal area, which will give rise to the splenic bud [18]–[20],[23]. An Nkx3.2 (Bapx1)-Cre line, in which one copy of the endogenous Nkx3.2 gene was replaced by a transgene encoding the Cre recombinase, had previously been generated [24],[25]. This transgenic mouse line faithfully replicates the endogenous expression of Nkx3.2 and directs Cre activity to the foregut mesenchyme and skeletal somites starting at e9.5 [24].
Given that pancreatic expression of the Nkx3.2-Cre transgene was not thoroughly analyzed in prior studies, we first crossed the transgenic mice to two reporter strains, the R26-LacZf/+ and the R26-YFPf/+ lines, which express LacZ or YFP, respectively, upon Cre-mediated recombination. YFP expression in Nkx3.2-Cre;R26-YFPf/+ embryos (from here on referred to as Nkx3.2/YFP) was not found in the endodermally-derived pancreatic epithelium marked by E-Cadherin and Pdx1 at e9.5 [26], but was ubiquitously detected in the surrounding mesenchyme (Figure 1A). Similarly, X-gal staining in Nkx3.2-Cre;R26-LacZf/+ (from here on referred to as Nkx3.2/LacZ) indicated LacZ expression was confined to the surrounding mesenchyme at e11.5 (Figure 1B). At p0, Nkx3.2/LacZ and Nkx3.2/YFP expressing cells with fibroblast-like morphology were observed around islets, ducts, and blood vessels (Figure 1C,C′,D). Importantly, we could not detect reporter genes' expression in either epithelial (Figure 1C,C′,D), endothelial, or neuronal cells (Figure S1), indicating that Nkx3.2-Cre activity is excluded from those compartments throughout pancreatic development. Thus, the Nkx3.2-Cre line directs Cre-activity exclusively to the mesenchyme during pancreas development and serves as a novel tool to specifically manipulate embryonic gene expression in this tissue.
General histological analysis implied that the relative proportion of mesenchyme to epithelium shifts during pancreas organogenesis as epithelial cell numbers expand. We therefore took advantage of Nkx3.2-Cre transgenic mice to quantify the mesenchymal area during different developmental stages. By measuring the percentage of the pancreatic area marked by Nkx3.2/LacZ and Nkx3.2/YFP cells at various developmental stages, we determined that while the relative mesenchymal area is significantly reduced during pancreas organogenesis, it still comprised 11% and 6% of the pancreatic area at e15.5 and e18.5, respectively (Figure 1E). Thus, although there is a dramatic reduction in their portion over time, embryonic mesenchymal cells are present throughout pancreas organogenesis.
Next, we tested the requirement for mesenchyme during pancreas organogenesis in vivo. Studies using cultured pancreatic rudiments as well as Fgf10 knockout mice demonstrated a crucial role for the mesenchyme in expanding the pool of epithelial pancreatic precursor cells at early developmental stages (e9.5–e11.5) [3],[14],[27],[28]. In order to determine the role of the mesenchyme during pancreas development in vivo, we decided to ablate this tissue by employing transgenic mice carrying the Diphtheria Toxin (DT) active A subunit (DTA) flanked by flox sites (R26-eGFP-DTA mice [29], from here on referred to as DTA). Upon Cre-mediated recombination, the DTA produced by the transgene inhibits protein synthesis, resulting in rapid apoptosis of Cre-positive cells within less than 24 h [29]. Given that Nkx3.2-Cre is expressed in mesenchymal cells surrounding the pancreas from the time organ morphogenesis is initiated (e9.5, Figure 1A), Nkx3.2-Cre;DTA embryos permit the study of mesenchymal requirement at early stages of pancreas development (illustrated in Figure 2A).
We first analyzed potential defects in e10.5 embryos. At this stage, Nkx3.2-Cre;DTA embryos presented with Pdx1+E-Cadherin+ epithelial pancreatic cells (Figure 2B,C). However, while non-transgenic control pancreatic epithelial cells were completely surrounded by E-Cadherin− mesenchymal cells, Nkx3.2-Cre;DTA embryos lacked most of the adjacent mesenchymal cell layer (Figure 2B,C). To assess potential defects in pancreatic bud morphology, we performed whole mount staining with the epithelial marker E-Cadherin. In wild-type embryos this staining revealed the expected organization of stomach, liver, and ventral and dorsal pancreatic buds (Figure 2D). In contrast, pancreatic buds of Nkx3.2-Cre;DTA embryos were severely reduced in size and did not evaginate from the foregut epithelium (Figure 2E).
Nkx3.2-Cre;DTA transgenic mice suffered from embryonic lethality starting at e15.5 as well as severe skeletal defects (Figure 2F,G) resulting from Nkx3.2-Cre activity in the somites [21],[22],[24]. Although the few viable Nkx3.2-Cre;DTA embryos recovered at e15.5 were only slightly smaller than non-transgenic littermates (Figure 2F,G), their gastrointestinal tract was dramatically reduced in size (Figure 2H,I), likely due to Nkx3.2-Cre-mediated expression of DTA in the mesenchyme surrounding these tissues [23],[24]. Notably, while pancreatic tissue was clearly detected in non-transgenic embryos at this stage (Figure 2H, demarcated by the white line), Nkx3.2-Cre;DTA embryos had no visible pancreatic tissue (Figure 2I). Histological analysis of gut rudiments confirmed the gross morphology observation and showed only intestine-like tissue in Nkx3.2-Cre;DTA embryos, with no discernable stomach, spleen, or pancreatic tissues (Figure 2J,K). Thus, elimination of mesenchyme at the earliest stages of pancreas formation leads to complete agenesis caused by the inability of pancreatic epithelium to evaginate from the forming gut and to expand.
Pancreas development is a multistep process during which the epithelium undergoes complex morphological changes while common precursor cells differentiate into the various cells types that form the adult pancreas [30]. To test whether mesenchymal cells play distinct roles during different stages of pancreas development, we depleted the mesenchyme at various time points by injecting DT into developing embryos. Unlike primates, rodent cells lack a high affinity receptor for DT and therefore do not endocytose the toxin [31]. Since DT internalization into the cell cytoplasm is crucial for its ability to trigger the apoptotic machinery, rodent cells are resistant to ectopically administrated DT. However, mouse cells expressing a human DT Receptor (DTR) transgene, encoding for the human heparin binding epidermal growth factor (hbEGF), gain sensitivity to DT and are rapidly eliminated upon exposure to the toxin [32]. Prior studies have established that cell specific expression of human DTR in transgenic mice allows the ablation of targeted cells within 6 h following DT administration [33]. By crossing transgenic mice in which DTR expression is activated upon Cre-mediated recombination (iDTR [34], from here on referred to as DTR) with the Nkx3.2-Cre mice (Nkx3.2-Cre;DTR) we were able to specifically ablate the mesenchyme at different embryonic time points during pancreas development upon DT injection.
To ensure efficient delivery of DT to the developing pancreas, we injected the agent directly into embryos intraperitoneally (i.p.; the experimental procedure is illustrated in Figure 3A and Figure S2A–D) [35],[36]. As early as 4 h following DT injection into e13.5 Nkx3.2-Cre;DTR embryos, we observed an increase in apoptotic mesenchymal cells compared to controls (Figure S2E,F). One day after DT injection we detected only E-Cadherin expressing cells in Nkx3.2-Cre;DTR pancreata (Figure S2G,H), strongly indicating that E-Cadherin-negative mesenchymal cells were eliminated. The loss of Nkx3.2-Cre;DTR-positive mesenchymal cells was further confirmed by direct staining for human DTR expression (Figure S2I,J).
At the end of gestation (e18.5), Nkx3.2-Cre;DTR embryos injected with DT at e13.5 were viable and appeared grossly normal, with normal body weight (Figure 3B–D). At e18.5 the transgenic embryos displayed skeletal dysplasia, gastrointestinal defects, and asplenia (Figure 3B,C and Figure S3), likely a result of ablation of Nkx3.2-Cre expressing cells in these organs, and they died at birth. Therefore, in utero injections of DT into Nkx3.2-Cre;DTR do not cause embryonic lethality and permit studying the effects of mesenchyme ablation on epithelial pancreas development during embryogenesis.
To elucidate the requirement of mesenchyme at different stages we injected Nkx3.2-Cre;DTR embryos and non-transgenic littermates in utero with a single dose of DT at embryonic days 11.5, 12.5, 13.5, 14.5 ,15.5, or 16.5 (illustrated in Figure 3E). Embryos were then allowed to develop in situ until e18.5 when pancreata were dissected and weighed. Surprisingly, both dorsal and ventral pancreatic regions were significantly reduced in size in treated transgenic embryos independent of time of DT administration (Figure 3F–H). The most dramatic reduction in pancreas mass, up to 80%, was observed when transgenic embryos were injected between e11.5 and e13.5 (Figure 3H). DT injection at later stages, e14.5 and e15.5, resulted in an approximately 50% loss of pancreas mass. Notably, mesenchymal elimination as late as e16.5 led to a marked reduction in pancreas size to about two-thirds of non-transgenic littermates (Figure 3H). These results demonstrate that the mesenchyme is continuously required for proper pancreas development and organogenesis.
Next, we performed an in-depth analysis of pancreas morphogenesis and cell differentiation in transgenic animals in which mesenchyme was depleted mid-way through organogenesis (DT injections into e13.5 Nkx3.2-Cre;DTR embryos followed by analysis at e18.5; DT e13.5→e18.5).
When compared to normal tissues [37], mesenchyme-ablated pancreata displayed an abnormal globular morphology. DT-treated Nkx3.2-Cre;DTR pancreata were smooth and lacked the typical extension of the left branches (Figure 4A,B) as well as the gastric lobe (Figure 3F,G). In addition, DT-treated transgenic pancreata presented with a rounded tail instead of the stereotypical anvil-shaped tail found in non-transgenic controls (Figure 4A,B) [37]. Histological analysis further revealed more compacted cellular distribution in mesenchyme-ablated pancreata as compared to control, as shown by severe reduction of typical acellular areas normally found between adjacent lobes (Figure 4C,D).
Staining for the endothelial cell marker PECAM1 revealed that pancreatic vasculature, known to be crucial for organ development [38],[39], was not overtly disrupted in DT-treated transgenic pancreata (Figure S4A–D). Similarly, Tuj1 (β-III Tubulin)-expressing neuronal cells, known to be required for proper endocrine differentiation [40],[41], could be found in pancreata of DT-treated Nkx3.2-Cre;DTR embryos (Figure S4E,F).
Previous in vitro studies have implied that mesenchymal cells may control the differentiation of pancreatic epithelial cells [6],[9]. In order to study the in vivo effect of the mesenchyme on epithelial cell differentiation, we analyzed pancreatic tissues from DT-treated Nkx3.2-Cre;DTR for the expression of exocrine and endocrine markers at the end of gestation (DT e13.5→e18.5). Normal expression patterns for both the duct cell marker Mucin1+ and the acinar cell marker Amylase+ in treated transgenic pancreas indicated normal exocrine differentiation (Figure 4E,F). Furthermore, endocrine differentiation was not disturbed by mesenchymal ablation as Insulin, Glucagon, and Somatostatin expressing cells could be detected in transgenic pancreata (Figure 4G,H). Endocrine cells were single-hormone positive, clustered in islet-like structures typical for this developmental stage, and were distributed throughout the pancreas in a normal pattern (Figure 4C,D,G,H). Moreover, β-cells from transgenic embryos expressed the transcription factor MafA (Figure 4I,J), which is critical for full maturation and glucose responsiveness [42], strongly indicating that mesenchyme ablation does not block their differentiation potential. Thus, while pancreas morphogenesis is impaired upon mesenchyme elimination after the first stages of pancreas formation, differentiation of the major cell types was not blocked.
Although each of the specific pancreatic epithelium lineages formed in mesenchyme-depleted pancreata, the dramatic reduction in pancreatic organ size suggests a decrease in the overall number of pancreatic epithelial cells. In order to understand whether mesenchymal ablation affects either endocrine or exocrine mass, Nkx3.2-Cre;DTR mice treated with DT at e13.5 (illustrated in Figure 5A) were analyzed at e18.5 for β- and acinar cell masses. Both Insulin+ β-cell and Amylase+ acinar-cell mass were significantly reduced in transgenic mice when compared to non-transgenic littermates (Figure 5B,C), suggesting a requirement for mesenchymal cells during the expansion of both exocrine and endocrine compartments/precursors. The observation that transgenic pancreata maintained a normal acinar to β-cell ratio (Figure 5D) indicates that both cell types depend in equal measures on mesenchymal signals for their proliferation.
To determine the developmental stage during which mesenchyme ablation affects pancreatic mass, we injected embryos at e13.5 and investigated pancreata 2 d later at e15.5 (DT e13.5→e15.5). At that stage, pancreas mass in transgenic embryos was already reduced by 80% as compared to controls (Figure 5E), similar to the reduction observed in pancreata injected at e13.5 and analyzed at e18.5 (Figure 3H). Since mesenchymal cells comprise only 11% of pancreatic tissue at e15.5 (Figure 1E), the observed reduction in pancreatic weight was likely due to a rapid and significant loss of the epithelial compartment of the organ.
Next, we investigated whether cells of either the endocrine or exocrine compartments were already affected in e15.5 Nkx3.2-Cre;DTR embryos that were DT-treated 2 d before (i.e., at e13.5). While present in DT-treated transgenic pancreata (Figure 5F,G), the number of cells positive for Neurogenin 3 (Ngn3), a transcription factor that marks endocrine precursor cells [43], was significantly reduced compared to littermate control mice (Figure 5H). Similarly, the number of cells expressing Ptf1a, a transcription factor found in exocrine precursor and differentiated acinar cells [44], was significantly reduced 2-fold in DT e13.5→e15.5 Nkx3.2-Cre;DTR pancreata as compared to non-transgenic controls (Figure 5I–K). Therefore, the reduction in β-cell and acinar cell mass detected at DT e13.5→e18.5 Nkx3.2-Cre;DTR embryos (Figure 5B,C) is, at least in part, due to the decreased number of Ngn3+ precursor cells and Ptf1a+ exocrine cells at earlier developmental stages.
Since pancreatic growth between e13.5 and e15.5 relies heavily on proliferation of precursor cells [45], we next analyzed the effect of mesenchymal depletion on the proliferation rate of these cell populations in e14.5 Nkx3.2-Cre;DTR embryos treated at e13.5 (DT e13.5→e14.5). Epithelial tip cells serve as multi-potent progenitors before they become committed to the exocrine lineage around e14.5 [44]. Staining these cells, identified as Carboxypeptidase 1 (Cpa1) expressing cells, with an antibody against phosphorylated Histone H3, a marker of cell proliferation, revealed a 50% reduction in proliferating tip cells in Nkx3.2-Cre;DTR embryos compared to controls (Figure 5L–N). In addition, we analyzed proliferation of Sox9 expressing cells, a transcription factor that marks epithelial precursor cells giving rise to exocrine cells as well as to Ngn3+ endocrine precursors [46]–[48]. The percentage of Sox9+ proliferating cells was slightly but significantly smaller in transgenic embryos (Figure 5O–Q). We could not detect apoptotic epithelial cells by TUNEL (terminal deoxynucleotidyl transferase dUTP biotin nick end labeling) assays (unpublished data), concluding that depletion of mesenchymal cells affects both endocrine and exocrine mass through reduced proliferative capacity of epithelial progenitor cells rather than their apoptosis.
The reduced proliferative potential of progenitor cells at e14.5 explained, at least in part, the reduction in pancreas mass in embryos treated with DT at e13.5. However, when mesenchyme was eliminated at e16.5 we also observed a significant reduction of about 35% in pancreas mass at e18.5, affecting both the endocrine and exocrine compartments (DT e16.5→e18.5; Figures 3H, 6A–D). By e16.5, the various pancreatic cell types are committed towards their final differentiation fate and present with many of their mature cell characteristics. Since pancreatic growth at those late stages of development is attributed to proliferation of these differentiated cells [49], the decrease in pancreatic mass could not be due to reduced proliferation of progenitor cells. While previous studies did not detect effects of mesenchymal cells on β-cell proliferation in culture [11], in vivo analysis of Nkx3.2-Cre;DTR embryos treated with DT at e16.5 and analyzed at e17.5 revealed decreased proliferative potential of both insulin and amylase expressing cells (Figure 6E–J). In agreement with what we had found at earlier stages, the ratio between Insulin+/Amylase+ areas was not affected in the DT-treated embryos (Figure 6D), suggesting mesenchymal factors have similar effect on cells of these two compartments.
Upon determining the requirement for mesenchymal cells to guide epithelial organ formation throughout development, we set out to identify signals and pathways critical for the mesenchymal effects. Canonical Wnt signaling is active in the developing pancreas, and both the mesenchyme and the epithelium express various Wnt ligands and receptors in a dynamic fashion [50]. At e11.5, Wnt signaling is observed in epithelial cells, and its level of activation declines in the following embryonic days [51], while its activity in the mesenchymal layer has been first reported around e13.5 [15],[52].
In order to directly investigate the role of mesenchymal Wnt signaling in pancreas development, we decided to block this pathway specifically in the mesenchyme by crossing transgenic mice carrying floxed alleles of β-catenin (βcatf/f), an essential mediator of canonical Wnt signaling, with Nkx3.2-Cre mice. In addition to its critical role in Wnt signaling, β-catenin has other functions within cells, most notably in maintaining cell-cell interactions as part of a complex with E-Cadherin. However, in pancreatic mesenchymal cells we failed to observe membrane-associated localization of the β-catenin protein (Figure S5A). Therefore, elimination of this gene in Nkx3.2-Cre;β-catf/f pancreata is unlikely to perturb cell-cell interactions but should reveal the requirement for β-catenin mediated Wnt signaling in mesenchyme.
As expected, elimination of β-catenin did not affect epithelial size at e12.5 (Figure S5B) prior to the reported onset of mesenchymal Wnt signaling. In contrast, Nkx3.2-Cre;β-catf/f pancreata were markedly reduced in size at e15.5 and e18.5 (Figure 7A,B), indicating that mesenchymal β-catenin signaling is critical for organ formation at later stages. In addition, Nkx3.2-Cre;βcatf/f pancreata exhibited aberrant morphology with diminished branching when compared to controls (Figure 7A,C,D).
In order to identify the potential effects on pancreatic epithelial development in Nkx3.2-Cre;βcatf/f embryos, we stained e18.5 knock-out pancreata for various cell markers and assessed acinar- and β-cell mass. All major pancreatic cell types, both of the exocrine (acinar and duct cells, Figure 7E,F) and of the endocrine compartments (α-, β-, and δ-cells, Figure 7G,H), were detected in the Nkx3.2-Cre;βcatf/f e18.5 pancreata. However, both β-cell and acinar-cell mass was significantly reduced in knock-out embryos (Figure 7I,J). Interestingly, the ratio between β- and acinar cells was maintained in Nkx3.2-Cre;βcatf/f pancreata (Figure 7K).
The Wnt signaling pathway was shown to become activated in the pancreatic mesenchyme around e13.5 [15],[52]. To address whether the reduction in pancreatic mass observed in Nkx3.2-Cre;βcatf/f at e15.5 and e18.5 is due to effects on epithelial growth at earlier stages, we studied epithelial proliferation in these mice at e13.5. At this stage, proliferating Cpa1+ tips cells serve as multipotent pancreatic progenitor cells for both endocrine and exocrine populations [44]. As shown in Figure 7L, the proliferation rate of Cpa1+ cells was significantly lower in Nkx3.2-Cre;βcatf/f embryos as compared to controls. Cell death was not apparent as we could not detect apoptotic epithelial cells by TUNEL assays or by staining for cleaved Caspase3 (unpublished data). Therefore, blocking mesenchymal Wnt signaling leads to reduced pancreatic mass by affecting the proliferation capacity of epithelial precursor cells.
Wnt signaling is known to regulate cell survival and proliferation [53]. Since pancreata from Nkx3.2-Cre;βcatf/f mice phenocopied those from DT-treated Nkx3.2-Cre;DTR mice, we wondered whether Wnt signaling is required for mesenchymal cell survival. Indeed, while e13.5 pancreatic tissue from wild type embryos contained both E-Cadherin-positive epithelial cells and E-Cadherin-negative mesenchymal cells (Figure 7M), we could detect only E-Cadherin expressing cells in Nkx3.2-Cre;βcatf/f tissues (Figure 7N), indicating ablation of the pancreatic mesenchymal layer in transgenic mice. Thus, our results point to mesenchymal Wnt signaling as a critical mediator of mesenchymal cell survival in vivo and therefore of epithelial growth and patterning.
Despite extensive efforts, the role of the mesenchyme during in vivo pancreas development has remained elusive. Due to the absence of suitable genetic tools, probing the role of the pancreatic mesenchyme during organogenesis has been mainly restricted to organ rudiment culture experiments. In addition to the inability to faithfully mimic the in vivo conditions, clean separation of mesenchyme and epithelium for culture experiments is limited to early stages of pancreas organogenesis before the epithelial layer has integrated into the overlying mesenchyme (prior to e12.5 in the mouse). Here, we show that Nkx3.2-Cre mice permit transgene manipulation specifically in mesenchyme, but not in other pancreatic tissues. By using this tool to eliminate pancreatic mesenchymal cells at will, we expand upon classical tissue culture studies and for the first time present a model system suitable for detailed analysis of the various functions of this supporting tissue in vivo at multiple gestational ages.
A key finding of our studies is the observation that pancreatic mesenchyme provides critical functions for the proper development of the epithelial compartment throughout organogenesis. As expected from previous studies [3],[14], depletion of mesenchymal cells at the onset of pancreas development using Nkx3.2-Cre;DTA mice arrests pancreas organogenesis. Mesenchymal ablation at later stages, by injecting DT into Nkx3.2-Cre;DTR transgenic embryos at various stages or by blocking mesenchymal Wnt signaling in Nkx3.2-Cre;βcatf/f mice, impairs pancreatic epithelium growth and branching. Notably, while the morphological changes observed are profound, we did not observe alterations in cell differentiation capacity of the three main pancreatic cell types, the endocrine, acinar, and duct cells. However, our results clearly demonstrate a requirement of the mesenchyme for the expansion of epithelial progenitor cells, as well as proliferation of differentiated pancreatic cells.
Factors secreted by the pancreas mesenchyme have previously been shown to regulate pancreas organogenesis [6], including Fgf10 whose function is required for the expansion of common epithelial progenitor cells during early stages of pancreas development [14]. The pancreatic defects we observe in Nkx3.2-Cre;DTA embryos are more severe than those previously reported for Fgf10−/− pancreata [14], a finding likely explained by the absence of mesenchymal cells, and thus reduction of all mesenchymal factors, in transgenic mice.
Our results further demonstrate a requirement for mesenchymal cells in promoting proliferation of various epithelial cell types, including precursors and differentiated cells. While it is theoretically possible that these functions are mediated by a limited number of factors throughout all stages of development, the dynamic activation of mesenchymal signaling pathways (summarized in [1]) would suggest a more complex interplay of a diverse set of molecules that changes over time. Our findings also suggest that mesenchyme supports proliferation of multiple distinct cell types, even during the same developmental stage. For instance, mesenchyme ablation has similar effects on proliferation of mature acinar and β-cells towards the end of gestation. This observation poses the question as to whether different epithelial cell types rely on the same mesenchymal factor(s) for their proliferation, or whether these processes are mediated by distinct signals. Future analysis is required to identify secreted factors expressed by the pancreatic mesenchyme at different developmental stages. The use of the Nkx3.2-Cre line will allow specific manipulation of the genes coding for these signals to ascertain their role during pancreas organogenesis.
Another important finding concerns the observation that the pancreatic mesenchyme is required for both endocrine and exocrine development in vivo. Previous reports had reached differing conclusions, with some demonstrating a positive role for the mesenchyme on exocrine formation but not endocrine cell development [6],[7],[9], and others indicating that mesenchymal factors promote proliferation of multi-potent pancreas progenitors that subsequently increase the formation of endocrine cells [11]. Some of these conflicting results can be explained by the different culture conditions used in each experiment.
In contrast to the cultured studies, in vivo depletion of the mesenchyme investigated here revealed similar requirements for this tissue with regard to the endocrine and exocrine cytodifferentiation. At this point, we cannot exclude that other cells types, including endothelial and neural-crest derived cells [38]–[41], or cells residing in the adjacent liver, stomach, gut, or kidneys might provide signals that guide epithelial cell differentiation in mesenchyme depleted embryos in vivo. In addition, mesenchymal cells that did not originate from Nkx3.2-Cre expressing cells might still be present in our in vivo model and could provide either instructive or permissive signals.
Prior organ culture studies proposed another model to explain the various effects of the mesenchyme on the epithelial compartments by demonstrating distinct effects of the mesenchyme on epithelial cells depending on the physical distance and contact between these tissues [5]. In these experiments, close proximity between epithelial and mesenchymal cells promoted exocrine differentiation while at the same time blocked endocrine formation. In contrast, mesenchyme factors supported endocrine differentiation at a distance, indicating that the physical relation between mesenchymal and epithelial cells is critical for endocrine versus exocrine differentiation. Our studies support the notion of mesenchymal signals being important for both endocrine and exocrine development. However, our lineage tracing experiments provide evidence of close physical contact between Nkx3.2/LacZ and Nkx3.2/YFP expressing cells with endocrine cells, indicating that close proximity between mesenchymal and epithelial cells does not necessarily interfere with endocrine differentiation. However, since mesenchymal cells surround islets, they are likely in close contact only with peripheral endocrine cells, such as α-cells, while direct interactions with centrally located β-cells might not be common. Whether the mesenchyme contributes to β-cell expansion by releasing secreted factors or through cell-cell interactions as well as how the mesenchyme affects other endocrine cells are questions that need to be addressed in future experiments. Furthermore, isolation and characterization of mesenchymal cells throughout development might reveal cell heterogeneity that could explain differential functions with regard to promoting endocrine versus exocrine development.
Our results also point to sustained mesenchyme function as a critical regulator of epithelial pancreas development and identify Wnt signaling as an essential mediator of mesenchyme survival. It is not clear as to whether Wnt signaling is activated in an autocrine or paracrine manner, as several Wnt ligands are expressed by both pancreatic epithelial and mesenchymal cells during development [50]. It is noteworthy that the defects we observe in Nkx3.2-Cre;βcatf/f only occur after the onset of canonical Wnt signaling in pancreas mesenchyme as measured by expression of transgenic Wnt-reporters (i.e., e13.5 [15],[52]). The implication of canonical Wnt signaling as the cause for the observed phenotypes is indirectly supported by a previous study using germ-line knock-out mice in which mPygo2, a critical component of the nuclear β–catenin/Tcf complex required for β-catenin transcriptional activity, has been eliminated [15]. mPygo−/− mice show pancreas hypoplasia and a reduction in endocrine mass [15], phenotypes that are not observed when this gene is specifically eliminated in pancreas epithelium. Thus, while mesenchyme specific depletion of mPygo2 has not been reported, the absence of pancreas hypoplasia upon epithelial-specific mPygo2 elimination suggests that at least some of the pancreatic defects are caused by reduced mesenchymal Wnt signaling. However, and in contrast to Nkx3.2-Cre;βcatf/f pancreata, the exocrine compartment is not affected in mPygo2−/− mutants and mesenchyme depletion was not reported in those mice. Since Wnt signaling is significantly reduced, but not completely blocked in the absence of mPygo2 [15], it is possible that low level of canonical Wnt signaling is sufficient for mesenchymal cell survival and the production of factors that promote exocrine cell development. Alternatively, β-catenin is known to regulate cell-cell interactions as part of Cadherin complexes and these additional functions might be crucial for the maintenance of the pancreatic mesenchyme. However, we did not observe β-catenin localized to membranes in mesenchymal cells. In order to study whether different levels of mesenchymal Wnt signaling have a different effect on endocrine and exocrine expansion, mice specifically lacking mesenchymal expression of various components of this pathway (such as mPygo2) would need to be examined.
In addition to Wnt signaling, other signaling pathways, such as the RA, BMP, and Hedgehog, have been implicated as mesenchymal factors regulating pancreas development [15]–[17],[54],[55]. Using Nkx3.2-Cre line as a novel tool to manipulate gene expression in the pancreatic mesenchyme will allow direct study of the role of these and potentially other pathways in pancreas organogenesis.
In summary, data presented here indicate continuous requirement of mesenchymal cells and/or mesenchyme-derived signals to regulate epithelial pancreas formation from the onset of organ morphogenesis until the end of gestation. Isolation of mesenchymal cells at different stages of pancreas formation might allow identification of candidate factors that regulate expansion of common and endocrine progenitors as well as of differentiated β-cells. Future therapies for both type I and II diabetes rely on renewable sources of functional insulin-producing β-cells [56]. Current protocols allow the formation of pancreas progenitor cells from human embryonic stem cells (hESC) in vitro, but not fully differentiated β-cells. Our results demonstrate that mesenchymal factors provide critical signals for the expansion of both precursors and differentiated endocrine and exocrine cells. Thus, mesenchymal signaling factors not yet identified will likely be useful for expansion of hESC derived pancreas progenitor and differentiated β-cells.
Mice used in this study were maintained according to protocols approved by the Committee on Animal Research at the University of California, San Francisco. Nkx3.2 (Bapx1)-Cre mice were described previously [24]. R26-YFPflox (Gt(ROSA)26Sortm1(EYFP)Cos), R26-LacZflox (Gt(ROSA)26Sortm1Sor), R26-eGFP-DTA (Gt(ROSA)26Sortm1(DTA)Jpmb), DTR (iDTR, Gt(ROSA)26Sortm1(HBEGF)Awai), and β-cateninflox (Ctnnb1tm2Kem) mice were obtained from Jackson Laboratories. Noon on the day a vaginal plug was detected was considered as embryonic day 0.5.
Injections were preformed as previously described [35],[36]. Briefly, pregnant females were anesthetized, a laparotomy was performed, and the uterus was delivered through the incision (as illustrated in Figure S2A–D). Each embryo was micro-injected with 8 ng/gr body weight Diphtheria Toxin (Sigma) diluted in 5 µl PBS. The uterus was placed back into the abdominal cavity and the laparotomy was closed. Embryos were allowed to develop in situ until indicated stages.
For immunofluorescence, dissected embryos and pancreatic tissues were fixed with Z-fix (Anatech) for 2–16 h, embedded in paraffin wax, and sectioned. For Ptf1a staining, tissues were fixed with Z-fix for 2 h, embedded in OCT (Tissue Tek), and cryosectioned. Tissue sections were stained using the following primary antibodies: rabbit anti-Amylase (1∶200, Sigma), goat anti-Cpa1 (1∶200, R&D), mouse anti-E-Cadherin (1∶200, BD), rabbit anti-Glucagon (1∶200, Linco), guinea pig anti-Insulin (1∶200, Linco), mouse anti-Ki67 (1∶200, BD), rabbit anti-MafA (1∶200, Bethyl), armenian hamster anti-Mucin1 (1∶200, Neomarker), guinea pig anti-Neurogenin 3 (1∶400, Millipore), rabbit anti-phosphorylated Histone H3 (1∶200, Millipore), rabbit anti-Pdx1 (1∶200, Millipore), rabbit anti-Ptf1a (1∶600, a gift from Dr. Helena Edlund), rat anti-Somatostatin (1∶200, Chemicon), rabbit anti-Sox9 (1∶200, Chemicon), and chicken anti-YFP/GFP (1∶400, Abcam) followed by staining with Alexa Fluor tagged secondary antibodies (1∶500, Invitrogen) and mounting with DAPI-containing Vectashield media (Vector). For TUNEL analysis, ApopTag Plus Fluorescein In Situ Apoptosis Detection kit (Millipore) was used according to the manufacturer's protocol.
For embryo wholemount staining, tissues were processed as previously described [57] and stained with rat anti-E-Cadherin (1∶1,000, CalBiochem), followed by staining with Alexa Fluor 555 anti-rat secondary antibody (1∶500, Invitrogen).
For x-gal staining, tissues were fixed with 2% PFA and 0.25% Glutaraldehyde for 2 h and incubated overnight with 0.5 mg/ml x-gal solution (Roche), followed by a second round of fixation in 4% PFA overnight. Tissues were then embedded in paraffin, sectioned, and counter-stained with nuclear Fast Red (Vector).
For histological analysis, dissected tissues were fixed with Z-fix (Anatech), for 4 h, and embedded in paraffin wax. Tissue sections were stained with Meyer's Hematoxylin (Sigma) followed by staining with Eosin (Protocol).
Images were acquired using Zeiss ApoTome, Leica MZ FL3 and SP5, and Olympus IX70 microscopes.
For all quantifications presented in this study, each transgenic tissue was processed and stained in parallel with a littermate control, with each analyzed group comprising at least three pairs of transgenic and control embryos (i.e., n≥3) as indicated in the figure legends. Throughout each analysis, images were acquired using the same exposure time and magnification. When MetaMorph software was used for image analysis, the same signal-to-noise threshold was applied throughout the experiment. For all measurements presented in this study, with the exception of the measurement of the mesenchymal area at e11.5 and Ptf1a+ cell numbers at e15.5, the following regimen was applied: the entire pancreatic tissue, including both dorsal and ventral buds, was embedded in paraffin wax and cut into 5 µm thick sections. Every fifth section (20% of total tissue) was then immuno-stained with indicated antibodies as described above. Images were acquired as detailed below and analyzed blindly.
For measurement of mesenchymal areas at e15.5 and e18.5, isolated pancreatic tissues from Nkx3.2-Cre;R26-YFPflox embryos were stained with an anti-YFP antibody and a fluorescent secondary antibody and entire sections were automatically imaged using Olympus IX70 widefield microscope and MetaMorph software. Over-exposure of the tissue and DAPI staining were used to determine the edges of the section. Images were analyzed using MetaMorph software, which automatically measured the positive area in each channel. To determine the percentage of mesenchymal area, total YFP-positive area was divided by total tissue area of each section.
For β- and acinar cell mass, isolated e18.5 tissues (including both dorsal and ventral tissues) were dissected and weighed. Following fixation, tissues were embedded in paraffin wax, sectioned as described above, and immuno-stained with anti-Insulin and anti-Amylase antibodies. Images were acquired as described above for mesenchymal area measurement, and areas positive for either Amylase or Insulin, as well as the total pancreatic area, were automatically measured using MetaMorph software. To determine the fractions of the β- and acinar cell areas, total Insulin or Amylase positive area was divided by total tissue area. Cell mass was calculated as the fraction of Amylase+ or Insulin+ areas of the total pancreatic area multiplied by gross pancreas weight.
To calculate the β-cell/acinar cell ratio, for each embryo Insulin+ and Amylase+ area was determined as described above for cell mass measurement, and Insulin+ area was divided by Amylase+ area. For clarity, the ratio obtained in non-transgenic controls was set to “1.”
For quantification of Ngn3-expressing cells, whole e15.5 pancreatic tissues were isolated and processed as described above. Sections were stained with anti-Ngn3 antibody followed by fluorescent secondary antibody and images were then acquired as described above for mesenchymal area measurement, but positive cells were counted manually. To accommodate for potential differences in the developmental stage of the various litters analyzed, the number of transgenic Ngn3-positive cells was normalized to the number of Ngn3 cells counted in the corresponding non-transgenic littermate controls.
For cell proliferation, whole pancreatic tissue (including both dorsal and ventral tissues) was isolated from embryos e15.5 and older. From embryos at e13.5 or e14.5, pancreatic tissue was isolated together with the adjacent stomach and duodenum. Following fixation, tissues were paraffin-embedded and sectioned as described above. Tissue sections were stained with indicated antibodies and imaged using Zeiss ApoTome or Leica SP5 microscopes. For each section, the percentage of proliferating cells was determined via manual counting of either Ki67 or pHH3 positive cells divided by the number of total target cells.
To determine mesenchymal area at e11.5, Nkx3.2-Cre;R26-LacZf/+ embryos were stained with X-gal as described above. Entire embryos were then cut to obtain 5 µm thick sections, and all sections were counterstained with FastRed dye and imaged using Zeiss ApoTome. The total dorsal pancreatic bud area, identified by its typical localization and morphology, and pancreatic mesenchyme area, identified by blue x-gal staining, were manually selected and measured using MetaMorph software.
For Ptf1a+ cell quantification, isolated e15.5 pancreatic tissue were fixed, embedded in OCT, frozen, and cryosectioned. 10 µm thick sections were used and every 10th section was stained (10% of total tissue). Whole sections were imaged using Leica SP5 confocal microscope and the number of positive cells was counted manually. To account for potential differences in developmental stage of each litter, the number of positive cells obtained for each transgenic animal was normalized to the number obtained from the non-transgenic littermate control.
P values were determined using unpaired, two-tailed student t test. Error bars in bar diagrams represent standard deviation of the samples.
|
10.1371/journal.pgen.1004712 | Copy Number Variation in the Horse Genome | We constructed a 400K WG tiling oligoarray for the horse and applied it for the discovery of copy number variations (CNVs) in 38 normal horses of 16 diverse breeds, and the Przewalski horse. Probes on the array represented 18,763 autosomal and X-linked genes, and intergenic, sub-telomeric and chrY sequences. We identified 258 CNV regions (CNVRs) across all autosomes, chrX and chrUn, but not in chrY. CNVs comprised 1.3% of the horse genome with chr12 being most enriched. American Miniature horses had the highest and American Quarter Horses the lowest number of CNVs in relation to Thoroughbred reference. The Przewalski horse was similar to native ponies and draft breeds. The majority of CNVRs involved genes, while 20% were located in intergenic regions. Similar to previous studies in horses and other mammals, molecular functions of CNV-associated genes were predominantly in sensory perception, immunity and reproduction. The findings were integrated with previous studies to generate a composite genome-wide dataset of 1476 CNVRs. Of these, 301 CNVRs were shared between studies, while 1174 were novel and require further validation. Integrated data revealed that to date, 41 out of over 400 breeds of the domestic horse have been analyzed for CNVs, of which 11 new breeds were added in this study. Finally, the composite CNV dataset was applied in a pilot study for the discovery of CNVs in 6 horses with XY disorders of sexual development. A homozygous deletion involving AKR1C gene cluster in chr29 in two affected horses was considered possibly causative because of the known role of AKR1C genes in testicular androgen synthesis and sexual development. While the findings improve and integrate the knowledge of CNVs in horses, they also show that for effective discovery of variants of biomedical importance, more breeds and individuals need to be analyzed using comparable methodological approaches.
| Genomes of individuals in a species vary in many ways, one of which is DNA copy number variation (CNV). This includes deletions, duplications, and complex rearrangements typically larger than 50 base-pairs. CNVs are part of normal genetic variation contributing to phenotypic diversity but can also be pathogenic and associated with diseases and disorders. In order to distinguish between the two, detailed knowledge about CNVs in the species of interest is needed. Here we studied the genomes of 38 normal horses of 16 diverse breeds, and identified 258 CNV regions. We integrated our findings with previously published horse CNVs and generated a composite dataset of ∼1400 CNVRs. Despite this large number, our analysis shows that CNV research in horses needs further improvement because the current data are based on 10% of horse breeds and that most CNVRs are study-specific and require validation. Finally, we analyzed CNVs in horses with disorders of sexual development and found in two male pseudo-hermaphrodites a large deletion disrupting a group of genes involved in sex hormone metabolism and sexual differentiation. The findings underline the possible role of CNVs in complex disorders such as development and reproduction.
| The significance of gene duplication in long-term evolutionary changes was already recognized over 40 years ago by Susumu Ohno [1]. Yet, systematic genome-wide discovery and functional interpretation of inter- and intraspecific copy number variations (CNVs) in genes and non-genic DNA sequences, started in the past decade with foundational studies in humans [2], [3] and mice [4], followed by genome-wide (GW) CNV discovery in chicken [5], cattle [6], dogs [7], [8] and other domestic species (see [9], [10]). It is now well established that CNVs are a common feature of vertebrate genomes. Typically, they are DNA sequence variants from at least 50 base-pairs (bp) to over several megabase-pairs (Mb) in size that are involved in deletions, insertions, duplications and translocations, causing structural differences between genomes [11], [12]. In terms of the total number of DNA base-pairs, CNVs are responsible for more heritable sequence differences (0.5–1%) between individuals than SNPs (0.1%) [11], [12], [13].
One of the central goals of CNV research has been determining their association with genome instability, genetic diseases and congenital disorders. It is thought that CNVs, as a major source of inter-individual genetic variation, could explain variable penetrance of Mendelian and polygenic diseases, and variation in the phenotypic expression of complex traits [14], [15]. Indeed, CNVs have been associated with common complex and polygenic disorders in humans affecting a broad range of biological processes, such as immune response, autoimmunity and inflammation [3], [16], [17]; musculoskeletal [18], [19] and cardiovascular systems [20], [21]; neurodevelopment, cognition and behavior [22], [23], and sexual development and reproduction [24], [25], [26], [27], [28].
The availability of whole genome (WG) sequence draft assemblies combined with the advances in array-based technologies and next generation sequencing (NGS), have prompted CNV research in all main domestic animal species (reviewed by [9], [10]) with the most advanced information currently available for cattle [6], [29], [30], pigs [31], and dogs [32], [33], [34].
In horses, five studies report about the discovery of CNVs in the whole genome [35], [36], [37], [38] or in gene exons [39]. Attempts have also been made to associate CNVs with equine diseases [36], adaptations [38] and phenotypic traits [37], [39]. While these studies set a foundation for understanding the role of CNVs in equine biology, the current information is inadequate for efficient discovery of variants affecting equine health and disorders. This is because the studies have used different CNV discovery platforms, the number of breeds and individuals in some studies is very limited, and the majority of reported CNVs are study-specific and not validated by two or more independent studies. Also, the available information has not been integrated into a composite dataset to facilitate the analysis of known CNVs and the discovery of new ones.
The aim of this study is to improve the current rather limited knowledge of CNVs in horses by their genome-wide discovery in multiple individuals of additional diverse horse breeds. Using a custom-made WG tiling array we generate a CNV map for the horse genome and integrate this with the previous CNV studies into a composite dataset. Finally, we carry out a pilot CNV analysis in horses with disorders of sexual development to test the utility of the array and the integrated dataset for the discovery of variants involved in equine complex disorders.
Texas A&M University (USA) and The University of Adelaide (Australia) collaborated to create a whole-genome (WG) 400K tiling array which was produced and printed by Agilent Technologies (Design ID #030025), and designated as the Texas-Adelaide array. The probes on the array represented 18,763 autosomal and X-linked genes, and intergenic, sub-telomeric and chrY sequences. Median genomic distance between the probes on the array was 7.5 kb; this distance was lower (4 kb) in sub-telomeric regions, and higher (∼20 kb) in the Y chromosome. Before using the array for CNV discovery in horses, the platform was tested for performance quality. Self-to-self control hybridizations (Figure S1a) showed 1.55% of False Discovery Rate (FDR) - an indication that the array design, fabrication, and array genomic hybridization (aCGH) procedures were optimal. As a proof-of principle, female-to-male hybridizations between two half-sib Thoroughbreds, Twilight (female) and Bravo (male), showed massive loss in the X chromosome and a gain in the Y chromosome in the male, whereas only one CNV was detected in an autosome, chr3 (Figure S1b). Hybridization quality was assessed by measuring Derivative Log Ratio Standard Deviation (DLRSD) which calculates probe-to probe log ratio noise and is typically <0.3 for good quality hybridizations. The DLRSD values for all hybridizations with blood DNA from Twilight and Bravo were <0.2. Therefore, and because the oligonucleotides on the array were derived from the sequences of these two horses, DNA of Twilight and Bravo was used as a reference for all aCGH experiments: Twilight for females and Bravo for males. Further, because our DNA collection from horse breeds contained samples isolated from blood and hair, an additional self-to-self hybridization was conducted using DNA from blood and hair of one male Quarter Horse QH3-H528 (Table S1). Blood DNA gave good quality results with DLRSD = 0.14, whereas consistent and high level hybridization noise was observed for hair DNA (DLRSD = 0.41) (Figure S1c). Due to this, CNVs in all samples were called with stringent criteria: log2 ratio alterations higher than 0.5 over 5 neighboring probes – a necessary compromise between calling CNVs with confidence and missing a few true calls. With median probe spacing of 7.5 kb on the array, this allowed detection CNVs of about 30 kb, and in probe-dense regions even smaller. We concluded that the performance of the equine 400K Texas-Adelaide whole-genome CGH array was optimal for the discovery of CNVs in the horse genome.
The aCGH data are available at NCBI GEO accession GSE55266. Collectively, 950 CNV calls were made across 36 horses, with an average of 26.4 calls per individual (Table 1; Table S3). The number of CNV calls was the highest in two American Miniature Horses (59 and 46) and the lowest in American Quarter Horses (12 and 14), whereas the number of calls per individual was not significantly different between blood and hair DNA (P = 0.07; Table 1) at the settings of log2±0.5 over 5 probes. The number and distribution of CNVRs in the two Przewalski horses were similar to those in domestic horses (Table 1, Table S4). Because the Thoroughbred served as a reference, by default all the 950 CNV calls recorded in other breeds were also present in the Thoroughbred, though inversely with respect to gains and losses. However, because the Thoroughbred was compared with multiple individuals, the same CNV had different log2 values, and that is why the Thoroughbreds were not included in the comparisons of CNV metrics.
The ADM-2 algorithm arranged adjacent and overlapping CNV calls (CNVs) within and between individual horses into 258 CNV regions (CNVRs; Table S5) of which 114 were shared between at least 2 individuals of the same or different breeds, while 144 were private and found only in one individual. Two CNVRs were found in two or more individuals of the same breed but not in other breeds and were tentatively considered as breed-specific: a 14 kb loss in chr9 in Exmoor ponies, and a 39 kb loss in chr20 in Swiss Warmblood horses (Table S4).
Based on the 258 CNVRs, a whole genome CNV map for the horse was constructed (Figure 1) details of which are summarized in Table 2. The mean size of CNVRs was 110 kb ranging from 1 kb to 2.5 Mb. The CNVRs occupied 1.15 % of the equine genome and were distributed over all horse chromosomes, except the Y, with the highest enrichment in chromosomes 12 (9.7%) and 20 (3.0 %). Even though chr12 is the gene richest chromosome in the horse genome (15 genes/Mb), there was no overall correlation between CNV enrichment and gene density. For example, the enrichment values for the second and third gene densest chromosomes, chr11 and chr13, were 0.02% and 0.28%, respectively (Table 2). Likewise, we did not observe CNV enrichment in sub-telomeres, as previously reported for humans [40]: the array contained 5,716 sub-telomeric probes, though only 10 CNVRs were detected in these regions in horses.
In general, losses (173; 67%) prevailed over gains (63; 24%), although 6 horses had more gains than losses (Table 1). Twenty-two CNVRs (8.5%) were complex involving both losses and gains in different individuals (Table 2, Table S3). Even though aCGH on diploid samples cannot discriminate between copies of alleles and thus, distinguish between heterozygous and homozygous CNVs, two gains and 14 losses were tentatively considered homozygous because of log2 alterations over 2.0 (Table S6). Homozygosity of 8 losses was confirmed by qualitative PCR (Fig. S2).
The majority (82%) of horse CNVRs contained one or more known Ensembl (http://www.ensembl.org/index.html) horse genes (158 CNVRs) or non-horse mammalian reference genes (54 CNVRs) (Table S7), while 46 CNVRs (18%) were located in intergenic regions (Table S8). Gene containing CNVRs were also predominant in individual chromosomes with the exception of chr31 which was enriched with intergenic variants Fig. 2. However, we consider calls for intergenic CNVRs tentative and subject to change as the annotation of the horse genome is still in progress.
Altogether, the CNVRs involved 805 protein-coding genes (750 Ensembl genes, 33 non-Ensembl genes and 22 horse mRNAs; Table S7) but also non-coding small and long RNA genes, and pseudogenes. The largest CNVRs with the highest number of genes corresponded to clusters of olfactory and non-olfactory G-protein coupled receptors (GPCRs) or to immunity related genes, such as immunoglobulins, T-cell receptors, and MHC protein complex genes - a typical feature of CNVRs in all mammalian genomes studied so far [3], [30], [32], [39], [41], [42]. Likewise, Gene Ontology (GO) analysis indicated that equine copy number variable genes are predominantly involved in biological processes and molecular functions related to transmembrane signal transduction, chemo-attractant sensory perception, immune response and steroid metabolism (Fig. 3; Table S9). Notably, 5 copy number variable genes from this study were associated with known OMIA (http://omia.angis.org.au/home/) phenotypes for immune, reproductive or neuromuscular diseases (Table 3), though none of the OMIA records involved horses or CNVs. The CNVR overlapping with the BMPR1B gene has been earlier reported in horses and is of interest because of a possible role in the regulation of the rate of ovulation [39].
Comprehensive knowledge of CNVs in normal horse populations, within and across breeds, is a prerequisite for the discovery of variants that contribute to equine genetic diseases and disorders. Therefore, we aligned the 258 CNVRs identified in this study with previously published CNV data for the horse [35], [36], [37], [38], [39]. Altogether, we found records of about 2041 CNVs and CNVRs (calling criteria vary between studies). These were further consolidated, based on adjacent locations or partial overlaps, into 1476 CNVRs of which 301 CNVRs (20%) were shared between two or more studies (Table S10, Fig. 4). The majority of shared CNVRs involved genes associated with olfactory reception (50 CNVRs) and membrane transport (49 CNVRs) but also genes involved in transcription (30 CNVRs), cell cycle regulation (12 CNVRs) and RNA genes (34 CNVRs). Expectedly, CNVRs that were found in more than 100 horses and reported by all 6 studies exclusively involved olfactory receptors. Comparative analysis also revealed that novel (study-specific) CNVRs predominated over shared ones in all 6 studies (Fig. 4). Novel CNVRs of functional interest from this study involved genes related to sperm-egg interaction and fertilization in chr4:19.8–19.9 Mb; a developmental gene SOX2 in chr19:20.1 Mb; an X-linked region harboring genes of circadian pacemaker function chrX:83.8–84.0 Mb, and a complex CNVR in chrUn:225–226 kb with cancer related genes. Notably, the latter two CNVRs were found in more than 10 horses each. Details of all novel and shared CNVRs are presented in Table S10.
Nineteen CNVRs were validated by quantitative PCR (qPCR) using array probe-specific primers (Table S2). The regions were selected upon three criteria – size, gene content and novelty. The smallest tested CNVR was 4 kb and the largest 2 Mb; 13 involved clusters of horse genes, and 6 were novel. A summary of qPCR results are presented in Figure S3 and Table S11. All selected CNVRs were first tested in the discovery horses and then analyzed in more individuals of the same breed to identify possible breed-specific tendencies. Overall, qPCR observations agreed well (P-value <0.05) with the array CGH data for all discovery horses and for other animals of the same breed. For example, it confirmed a complex CNVR in chr27 involving CSMD1 gene (CUB and Sushi multiple domains 1) which encodes a transmembrane and a candidate tumor suppressor protein [43]. Copy numbers in this region were tested on 11 breeds with at least 2 individuals each and showed a gain in native ponies, draft breeds and the Przewalski horse, and a loss in American Miniature horses in relation to the Thoroughbred (Fig. 5A–B). Likewise, qPCR confirmed a CNVR in chr20 (Fig. 5C) which has been found only in this study and in indigenous plateau horses [38]. However, we found some differences too between the two data sets: e.g., while qPCR confirmed a loss in chr20:32.0–32.4 Mb and chr17:18.8–19.0 Mb in the discovery Swiss Warmblood and Mongolian horses (Table S3), respectively, inclusion of additional horses from the same breeds resulted in a significant gain in these regions (Fig. S3). Also, initial qPCR confirmed a loss in chr7:74.8–74.9 Mb in the two discovery Swiss Warmblood horses (Table S3) but no significant losses were found when more individuals were added. These minor discrepancies can be attributed to intra-breed variation: array CGH was based on 2 to 4 individuals, while qPCR involved 4 or more horses per breed (Figure S3, Table S11).
Two CNVRs, a complex 200 kb gain-loss region in chr1:114.0–114.2 Mb and a 2.2 kb gain in chrUn: 529–531 kb) were validated by FISH using CNV-containing CHORI-241 BAC clones 132B13 (Fig. S4) and 91B23 (Fig. 6), respectively. Clear differences in copy numbers between individual horses, as well as between homologous chromosomes of the same horse were observed. Additionally, the CNVR in chrUn was mapped to horse chr19q12–q13 (Fig. 6).
Finally, we carried out a pilot study to test the utility of the tiling array and the integrated CNV data set (Table S10) for the discovery of CNVs involved in equine XY disorders of sexual development (XY DSD). Selection of the phenotype was based upon studies in humans suggesting contribution of CNVs to XY DSDs [25], [27], [28]. Array CGH experiments were carried out in 6 affected horses (Table 4): all had normal male 64,XY karyotype with an intact SRY gene, abnormal male or female gonads, and female or female-like external phenotype [44]. We determined 179 CNVs (average 30 calls per individual) and 107 CNVRs, of which 83 were common and shared with normal equine populations, and 24 CNVRs were novel (Table 5). Only 3 novel CNVRs were shared between two or three XY DSD horses, while the remaining 21 were private and present in just one animal. Protein coding or miRNA genes with functions in cell cycle regulation, transcription and posttranscriptional processing were involved in 14 novel CNVRs. None of the CNV-genes had known functions in sexual differentiation or development.
Analysis of common CNVRs for highly aberrant log2 values detected two likely homozygous deletions (Table 5): a 26 kb loss in chr7 (log2 −2.2) and a ∼200 kb loss in chr29 (log2 −3.5). The latter was of particular interest because it was found in two closely related American Standardbreds with very similar male-pseudohermaphrodite phenotypes (H348 and H369; Table 4). The CNVR was also present in 10 out of the 38 normal horses (Table S3) including one American Standardbred, though with a moderate aberration value (log2average −0.7) compared to log2 = −3.5 in the two XY DSD horses. Most notably, the CNVR involved at least 4 members of the aldo-keto reductase AKR1C gene family, known to be critical in the backdoor pathway of dihydrotestosterone (DHT) synthesis and sexual development [45], [46]. A schematic overview of the CNVR, including the involved genes and aberration profiles of all 47 array probes in the region, is presented in Fig. 7. Homozygosity of the deletion was confirmed by fluorescence in situ hybridization (FISH) with a BAC clone (CHORI-241-23N13) spanning the deletion. The BAC hybridized to chr29 in control animals but not in the two XY DSD horses, whereas a control BAC (CHORI-241-76H613) with the CREM gene from a non-CNVR in chr29 [47] hybridized equally in the XY DSD horses and controls (Fig. 7). Homozygosity of the deletion was further confirmed by PCR showing that primers designed inside the CNVR amplified genomic DNA of control horses and the remaining 4 XY DSD horses, but not of the two male-pseudohermaphrodite American Standardbreds (H348 and H369; Fig. 7). Though primers designed outside the CNVR, amplified the DNA by PCR in all horses – an evidence that the DNA quality of the two Standardbreds was acceptable. We theorized that the homozygous deletion involving AKR1C genes in the two male-pseudohermaphrodite horses might be the risk factor for abnormal sexual development.
During just the past two years, five studies have addressed the phenomenon of copy number variation in the horse genome [35], [36], [37], [38], [39] contributing to our knowledge about the genomic landscape of CNVs and their role in inter-individual variation in horses. Despite the progress, lessons from humans [48], [49], [50] and more recently from dogs [7], [34], show that efficient biomedical application of this information requires integration of data from many more populations and individuals and the use of comparable methodological platforms[48], [49], [50].
Here we report about the construction of a 400K high-density WG tiling oligoarray for the horse and its application for the discovery of CNVs in 38 normal horses of 16 diverse breeds, as well as in 6 horses with congenital disorders. Probes on the array were designed to detect CNVs in 18,763 equine autosomal and X-linked genes but also in intergenic, sub-telomeric and Y chromosome sequences. Regarding genome coverage, our CNV discovery platform most closely resembled the recently reported WG 1.3 M NimbleGen CGH array [38], but essentially complemented the exon CGH array by Doan and colleagues [39] and the studies based on WG SNP50 BeadChip [37], [51]. The latter is of a magnitude lower density and not specifically designed for CNV capture. Also, as shown in humans and cattle, the efficiency of CNV discovery is lower in SNP platforms compared to CNV focused arrays [29], [50]. While the future direction for CNV research in any species is probably next generation sequencing (NGS), the approach has as yet found only limited application in horses: for the discovery of CNVs in the genome of a Quarter Horse mare [35] and for the discovery of segmental duplications in 6 horse breeds and the donkey [52].
A unique feature of our CGH array was the inclusion of probes from the Y chromosome and sub-telomeric regions. This was because CNVs and segmental duplications are known to be an integral part of the architecture of the mammalian Y chromosome [53], [54], while sub-telomeres are hotspots of DNA breakage and repair, and undergo structural rearrangements more frequently than the rest of the genome [40], [55]. Despite the almost 6,000 sub-telomeric probes with lower than average spacing (∼4 kb vs. ∼7 kb across the genome) on the array, only 10 CNVs were detected in sub-telomeres and none in the Y chromosome (Table 2). It is likely that the complex sub-telomeric sequences are either missing or underrepresented in the current horse sequence assembly [56], due to which it is possible that the probes designed from the ends of the chromosomes, did not originate from actual sub-telomeres.
Poor representation of centromeric/pericentromeric and telomeric/sub-telomeric sequences is a common shortcoming of all draft genome assemblies. Whilst the horse may be different to humans or other species in terms of subtelomeric sequences, this can only be rigorously shown by sequencing BAC clones from these regions, preferably with long-read single molecule technology such as a Pacific Biosciences instrument to resolve long repeats. Such an approach was recently successfully applied to resolve regions of segmental duplications in the finished genome sequence of humans [57].
The Y chromosome, on the other hand, has acquired and amplified novel sequences, as well as sequences from the rest of the genome [58]. Thus, it is likely that many potential copy number variable Y probes did not pass the ‘uniqueness’ test by BLAST and were dropped from the array (see Material and Methods for details).
The present and all previous CNV studies in horses [35], [36], [37], [38], [39] differ by discovery platforms, genome coverage, resolution, the study cohorts and analytical methods (Table 6). Therefore, the overall numbers, size ranges and chromosomal distribution of CNVs vary between the studies. For example, it has been shown that due to analytical reasons, CGH-based studies tend to detect more losses than gains [59]. This holds true for the Agilent WG array in the present study and also the Nimblegen WG array [38], though [38]slightly more gains were detected with the Agilent exon array [39] (Table 6). The latter was attributed to the large number of losses in the reference animal compared to the Thoroughbred (Twilight) genome sequence assembly EquCab2 [56]. In contrast, gains vastly predominate (97%) among the CNVs found by NGS in a Quarter Horse mare [35]. Apparent differences in CNV calling algorithms and thresholds (Table 6), on the other hand, are responsible for the variation in the number of CNVs, their size and the criteria for merging individual CNVs into CNVRs. For example, in this study we mainly reported CNVRs because this is how the ADM-2 algorithm analyses and assembles the CNV calls (CNVs) within and across individuals. Further, specific features of the probe/array design, and not necessarily the number of probes, are responsible for the differences in the genomic distribution of discovered CNVs. So far, X-linked CNVs have been found only in this study and by Doan & colleagues [39], and CNVs in chrUn only in this study. Surprisingly, the study with a three times denser 1.3 M Nimblegen array failed to detect CNVs in chrX, as well as in [38] chrs30 and 31 [38]. At the same time, the latter two small autosomes show the highest number of CNVs in the Quarter Horse mare [35]. Major differences are also in the size, diversity and origin of the study cohorts, ranging from just a few breeds and individuals [35], [38] to over 15 breeds (this study and [37]) and hundreds of individuals [36], [37] (Table 6).
The many variables between the six studies (Table 6) obviously confound assessments based solely on CNV metrics, and it would probably be more appropriate to compare the actual CNVs/CNVRs reported. Therefore, and in order to obtain a comprehensive overview about the status of CNV discovery in horses, we integrated the CNVs or CNVRs from all six studies ([35], [36], [37], [38], [39], this study) according to their genomic locations into a composite dataset of 1476 CNVRs (Table S10). Of these, 301 are reported by at least two studies, while the remaining 1174 CNVRs are study-specific (novel; Fig. 4) and require further validation.
The integrated dataset is a needed resource for evaluating new CNV discoveries and gives an idea about the most intrinsic features of the CNV profile in horses. Copy number variants account for about 1 to 3 % of the horse genome and there are more CNVs that involve genes than those located in intergenic regions. Though, the number of intergenic CNVs is possibly deflated because all tiling arrays [38], [39], including ours, have been biased towards probes for gene exons. For example, 20% of the probes in the Texas-Adelaide WG array represent protein coding genes, whereas these genes make up only about 2–3% of the mammalian genome. Notably, all studies find chr12 as the most CNV-enriched (Table 6) and not because of many CNVs, but because of a few very large clusters of olfactory receptors and immunity-related genes (Tables S8, S10).
Studies in human [3], [60], dogs [8] and cattle [30] have noted strong correlation between CNVs and segmental duplications (SDs). This is because SDs share 90% sequence similarity with another genomic location and can promote CNV formation by non-allelic homologous recombination [61]. Similar tendency has been observed in horses [39], although horse SDs are relatively small (largest ∼60 kb) and comprise only about 0.5–0.6 % of the genome [56], thus less than the portion involved in CNVs (Table 6). Low level of SDs or low copy number repeats was also reported by a recent de novo analysis of the equine genome where no novel classes or types of interspersed repeats were identified [62]. An additional 0.4% of SDs are in unplaced contigs (chrUn) [56], though in this study only 0.04 % of chrUn sequences had CNVs (Table 2). Likewise, chr25 which is the most SD-rich chromosome (1.7%) according to EquCab2 genome assembly [56], was only moderately enriched with CNVs (0.35%) in this study. Yet, findings by us and others support the correlation between CNVs and SDs in some genomic regions. For example, a known large (750 kb) segmental duplication at the boundary of ELA class I and class III [63] falls into a large common CNVR in chr20:30,127,886–31,231,182 (Table S10); further, low copy number directional repeats have been associated with large deletions in the horse Y chromosome [44] or, GO categories, such as olfactory reception and immune response, prevail among the genes involved both in CNVs and SDs [52]. Therefore, for improved understanding of the genomic architecture of CNVs and their relation to genes and phenotypes in horses, it would be worthwhile to focus future CNV research on associations between CNVs and SDs, as recently successfully done in dogs [8].
It is noteworthy that regardless of the discovery methodology and study cohorts, functional groups of genes that are most affected by CNVs remain the same in all studies. These include genes for transmembrane signal transduction and chemo-attractant sensory perception (olfactory and non-olfactory G-protein coupled receptors, GPCRs), immune response (immunoglobulins, T-cell receptors, MHC protein complexes), and steroid metabolism (Table S9). Not coincidentally, CNVs are associated with the same groups of genes in humans [3], [64], cattle/ruminants [30], [65], [66], pigs [31], dogs [32] and even chicken [67], suggesting the importance of inter-individual variation in these genes for adaptive plasticity [68]. Indeed, genetic diversity and fine functional tuning of sensory receptors, immunoglobulins, natural killer and Toll-like receptors is further enhanced by additional mechanisms, such as asynchronous replication which increases the rate of tandem duplications, and monoallelic expression, so that each sensory neuron or lymphoid cell expresses only one allele of a gene [69], [70]. Conserved linkage between distinct olfactory receptor genes and the MHC in several mammalian species suggests their concerted function - in this case, MHC-influenced mate choice in reproduction [71]. Olfactory receptors are also thought to function as chemo-sensing receptors to regulate sperm density, motility, acrosome reaction and sperm-egg interaction in fertilization [71], [72]. Thus functionally, the CNV-enriched genes in horses and other mammals fall into just three large categories: sensory perception, immunity and reproduction.
Among the 258 CNVRs detected in this study, 20% were located in intergenic regions. These CNVRs were relatively small (average 50 kb, median 35 kb) and represented predominantly losses (Fig. 2, Table S8). Prevalence of losses among intergenic CNVRs has also been found in dogs [32]. Although there is no information about possible implication of these regions on the function of genes in animal genomes, studies in humans show that intergenic deletions are significantly enriched among gene expression-associated CNVs [73]. Thus, with the improvement of genome sequence assembly and annotation in horses, intergenic CNVRs will be of interest for future studies. We also anticipate that as gene models are revised and converge more with the underlying reality of the genes, some intergenic CNVRs may become genic and vice versa.
One of the goals of CNV research in horses is to find variants that distinguish between breeds or groups of breeds and could be associated with specific adaptations and phenotypic traits of interest. In order to visualize the breeds and the degree of diversity represented in this and previous studies, we performed a phylogenetic analysis using population data of 15 microsatellite loci [74] for the breeds involved (E.G. Cothran, unpublished). The dendrogram in Figure 8 shows that while the major clades of domestic horses are represented, there is a clear preponderance of the breeds with Thoroughbred ancestry. It is therefore noteworthy that data for 11 new breeds, mainly representing native ponies and draft horses, were added in this study. Nevertheless, the current tally of horse breeds studied for CNVs is 41 (Table S12) which is less than 10% of the over 400 horse breeds known worldwide [75]. Furthermore, given that just 7 breeds have been involved in 2 or more studies (Fig. 8, Table S12) and several breeds are represented by one individual [38], [39], any CNV reported to be breed-specific should be taken with caution. For example, our composite CNV dataset (Table S10) shows that the 18 CNVs reported to be specific for Hanoverians [37] are present in other breeds. Likewise, only one (chr13: 1,497,390.00–1,508,926.00; EIF2AK1) of the 7 plateau-breed-specific CNVs in heme binding genes [38] is not found in other breeds. The same happened with our data where initially we identified over 10 putative breed-specific CNVs which, after comparison, reduced to 2 - one in Exmoor pony, another in Swiss Warmblood horse (Table S4). Interestingly, no unique CNVs were found in the Przewalski horse which shared similarity mainly with ponies and draft breeds (Table S3). Besides, only 9 of the 25 CNVs in Przewalski horses were shared between the two individuals studied. Similar tendency for intra-breed individual variation was observed for domestic horses where private CNVs predominated over the shared ones. Nevertheless, as suggested by other studies in horses [39], cattle [29], pigs [31] and dogs [33],we anticipate that a small percentage of CNVs might remain unique to their respective breeds, though this requires analysis of much larger and more diverse equine populations. On the other hand, most horse breeds are of recent origin with a good deal of cross-breeding until closed breeds were established which has led to a high degree of haplotype sharing [56], [76], and thereby decreased chances for finding breed-specific CNVRs compared to species like dogs [34].
Probably the most exciting goal of CNV research in any species is the discovery of pathogenic variants responsible for complex diseases and congenital disorders. Among these, disorders of sexual development (DSDs) are not uncommon in horses, though causative mutations have been identified for just a few: Y chromosome deletions in SRY-negative XY sex reversal mares [44] and a point mutation in the androgen receptor gene in 3 related SRY-positive XY mares [77].
Here, we conducted the first pilot CNV analysis in horses with XY DSD and identified a large autosomal (chr29) deletion in 2 related American Standardbreds (H348 and H369, Table 4). The animals were classified as male pseudo-hermaphrodites with XY male genotype, immature testes-like abdominal gonads, and female-like external phenotype (Table 4). The deletion in chr29:28.6–28.8 Mb was homozygous as confirmed by FISH and PCR, and involved at least 8 genes of which 4 belonged to the aldo-keto reductase family 1, member C (AKR1C; Fig. 7). Annotation of these genes in the equine genome is, as yet, preliminary and based on the alignment with human AKR1C proteins in the UCSC Genome Browser (http://genome.ucsc.edu/index.html) and mammalian homology in Ensembl (http://www.ensembl.org/index.html). Therefore in Fig. 7, three genes are denoted as AKR1CL1 and one gene has three labels, corresponding to AKR1C2 in chimpanzee, AKR1C3 in human, and AKR1C4 in cattle.
The AKR1C genes are members of the aldo-keto reductases (AKR) superfamily [78]and encode for 3α-hydroxysteroid dehydrogenases [78] which are critically involved in steroid hormone metabolism [79]. In the human genome, there are 4 family members - AKR1C1, ALR1C2, AKR1C3 and AKR1C4, which share 86% sequence identity and are clustered in HSA10p15-p14 [78], [79]. The human AKR1C genes are not widely expressed: AKR1C1 in brain, kidney, liver and testis, AKR1C2 in prostate and brain, AKR1C3 in prostate and mammary gland, and AKR1C4 in liver, whereas the rat has a single AKR1C gene expressed in liver[79], [80], [81]. Among other functions, the AKR1C genes are involved in the biochemical pathway that leads to dihydrotestosterone (DHT) synthesis without testosterone intermediate. As opposed to ‘classical’ DHT synthesis from cholesterol and testosterone, this pathway is known as ‘the backdoor pathway’ and was originally discovered in marsupials [82] and thereafter in eutherian mammals [45], [46], [83], [84]. The importance of the ‘backdoor pathway’ and AKR1C genes in male sexual development was recently demonstrated by a study in humans showing that mutations in AKR1C2 and AKR1C4 genes cause abnormal virilization and disordered sexual development, including XY sex reversal [46], [84]. Even though no mouse knockout models are available for any of the AKR1C genes (MGI; http://www.informatics.jax.org/), it is tempting to speculate that the homozygous deletion in horse chr29 is a causative or a risk factor for some forms of equine XY DSDs, such as male-pseudohermaphroditism, as observed in this study. It is also worth mentioning that a CNV analysis of human XY DSDs detected a clinically significant de novo 64 kb duplication in HSA10p14 [28] - a genomic segment next to the AKR1C gene cluster (UCSC: http://genome.ucsc.edu/cgi-bin/hgGateway). Whether this is a coincidence or the region includes more copy number variable factors contributing to DSDs, needs further investigation. [45], [46], [84] [84] [28].
Our findings in horses might be of even broader interest because the two deletion carrying horses were elite American Standardbred pacers, Martha Maxine and Arizona Helen (Table 4), whose problematic sexual identity has become public, making headlines in The New York Times [85] and The Horse [86]. Thus, studies are underway to precisely determine the deletion breakpoints and develop molecular tests for detecting other horses with a similar deletion, as well as heterozygous carriers. Finally, the fact that only 2 XY DSD horses out of 6 had this mutation underscores the phenotypic and genetic heterogeneity of these disorders.
This study represents an important contribution to CNV research in horses by identifying new CNVs and developing an integrated datset of 1476 CNVRs to facilitate the discovery of variants of biomedical importance. However, despite progress, the majority of the CNVRs reported for the horse require proper validation by methodologically comparable studies invloving more diverse breeds and individual animals. Last but not least, due to the very nature of CNVs, these regions are likely to have sequence assemblies not as accurate as non-variable regions. Thus, the findings also identified potential targets for genome re-sequencing and -assembly.
Procurement of peripheral blood and hair was performed according to the United States Government Principles for the Utilization and Care of Vertebrate Animals Used in Testing, Research and Training. These protocols were approved by Texas A&M Office of Research Compliance and Biosafety as AUP2009-115, AUP2012-0250. CRRC09-32 and CRRC09-47.
A horse WG tiling array was designed using the horse genome draft sequence (EquCab2, http://www.ncbi.nlm.nih.gov/assembly/286598; [56], Oligowiz2.0 (http://www.cbs. dtu.dk/services/OligoWiz/), ArrayOligoSelector (http://arrayoligosel.sourceforge.net/), and ArrayDesign [87] software packages. The array comprised 417,377 60-mer oligonucleotide probes: 85,852 probes corresponded to one or more exons of the 18,763 annotated equine genes (http://www.ncbi.nlm.nih.gov/genome/genomes/145?); 305,416 probes originated from intergenic regions (excluding sub-telomeres); 5,716 probes were designed from sub-telomeres (the terminal 1 Mb of each chromosome), and 519 probes represented the horse Y chromosome [58]; our unpublished data). [87]For intergenic probes, including chrUn, repeat-masked (http://www.repeatmasker.org/) sequences were used. For reference genes, we first designed probes from exons. If these were not specific, attempts were made to design probes from introns and upstream/downstream flanking regions of those genes. Before inclusion in the array, the specificity of all sequences were analysed with BLAT (http://www.kentinformatics.com/) and BLAST (http://blast.ncbi.nlm.nih.gov/Blast.cgi) against the EquCab2 reference genome sequence. Probes with more than one hit in the genome were discarded. Possible cross-hybridization of the probes was further evaluated using Kane's parameters [88] and all probes that had a total percent identity >75–80% with a non-target sequence, or probes with contiguous stretches of identity >15 nucleotides with a non-target sequence were discarded. Only probes with high specificity were kept in the final array. A Cytoband file was generated to align the horse draft sequence assembly with the cytogenetic map [89]. The array, designated as the Texas-Adelaide horse WG tiling array, was fabricated by Agilent Technologies using Agilent SurePrint G3 technology and 2×400K chip format (two arrays on a single slide). The array is available at Agilent Technologies; Design ID #030025, Cat. No G4124A.
The CNV discovery cohort comprised 38 horses representing 16 diverse breeds and the Przewalski's horse (Table S1). Horse breeds were selected according to the recent population studies [51], [56], [76], [90] with an aim to maximize the genetic diversity among samples and to encompass the common warm blood, cold blood (draft) and native pony breeds. An additional cohort of 52 normal horses representing the same 16 breeds was used for quantitative PCR validation of CNVs. Finally, a pilot study testing the utility of the tiling array for the discovery of CNVs contributing to equine congenital disorders used 6 horses previously diagnosed with XY disorders of sexual development (XY DSDs; Table 4) [44].
Genomic DNA was isolated from peripheral blood or hair follicles using QIAGEN Gentra PureGene Blood kit (Qiagen) according to manufacturer's protocol. The DNA was cleaned with DNeasy Blood and Tissue kit (Qiagen) and quality checked by gel electrophoresis and by Nanodrop spectrophotometry (Thermo Scientific).
Probe labeling and array CGH experiments were performed according to Agilent Technologies Protocol Version 7.3, March 2014 (http://www.chem.agilent.com/Library/usermanuals/Public/G4410-90010_CGH_Enzymatic_7.3.pdf). All hybridizations comprised of a pair of differently labeled probes, one of which was always the reference DNA – a Thoroughbred mare Twilight for females and a Thoroughbred stallion Bravo for males (see explanations below). The genomic DNA (gDNA) was cleaved to 200–500 bp fragments with RsaI and AluI (Promega) and labeled with Cy3 (the reference DNA) or Cy5 (sample DNA) by random priming using Genomic DNA Enzymatic Labeling Kit (Agilent Technologies). The products were cleaned with 30 kDa filters (Amicon) and the yield and specific activity of labeled DNA was determined with a Nanodrop spectrophotometer. Typical yield for 1 µg of starting DNA was 6–8 µg; specific activity for Cy3 was 25–40 pmol/µg and for Cy5 20–35 pmol/µg. The hybridization mixture was prepared using Agilent Oligo aCGH Hybridization Kit and contained equal quantity of Cy3 and Cy5 labeled probes, 1 µg/µL horse Cot1 DNA, 10× blocking agent, and 2× Hi-RPM buffer. Denatured and pre-annealed probe mixture was applied onto gasket slide, placed in Agilent SureHyb hybridization chamber, ‘sandwiched’ with an array slide and incubated in Agilent hybridization oven at 65°C for 40 hours. The array slides were washed with Agilent aCGH Wash Buffers 1 and 2 and dried with Acetonitrile and Stabilization and Drying Solutions (Agilent Technologies).
The slides were scanned with Agilent SureScan DNA Microarray Scanner and Scanner Control software v8.3. The data were extracted and normalized with Agilent Feature Extraction software v10.10.1.1 and saved in.fep format. The Feature Extraction software also checks the quality of aCGH by measuring Derivative Log2 Ratio Standard Deviation (DLRSD), Signal-To-Noise Ratio (SNR) and Background Noise (BGNoise). The data were analyzed with Agilent Genomic Workbench 5.0 software. In each array spot log2 ratios of Cy3 versus Cy5 were computed with the default P-value threshold 0.05 and overlap threshold value 0.9. The CNVs were represented by gains and losses of normalized fluorescence intensities relative to the reference and called by conservative criteria which required alterations of >0.5 log2 ratios over 5 neighboring probes. Homozygous losses were called when signal log2 ratio was <−2.0. Copy number variable regions (CNVRs) were determined by ADM-2 algorithm [91] by combining overlapping and adjacent CNVs in all samples across the CGH experiments. Output files were generated with genomic coordinates and cytoband locations for all CNVs. The raw data were submitted to NCBI Gene Expression Omnibus (GEO) accession GSE55266.
To evaluate baseline variations and determine FDR [92], [93] female and male self-to-self, and female-to-male control hybridizations were conducted using blood DNA from one female and one male Thoroughbred horses. The female Thoroughbred, Twilight, was the DNA donor for the horse reference sequence EquCab2 [56] and the origin of the probes on the tiling array. The male Thoroughbred, Bravo, a half-sibling to Twilight, was the DNA donor for the CHORI-241 BAC library (http://bacpac.chori.org/equine241.htm) and the origin of all Y chromosome probes on the array. The FDR was calculated as a percentage of the ratio of CNVs in self-to-self hybridization to the total number of CNVs in all experiments. Additionally, array performance was evaluated by self-to-self hybridizations with blood and hair DNA from one Quarter Horse (H528, Table S1). Hybridization quality was assessed by DLRSD which calculates probe-to probe log ratio noise of an array; (http://www.chem.agilent.com/Library/applications/5989-6624EN.pdf): DLRSD <0.2 was considered excellent; 0.2≥DLRSD≤0.3 was good, and values >0.3 indicated poor quality hybridization.
Horse chromosome enrichment percentage was determined by the total length of CNVRs present in each chromosome, divided by chromosome length (Ensembl, http://www.ensembl.org/index.html).
Ensembl gene list (Ensembl Genebuild 73.2) along with their position in the horse genome was added to Agilent Genomic Workbench as a custom track to determine the genic and intergenic CNVs. Gene Ontology analysis (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of the genes present in CNVs were performed using DAVID bioinformatics tool with default settings [94], [95]. Because only a limited number of genes in the horse genome have been annotated, horse gene IDs were converted to orthologous human Ensembl gene IDs by BioMart, followed by GO and pathway analyses, as described above. Biological functions of the genes in CNVRs were further analyzed manually by data mining in Ensembl (http://www.ensembl.org/index.html), UCSC (http://genome.ucsc.edu/) and NCBI (http://www.ncbi.nlm.nih.gov/) Genome Browsers searching for data for equine orthologs in other mammalian species. CNVs present in intergenic regions were analyzed in UCSC genome browser and NCBI and GeneCards (http://www.genecards.org/) for similarities to known mammalian genes.
A composite CNV dataset for the horse (Table S10) was generated by aligning genomic positions of CNVs/CNVRs from this and all previously published studies [35], [36], [37], [38], [39]. Partially or completely overlapping and adjacent CNVs (the end position of a previous CNV and the start position of the next CNV are the same) were consolidated into one CNVR.
Genomic copy number changes as detected by aCGH were validated by quantitative PCR (qPCR) for 18 selected CNVRs using 22 probe-specific primers. Additionally, 8 putative homozygous deletions were validated by regular (qualitative) PCR. Primers (Table S2) were designed inside CNVRs using array probe sequences and the horse whole genome sequence information (EquCab2 at UCSC: http://genome.ucsc.edu/and Ensembl: http://www.ensembl.org/index.html) and Primer3 software (http://bioinfo.ut.ee/primer3-0.4.0/primer3/input.htm). The qPCR experiments were performed with LightCycler 480 (Roche Diagnostics) in triplicate assays. Each assay was done in triplicate 20 µL reactions containing 50 ng of template DNA, 10 µM primers and the SYBR Green PCR kit (Roche). Relative copy numbers of the selected regions were determined in comparison to the reference sample (Thoroughbred and Quarter Horse) and normalized to an autosomal reference gene GAPDH. The cycling conditions were 1 cycle 5 min at 95°C; 45 cycles 10 sec at 95°C, 5 sec at 58°C, and 10 sec at 72°C; 1 cycle for melting curve 30 sec 95°C, 30 sec 65°C and final cooling 20 sec at 50°C. Quantification of the copy number was carried out using the comparative CT method (2ΔΔCt) [96], [97] with p<0.05 as a cut-off threshold for statistical significance. Qualitative PCR results were analyzed by agarose gel electrophoresis.
CNV specific primers were used to screen CHORI-241 BAC library (http://bacpac.chori.org/equine241.htm) by PCR (Table S2); BAC DNA was isolated by Plasmid Midiprep kit (Qiagen), labeled with biotin-16-dUTP or digoxigenin-11-dUTP using Biotin- or DIG-Nick Translation Mix (Roche), and hybridized to metaphase chromosomes of CNV carriers and control horses following standard protocols [98]. A BAC clone representing a non-CNV region was used as a control in each FISH experiment. Images for a minimum of 20 metaphase and/or interphase cells were captured for each experiment and analyzed with a Zeiss Axioplan2 fluorescent microscope equipped with Isis v5.2 (MetaSystems GmbH) software.
Genotypes for 15 microsatellite loci [74]; E.G. Cothran, unpublished) were available for 32 out of 41 horse breeds involved in CNV studies (see Table S12). Majority-rule consensus of Restricted Maximum Likelihood (RML) trees were constructed and visualized as described elsewhere [74]. The Przewalski Horse population was used as an out-group.
|
10.1371/journal.pgen.1003456 | Comprehensive Assignment of Roles for Salmonella Typhimurium Genes in Intestinal Colonization of Food-Producing Animals | Chickens, pigs, and cattle are key reservoirs of Salmonella enterica, a foodborne pathogen of worldwide importance. Though a decade has elapsed since publication of the first Salmonella genome, thousands of genes remain of hypothetical or unknown function, and the basis of colonization of reservoir hosts is ill-defined. Moreover, previous surveys of the role of Salmonella genes in vivo have focused on systemic virulence in murine typhoid models, and the genetic basis of intestinal persistence and thus zoonotic transmission have received little study. We therefore screened pools of random insertion mutants of S. enterica serovar Typhimurium in chickens, pigs, and cattle by transposon-directed insertion-site sequencing (TraDIS). The identity and relative fitness in each host of 7,702 mutants was simultaneously assigned by massively parallel sequencing of transposon-flanking regions. Phenotypes were assigned to 2,715 different genes, providing a phenotype–genotype map of unprecedented resolution. The data are self-consistent in that multiple independent mutations in a given gene or pathway were observed to exert a similar fitness cost. Phenotypes were further validated by screening defined null mutants in chickens. Our data indicate that a core set of genes is required for infection of all three host species, and smaller sets of genes may mediate persistence in specific hosts. By assigning roles to thousands of Salmonella genes in key reservoir hosts, our data facilitate systems approaches to understand pathogenesis and the rational design of novel cross-protective vaccines and inhibitors. Moreover, by simultaneously assigning the genotype and phenotype of over 90% of mutants screened in complex pools, our data establish TraDIS as a powerful tool to apply rich functional annotation to microbial genomes with minimal animal use.
| Salmonella Typhimurium is a major cause of human diarrhoeal infections, usually acquired from chickens, pigs, cattle, or their products. To understand the basis of persistence and pathogenesis in these reservoir hosts, and to inform the design of novel vaccines and treatments, we generated a library of 7,702 S. Typhimurium mutants, each bearing an insertion at a random position in the genome. Using DNA sequencing, we identified the disrupted gene in each mutant and determined its relative abundance in a laboratory culture and after experimental infection of mice, chickens, pigs, and cattle. The method allowed large numbers of mutants to be investigated simultaneously, drastically reducing the number of animals required to perform a comprehensive screen. We identified mutants that grow in culture but do not survive in one or more of the animals. The genes disrupted in these mutants are inferred to be important for the infection process. Most of these genes were required in all three food-producing animals, but smaller subsets of genes may mediate persistence in a specific host species. The data provide the most comprehensive map of virulence-associated genes for any bacterial pathogen in natural hosts and are highly relevant for the design of control strategies.
| Salmonella enterica is a facultative intracellular pathogen of worldwide importance, associated with c. 21.7 million cases of systemic typhoid fever and 93.8 million cases of non-typhoidal gastroenteritis in humans each year [1], [2]. Around 86% of human cases of non-typhoidal salmonellosis are the result of food-borne infections [2], and chickens, pigs and cattle are key reservoirs of infection [3]. The major S. enterica subspecies enterica encompasses a wide variety of serovars. Some of these, such as S. Typhimurium and S. Enteritidis, exhibit a wide host range, whereas others such as S. Typhi are largely restricted to a single host species. The molecular basis of the host- and tissue-tropism of S. enterica has long eluded researchers and there has been a disproportionate emphasis on the basis of Salmonella persistence and pathogenesis in murine models of colitis and typhoid fever. Comparative analyses of whole genome sequences have associated host-restriction of S. enterica serovars with gene decay. Interpretation of the impact of variation in the repertoire, sequence or expression of S. enterica genes requires an understanding of the roles of those genes in relevant hosts. Random transposon insertion mutants have been screened individually in chickens [4], but high-throughput simultaneous analysis of mutant phenotypes was made possible by the advent of signature-tagged mutagenesis (STM) [5]. STM allows the survival of individual mutants within a pool to be assessed qualitatively through hybridization of a probe to a unique tag sequence within the transposon. Comparison of hybridization signals obtained from “input pools” of mutants grown in vitro with the those obtained from the same set of mutants screened for survival in a model of infection (“output pool”) allows attenuated mutants to be identified. The insertion sites can then be identified by subcloning and sequencing. Analysis of pools of signature-tagged S. Typhimurium mutants in mice [5] led to the discovery of Salmonella Pathogenicity Island (SPI)-2 [6], a gene cluster that encodes a type III secretion system (T3SS-2) that acts as a molecular syringe for the secretion of effector molecules and influences systemic virulence and intracellular survival [7]. A distinct T3SS, encoded by SPI-1, was already known to be essential for infection of mice via the oral route [8]. Subsequently, STM libraries constructed in a range of Salmonella serovars were examined for their ability to colonize multiple hosts [9]–[14]. Comparative analysis of a library of 1045 mutants in chickens, pigs and calves suggested that S. Typhimurium deploys both conserved- and host-specific virulence factors [10], [12]. Notably, SPI-1 and -2 were vital in intestinal colonization of calves and pigs and to a lesser extent chickens [10], [12], yet a Type I protein secretion system encoded by SPI-4 appeared to influence infection of calves, but not chickens [12] or pigs [10].
Although STM has provided valuable insights into Salmonella pathogenesis, the technique is limited by the number of unique tags available, and the time and effort required to construct the library and identify attenuating mutations. Moreover, only negatively-selected mutants tend to be investigated and subjective judgments are used to compare signal intensities relative to the input and to other screened mutants. Transposon-directed insertion-site sequencing (TraDIS) [15], one of a new generation of STM-like techniques, addresses some of these limitations. TraDIS exploits Illumina sequencing [16] to obtain the sequence of the genomic region flanking each transposon. The massively parallel nature of the sequencing permits comparison of the number of specific reads derived from the input pools and the output pools after animal infection, providing a numerical measure of the extent to which mutants were negatively- or positively-selected during colonization (see Figure 1). TraDIS-like sequencing methods have been used to identify the essential gene complement of S. Typhi [15] and Streptococcus pneumoniae [17], genes involved in virulence of Haemophilus influenzae [18] and S. pneumoniae [19] in mice, and genes required for survival of the symbiont Bacteroides thetaiotaomicron in the murine gut [20]. Here we apply TraDIS to simultaneously assign the genotype and relative fitness of 7702 distinct S. Typhimurium mutants during intestinal colonization of chickens, pigs and cattle, providing highly relevant data for the control of zoonotic and animal salmonellosis.
To validate the quantitative nature of TraDIS, we applied it to investigate pools of Mu and mini-Tn5 mutants of S. Typhimurium strain SL1344 before and after intravenous infection of BALB/c mice (see Text S1). These mutant pools had been characterized previously using transposon-mediated differential hybridization (TMDH) [21], a microarray-based method that relies on hybridization of run-off transcripts arising from transposon-encoded T7 and SP6 promoters to high-density oligonucleotide arrays. In total, using TraDIS, 9792 distinct transposon insertions (4992 Mu and 4800 Tn5) were unambiguously mapped at the level of the single nucleotide to the SL1344 genome, providing relative fitness scores for 94.4% of the 10368 mutants screened, This is likely to be an underestimate of the performance of TraDIS, since it is likely that there were siblings of mutants with mapped insertions within the pool of 10368 mutants. The fitness scores assigned by TraDIS are defined as the log2-fold change in the number of sequence reads obtained across the boundaries of each transposon insertion between the input and output pools (see Materials and Methods), and were significantly correlated with the existing TMDH data (Figure S1; P<2.2×10−16), verifying the quantitative nature of TraDIS. TraDIS allowed identification of a number of mutants missed by TMDH (see Text S1), and provided finer mapping of insertion sites than can be achieved by hybridization of transposon-flanking sequences to tiling arrays, extending the conclusions of the earlier study and demonstrating the superiority of the TraDIS approach.
Though useful, previous attempts to assign comprehensively the role of S. Typhimurium genes relied on parenteral infection of atypically susceptible mice [21]–[23], and do not reflect the roles of Salmonella genes during intestinal colonization of food-producing animals infected via the natural oral route. To identify genes relevant to colonization of animal reservoirs and therefore zoonosis, we generated a library of 8550 mini-Tn5 mutants of S. Typhimurium strain ST4/74 and applied TraDIS to assess the survival of these mutants during oral infection of chickens, pigs and calves. Pools of 475 mutants were screened in individual pigs and calves and pools of 95 mutants were screened in duplicate chickens. Pilot studies in which selected pools had been repeatedly screened in each species indicated the reliable negative selection of the same mutants at these pool complexities in the absence of stochastic loss that may be due to population bottlenecks (data not shown). TraDIS mapped 7702 distinct insertions to the nucleotide level, representing at least 90.1% of the mutants screened, and demonstrated the random distribution of the transposon insertions around the genome (see Figure 2), which disrupt 2715 different genes. TraDIS analysis revealed evidence of random drop-out of mutants from the pools in one chicken, two pig and two calf experiments, likely owing to recovery of output pools of an inadequate size, and data from these animals were omitted from subsequent analysis (see Text S1).
TraDIS assignments of the insertion sites and fitness scores of mutants are listed in Table S1 (BALB/c mice dosed intravenously) and Table S2 (chickens, pigs and calves dosed orally). The raw TraDIS sequence data are available from the NCBI Short Read Archive (accession numbers ERA000172 and ERP000286). To facilitate exploration of the TraDIS data, a user-friendly online genome browser was constructed with which the insertion site and fitness score can be viewed in the context of the linear genome, GC content, transcription start sites and existing annotation (http://www-tradis.vet.cam.ac.uk). Figure 3 shows the fitness scores obtained by TraDIS analysis of S. Typhimurium mutants screened in mice, chickens, pigs and calves, plotted against read coverage (equivalent to the “M/A” plots commonly used to display microarray data). The proportion of significantly attenuated mutants identified during intestinal colonization of food-producing animals was greater than in the murine typhoid model. In each host, a large proportion of mutations did not exert a strong negative or positive effect, indicating that a high number of accessory or redundant functions exist. P values were estimated using the available biological replicates (all pools were screened in duplicate in chickens, 2 pools were screened in duplicate calves and 3 pools were screened in triplicate pigs), and attenuated mutants were defined as those with a negative fitness score and P≤0.05.
To summarize the dataset further, genes were scored as potentially important in colonization if they were disrupted in at least one significantly attenuated mutant in any of the four host species. This enables comparisons between datasets derived from different transposon libraries (such as the mouse and chicken/pig/cattle datasets), and visualization of the data in comparison with other genome-wide datasets. For some genes, insertions at different subgenic locations can have divergent effects on the encoded protein resulting in contrasting fitness scores, so it is important to consider the genetic context of each individual transposon when interpreting the TraDIS data in detail. This is true for all transposon-based mutant screens, although earlier technologies such as STM and TMDH lacked sufficient resolution to permit such considerations. We recommend the use of the TraDIS browser (http://www-tradis.vet.cam.ac.uk) to assist interpretation of our data in the context of the genome annotation.
Fitness scores were obtained for 3194 distinct genes disrupted by transposon insertions in the mouse screen, and 2715 genes in the chicken, pig and calf screens, of which fitness score existed for 2435 genes in all three food-producing animals. Fitness scores were available from all four hosts for 1935 genes, of which 1069 had a significantly attenuated mutant in at least one host. Venn diagrams showing the numbers of significantly attenuated mutants, and the numbers of genes potentially important for colonization in chickens, pigs and cattle are shown in Figure 4. A further Venn diagram, combining the chicken/pig/cattle and mouse datasets, is available in Figure S2, and Figure S3 shows a comparison between the chicken, pig and cattle TraDIS data, and the data obtained from equivalent mutants in the earlier STM screens [10], [12]. A table of all genes disrupted by a transposon insertion, indicating if any of the mutants in each gene was significantly attenuated, is available in Table S3. To facilitate exploration of the TraDIS data, custom files have been prepared that allow proteins in KEGG metabolic pathway diagrams [24] to be coloured blue if an attenuated mutant was found in the encoding gene, or red if the gene was mutated but no significant attenuation was observed. The KEGG colour files are available from http://www-tradis.vet.cam.ac.uk. As an example, Figure S4 shows the effect of mutations affecting multiple steps in chorismate biosynthesis, which is known to influence persistence of S. Typhimurium in vivo. It is evident that mutations affecting sequential steps in the pathway are attenuating, with just two exceptions: the initial condensation of D-erythrose-4-phosphate and phosphoenolpyruvate into 3-deoxy-D-arabino-heptulosonate-7-phosphate, and the conversion of shikimate to shikimate-3-phosphate, both of which can be catalyzed by the products of multiple genes (aroFGH and aroKL, respectively). Thus, TraDIS identifies pathways in which defects exert a common effect, but also reveals steps at which functional redundancy exists. During the analysis of the TraDIS data it became clear that many regions of the genome with low GC content are important for intestinal colonization of chickens, pigs and cattle (see Text S1 and Figure S5).
Twelve genes were selected for further investigation based on the TraDIS data: carB, clpB, ilvC, mig-14, pagN, SL1344_0084 (STM0084), SL1344_4248 (STM4312), SL1344_3128 (STM3154), trxA, virK, ytfL and zirT (SL1344_1599). These targets were chosen based on their fitness scores (at least one mutant in each gene shows significant attenuation), and to include some genes with established roles in colonization in chickens (clpB [4]) or mice (trxA [25], mig-14 [26], virK [27]), genes with a postulated role in colonization (pagN [28], SL1344_0084 [12]), the mouse anti-virulence factor zirT [29], genes demonstrating variable fitness scores (SL1344_0084, SL1344_3128 and ytfL) and genes that demonstrate putative host-specific effects on colonization in the TraDIS data (clpB, ilvC and ytfL). Each gene was inactivated separately by λRed recombinase-mediated integration of linear PCR amplicons by homologous recombination [30]. Mutant phenotypes were evaluated in groups of 3 chickens per mutant per time interval. For each mutant, competitive indices (CIs) were derived 4, 6 and 10 days post-inoculation of age-matched chickens with the kanR-tagged mutant and ST4/74 nalR wild-type strain in a 1∶1 ratio (see Table 1).
With a single exception (SL1344_3128) all mutants were negatively-selected relative to the parent strain at day 4 post-inoculation, which corresponds to the time at which mutants were recovered for the TraDIS analysis. The difference in the mutant∶wild-type ratio was significantly different from the ratio in the inocula for 8 of the mutants at day 4. For a further 2 mutants, significant differences were detected at later time points. Taken together with comparisons to existing datasets for signature-tagged mutants of the same strain in the same animal models [10], [12] (see Text S1), the data strongly support evidence of attenuation detected by TraDIS. Variance from TraDIS fitness scores is likely to reflect differences in competition dynamics for a given mutant relative to co-screened wild-type or mutant bacteria. For one gene (SL1344_3128 ), no evidence was found of any attenuation of the defined mutant, which performed comparably to the wild-type at all time intervals. This gene was chosen for further investigation because its mutants exhibited a wide range of TraDIS fitness scores (−1.02 to −9.20). Interestingly, mutants in the gene cluster SL1344_3128-30 are predicted to be deficient in swarming motility [31], suggesting the possibility that such motility may be an occasional but not universal requirement for colonization.
The TraDIS dataset is a powerful resource for understanding intestinal colonization of a range of highly relevant hosts by Salmonella, and thus zoonotic transmission and animal disease. The data suggest that the definition of what constitutes a colonization gene is not straightforward, encompassing genes involved in metabolism, stress responses and transcriptional regulation, together with genes with well-established roles in virulence. The T3SSs encoded by SPI-1 and SPI-2 are both essential for infection in chickens, pigs and cattle, although there are some mutants within both regions that are not attenuated or exhibit a less pronounced phenotype in chickens. T3SSs allow the secretion of effector molecules into the host cytoplasm, these effectors being encoded both within SPI-1 and 2 and distally. Most of the known effector genes, including sopA, sopB, sopE2, sipA, avrA, sipC, sseG, sseI, sifA, sseK1, pipB2 and sopD2, were identified by TraDIS as being important for infection of all three food-producing animals, although as with the T3SS structural genes, the phenotype was often less pronounced in chickens. Other T3SS effectors, including sptP, slrP, gogB and sspH2, could be disrupted without affecting colonization. Of these, slrP has been implicated as a host-specificity factor, essential for oral infection of mice but not required for calf infection [14]. Null mutants of sptP are not impaired in their interactions with cultured macrophages or epithelial cells [32]–[34], and sspH2 mutants do not show any defect in vacuole-associated actin polymerization [35]. For both sptP and sspH2, the lack of phenotype was suggested to be due to functional redundancy amongst T3SS effectors.
There are also attenuated mutants that harbour insertions within the other recognized Salmonella pathogenicity islands [36]. In SPI-3, mutants in some genes (mgtC, marT, SL1344_3717 and SL1344_3721) were attenuated, with others (misL, sugR, slsA and mgtB) showing no attenuation. Interestingly, marT, which encodes a transcriptional regulator, is a pseudogene in S. Typhi, and restoring it reduces survival during infection of a human cell culture [37]. A role for marT in infection of chickens, pigs and cattle suggests a selection pressure for its retention in the S. Typhimurium genome. SPI-4 was previously thought to play a role in infection of cattle but not chickens or pigs, based on STM screens [10], [12]. TraDIS suggests a role for SPI-4 in all three species, although the phenotypes in chickens and pigs are more subtle than in calves, highlighting the increased sensitivity of TraDIS relative to STM. Some transposon insertions within the central highly repetitive region of siiE are tolerated in chickens and pigs, but are attenuated in cattle. Interestingly, siiE is split into two ORFs in both S. Typhi genome sequences, and part of the repetitive central region is absent from the genome of S. Paratyphi A. SiiE is secreted [38], indicating that in trans complementation by co-screened mutants does not obscure the identification of secreted colonization factors by TraDIS. All of the genes of the enteritis-associated SPI-5 that were disrupted by a transposon (pipACD, sopB and orfX) were required in all three species, although often with a milder phenotype in chickens. Several clusters of attenuated mutants were also found in the Salmonella chromosomal island (SCI, also known as SPI-6 in S. Typhi), including mutants in the hypothetical genes sciJ, sciQ, SL1344_0286A and sciX, the fimbrial subunit safA and its chaperone safB, the regulator sinR, SL1344_0301 (STM0305) which encodes a putative cytoplasmic protein, the pagN adhesin (STM0306) and sciZ (STM0307), a homologue of Shigella virG. Deletion of SCI affects invasion and virulence in a mouse intraperitoneal infection model [39], and the phenotype of a defined safA mutant has been confirmed in pigs [10].
Fimbriae play a well-established role in Salmonella attachment and intestinal colonization [40]. All twelve fimbrial operons were disrupted by multiple transposons in the TraDIS screen. No obvious host-specific phenotypes were seen, with a common pattern that mutants of fimbrial subunit genes were attenuated, whereas assembly genes were often dispensable, suggesting cross-talk in the assembly pathways. Stress responses are also important in the infection process, as Salmonella is subjected to a range of stresses including low pH, oxidative stress and heat shock [41]. The genetic components of these stress responses overlap [42], and many of these genes harboured transposons that resulted in attenuation. These included the sigma factor gene rpoE and its anti-sigma factor resA, the heat shock chaperone genes dnaK and dnaJ and the heat shock protease gene degP (htrA). Interestingly, several stress response genes are variably attenuated in the different hosts, suggesting species-specific stresses. These include the two-component regulatory system genes envZ and ompR, and the oxidative stress response genes dps, katE and proV which are all attenuated in pigs and cattle but show little or no attenuation in chickens. Conversely, transposon mutants in clpB, clpP and clpX, which encode proteases and are involved in the regulation of rpoS, are attenuated in chickens but not pigs or cattle.
Many S. Typhimurium genes beyond the classical virulence factors and stress response genes were revealed to be important for oral infection of livestock species. These include genes involved in nucleotide metabolism (pyrCD, purADGH, dgt, dcd, guaA, pyrCD and carAB), aromatic amino acid biosynthesis (aroABCDE), inorganic ion transport (trkAH, znuABC, fepCDG), protein synthesis (tufAB, fusA, efp, rplI, rpsK), protein export (tatABC, yajC) and many genes involved in carbohydrate metabolism. Additionally, numerous low GC clusters of genes with putative metabolic functions and multiple attenuating mutations were identified. Several global regulators, including crp, smpB and dam, result in attenuation in all three hosts, whereas another, fnr, appeared only to be important for infection in chickens. On occasion, TraDIS revealed functional data at a sub-genic level. For example, most of the insertions that disrupt the SPI-1 gene sptP result in attenuation, but insertions close to the 3′ end of the gene are tolerated. The gene rpoC, which encodes the β′ subunit of RNA polymerase, is essential in S. Typhimurium [43]. However, one transposon insertion in the chicken, pig and cattle dataset, and two in the mouse dataset, were identified close to the 3′ end of rpoC. These insertions would disrupt the extreme C-terminal end of the encoded protein, and were found to reduce the fitness of the mutants in the animal screens. Similarly, an insertion was found at the 3′ end of the essential polA gene which encodes DNA polymerase I, and this mutant was significantly attenuated in chickens, pigs and cattle.
The recent RNAseq-based analysis of the S. Typhimurium SL1344 transcriptome [44] identified a number of small-regulatory RNAs. As indicated in Table S4 several of these were implicated in colonization by TraDIS. For many it is difficult to demonstrate conclusively a colonization-associated phenotype from the TraDIS data alone, since we cannot preclude the potential for polar effects on adjacent genes. This is the case for InvR, which is encoded within SPI-1. Table S4 details only the sRNA genes annotated in the SL1344 genome (which differs from ST4/74 by just 8 SNPs [45]), but in the TraDIS data there are a number of attenuating transposons within large intergenic regions that could reveal the presence of novel sRNA genes.
The chicken, pig and cattle TraDIS data presented in Figure 4 indicate that a shared core set of 611 genes is required for efficient colonization of all three species, with a smaller set of species-specific colonization factors. The core set comprises approximately two thirds of the genetic requirements for infection of each individual species, and 48% of the total set of colonization-associated genes. There are 259 genes which are required for systemic infection of mice for which comparable data are available from the food-producing animals (Figure 4); of these, 140 also contribute to oral infection of chickens, pigs and cattle, and only 43 are putative mouse-specific factors. Many of the differences between the mouse and chicken/pig/cattle datasets may arise from the additional genetic requirements for infection via the oral route.
Although most colonization factors were necessary for infection of chickens, pigs and cattle, there were some patterns amongst the colonization factors that appeared to function in a host-specific manner that may reflect underlying differences in host biology. There are many genes associated with flagellar motility that are essential for infection of pigs but not required for chicken or calf infection, including fliY, flgK, fliN, flgN, fliB and fliZ. Several other flagellum-associated genes (flgB, flgL, fliL) are required for infection of cattle but not chickens or pigs. In chickens, many genes that are involved in anaerobic growth are required; these include genes involved in the production of group I hydrogenase (hypOBF, hybABCDF), fumarate reductase (frdAD), pflB, pfkA, rNTP reductase (nrdDG) and the global regulator Fnr. Differences in oxygen tension proximal to villus tips have been detected that modulate the regulation of Shigella virulence genes [46], therefore the requirement for distinct respiratory pathways by S. Typhimurium in food animals may reflect differences in the niches occupied. Also required in chickens, but apparently not calves or pigs, are the virK homologue ybjX, ilvGE, clpB and the his operon. The observation that many of the host-specific phenotypes were observed independently in multiple genes affecting the same pathways strongly suggests that these effects are due to differences in the within-host environment. For example, in relation to fumarate reductase there were nine independent frdA mutants that were all significantly attenuated in chickens; of these only one showed significant attenuation in pigs, and none in calves. Similarly, in relation to group I hydrogenase, from a total of 15 mutations affecting the hybABCDF genes, 12 showed significant attenuation in chickens, but none in pigs or calves (Table S2).
The serovar Typhimurium strain ST4/74 investigated here is a natural bovine isolate that elicits pathology typical of clinical salmonellosis in all the host species used in this study. However, some atypical S. Typhimurium strains exist that have lost the capability to colonize a broad range of hosts. For example, the laboratory-adapted strain LT2 and its derivatives tend to be avirulent or less virulent in mice relative to natural isolates of serovar Typhimurium [47] and ST4/74-based strains. It is noteworthy that the rpoS gene encoding the sigma factor σS, which is defective in LT2 and associated with the relative avirulence of this strain [48], was found to be required in all four species tested by screening of the S. Typhimurium libraries we describe (Table S3). TraDIS identified additional regions of the ST4/74 genome that are required for infection, but which are absent from the LT2 genome. These include several genes that are encoded within the same phage element: SL1344_1965, which is required for infection of mice, and SL1344_1929/30 and SL1344_1976, which are required for infection of chickens, pigs and cattle. Moreover, some Typhimurium strains have become adapted to a particular host, and the genome sequence of a human-adapted variant [49] reveals the decay of a number of genes important in colonization of food animals (e.g. allP, sseI, pipD, ydeE), but also other pseudogenes for which no role in food animals for the intact gene could be detected (e.g. ratB, ygbE, yhjU). Integration of genome sequences with high-resolution functional data of the kind we describe will provide further clues to explain the differential virulence of host-adapted or laboratory strains relative to natural isolates.
Our study represents the first comprehensive genome-wide survey of the role of thousands of Salmonella genes during colonization of the primary reservoirs of human non-typhoidal salmonellosis. TraDIS simultaneously assigned the genotype and relative fitness of 7702 distinct S. Typhimurium insertion mutants in chickens, pigs and cattle, representing over 90% of the mutants screened in pools of up to 475 mutants per animal. TraDIS therefore represents a significant advance in the reduction, refinement and replacement of animal models relative to STM, where only negatively-selected mutants tend to be interrogated and the vast majority of insertion sites and phenotypes are unreported [50]. Multiple lines of evidence suggest that the TraDIS data are robust and reliably reflect the fitness of the screened mutants in each host animal. Many of the attenuated mutants were found to harbour transposons in genes known to be involved in colonization. The TraDIS fitness scores correlated well with established datasets obtained using STM [10], [12] and TMDH [21]. Multiple mutations within the same gene or pathway usually gave comparable phenotypes, and most of the attenuated mutants demonstrated the same phenotype independently in the three different food-producing animal hosts. The examples of putatively host-specific attenuation tended to be restricted to particular pathways with multiple independent mutations. Finally, analysis of defined knockout mutants of targets chosen based on the TraDIS data reproduced attenuated phenotypes in all but one case.
Many novel colonization-associated genes were identified within the S. Typhimurium genome and the data provide an invaluable resource for the community to mine and extend. Moreover, TraDIS indicated that thousands of mutations exerted little or no effect in vivo, implying functional redundancy that may limit and refine the selection of targets for novel inhibitors, as previously suggested [51]. Unlike library screens conducted in murine typhoid models to date, we provide highly relevant data for control of intestinal S. enterica infections in food-producing animals, and thus zoonosis. Attenuating mutations may be suitable for selection and refinement of live vaccines for food-animals, and these in turn may express heterologous antigens. Further, the data will guide the interpretation of existing and fast emerging datasets on the repertoire, sequence and expression of Salmonella genes and aid the modelling of virulence in a wider evolutionary and ecological context. Our data reflect mutant phenotypes at a specific time and site, and further studies on the temporal and spatial role of Salmonella genes are likely to be informative. This study also establishes TraDIS as a quantitative technology in functional genomics, which has potential for widespread application beyond the realm of microbial pathogenicity.
Animal experiments were conducted according to the requirements of the Animals (Scientific Procedures) Act 1986 (project license number 30/2485) with the approval of the local Ethical Review Committee.
For full details of experimental animals, bacterial strains, materials, molecular biological techniques and statistical methods see Text S1. Briefly, a library of 8550 mini-Tn5 mutants was generated in a spontaneous nalidixic acid resistant variant of S. Typhimurium ST4/74. The mutants were combined into pools of 95 for chickens, and 475 for pigs and calves. Animals were inoculated orally and killed humanely 4 days (chickens and calves) or 3 days (pigs) after infection, or earlier if the clinical endpoint was reached. A section of an appropriate tissue (whole caeca for chickens, spiral colonic mucosa for pigs and distal ileal mucosa for calves) was homogenized and grown overnight on MacConkey agar plates to isolate the output bacteria.
Genomic DNA was prepared from the inocula and output samples, and fragmented to ∼300 bp. An Illumina adapter was ligated to the fragments, and PCRs were performed using an adapter-specific primer in conjunction with primers homologous to each end of the transposon. The sequences of all oligonucleotide primers used in this study are detailed in Table S5. The resultant products were sequenced on single end Illumina flowcells using a sequencing primer designed to read a 10 bp tag of transposon-derived sequence, plus 27 bp of flanking genomic DNA. Sequences containing the tag were mapped to the S. Typhimurium SL1344 genome sequence. A transposon was inferred to be present if there were corresponding reads derived from each end of the transposon in the input pool.
The number of reads corresponding to each transposon in the input pool, and the number of reads mapping to the equivalent position in the output pool data, were compared using DESeq [52]. The ratio of input∶output read counts was determined, after normalisation to account for variations in the total number of reads obtained for each sample, and expressed as log2(fold change), referred to as the fitness score. A negative fitness score indicates an attenuated mutant, a positive score indicates a mutant which was more abundant in the output pool than in the input. For strongly attenuated mutants, no reads were obtained in the output pool, so it was not possible to calculate a finite log2(fold change); such mutants were assigned an arbitrary fitness score of −15. For each individual mutant, the hypothesis that the fitness score was equal to zero (i.e. that the mutant was present at equivalent levels in the input and output pools) was tested using the negative binomial distribution as implemented in DESeq. DESeq models variance under the assumption that mutants with comparable levels of sequence coverage exhibit similar levels of dispersion. We exploited this model to estimate P values for all mutants whilst minimising the number of biological replicates by fitting using only those mutants for which replicate data points were available, and applying the resultant model to the data derived from all mutants.
Defined null mutants were obtained for twelve genes identified as attenuated in the TraDIS screen, and assessed in competition with wild-type ST4/74 during oral infection of chickens. The ratios of mutant∶wild-type bacteria from caecal isolates at day 4, 6 and 10 were compared with those in the inoculum, and the significance of any differences was tested using Student's t-test.
|
10.1371/journal.pgen.1001072 | Survival and Growth of Yeast without Telomere Capping by Cdc13 in the Absence of Sgs1, Exo1, and Rad9 | Maintenance of telomere capping is absolutely essential to the survival of eukaryotic cells. Telomere capping proteins, such as Cdc13 and POT1, are essential for the viability of budding yeast and mammalian cells, respectively. Here we identify, for the first time, three genetic modifications that allow budding yeast cells to survive without telomere capping by Cdc13. We found that simultaneous inactivation of Sgs1, Exo1, and Rad9, three DNA damage response (DDR) proteins, is sufficient to allow cell division in the absence of Cdc13. Quantitative amplification of ssDNA (QAOS) was used to show that the RecQ helicase Sgs1 plays an important role in the resection of uncapped telomeres, especially in the absence of checkpoint protein Rad9. Strikingly, simultaneous deletion of SGS1 and the nuclease EXO1, further reduces resection at uncapped telomeres and together with deletion of RAD9 permits cell survival without CDC13. Pulsed-field gel electrophoresis studies show that cdc13-1 rad9Δ sgs1Δ exo1Δ strains can maintain linear chromosomes despite the absence of telomere capping by Cdc13. However, with continued passage, the telomeres of such strains eventually become short and are maintained by recombination-based mechanisms. Remarkably, cdc13Δ rad9Δ sgs1Δ exo1Δ strains, lacking any Cdc13 gene product, are viable and can grow indefinitely. Our work has uncovered a critical role for RecQ helicases in limiting the division of cells with uncapped telomeres, and this may provide one explanation for increased tumorigenesis in human diseases associated with mutations of RecQ helicases. Our results reveal the plasticity of the telomere cap and indicate that the essential role of telomere capping is to counteract specific aspects of the DDR.
| The telomeric DNA of most eukaryotes consists of G-rich repetitive DNA with a 3′ single stranded DNA (ssDNA) overhang. In human and budding yeast (Saccharomyces cerevisiae) cells, the 3′ ssDNA overhang is bound by essential telomere capping proteins, POT1 and Cdc13 respectively. Maintenance of telomere capping is essential for the survival of cells. The RecQ helicases are a family of highly conserved proteins involved in the maintenance of telomere and genome stability. Loss of function of three RecQ helicases in humans results in cancer predisposition disorders Bloom's syndrome (BS), Werner's syndrome (WS), and Rothmund Thomson syndrome (RTS). Here we found that the RecQ helicase in budding yeast, Sgs1, plays a critical role in the resection of uncapped telomeres. Strikingly, simultaneous inactivation of Sgs1, the exonuclease Exo1, and checkpoint protein Rad9 allows budding yeast cells to divide in the absence of Cdc13, indicating that the essential role of the telomere cap is to counteract specific components of DNA damage response pathways. We speculate that, in certain genetic contexts, mammalian RecQ helicase also inhibit growth of cells with telomere capping defects, and a defect in this role could contribute to increased levels of tumorigenesis in BS, WS, and RTS patients.
| The ends of linear chromosomes pose two major threats to the proliferative potential and genetic stability of eukaryotic cells: inappropriate activation of the DNA damage response (DDR) and progressive shortening of the chromosome ends due to the end replication problem. The telomere, a specialised structure consisting of G-rich repetitive DNA and associated protein complexes at the end of linear chromosomes helps to overcome both of these problems by recruiting telomere capping proteins and telomerase to telomeres [1]. In metazoan organisms, maintenance of telomere integrity is critical for protecting against the processes of cancer and ageing [2], [3].
The mechanisms of chromosome end protection by telomere capping proteins are conserved in eukaryotes. The telomeric DNA of most eukaryotes consists of G-rich repetitive DNA with a 3′ single stranded DNA (ssDNA) overhang. In mammalian cells, telomeric DNA is bound by the shelterin complex [4], [5]. Two proteins of shelterin, TRF1 and TRF2, bind to the double stranded telomeric repeat and recruit TIN2, Rap1, TPP1 and POT1 to the telomere and help to cap the chromosome end [4], [5]. In addition to the shelterin complex, another conserved telomere capping complex, the CST complex, which consists of CTC1, STN1 and TEN1, has been recently described in mammal and plant cells [6]–[8]. In Saccharomyces cerevisiae, telomeric ssDNA is capped by the essential Cdc13-Stn1-Ten1 complex, analogous to the CST complex in other cell types [7], [9].
ssDNA binding proteins like POT1 and Cdc13 bind to the telomeric 3′ ssDNA overhang and play multiple roles to protect and maintain the chromosome end [4], [5]. Deletion of POT1 or CDC13 results in lethality in both mammalian and yeast cells [10]–[12]. Conditional inactivation of POT1 in mammalian cells leads to telomeric ssDNA generation, ATR-dependent checkpoint activation, deregulation of telomerase, telomere recombination and telomere fusion [4]. Similarly, acute inactivation of Cdc13 by the temperature sensitive cdc13-1 allele in budding yeast induces telomeric ssDNA generation, recombination and Mec1 (ATR orthologue) dependent cell cycle arrest [10], [13].
The RecQ helicases are a family of highly conserved proteins involved in the maintenance of genome stability and at telomeres [14]. There are five RecQ helicases in humans. Loss of function of three of these results in cancer predisposition disorders Bloom's syndrome (BS, defective in BLM), Werner's syndrome (WS, defective in WRN) and Rothmund Thomson syndrome (RTS, defective in RECQ4) [14]. WS and RTS are also characterised by various features of premature ageing. There is only one RecQ helicase in Saccharomyces cerevisiae - SGS1. RecQ helicases are 3′-5′ DNA helicases that unwind a variety of DNA replication and recombination structures. It is believed that in the absence of RecQ helicases, genomic rearrangements that arise due to a failure to unwind pathological DNA structures formed at stalled replication forks are a major cause for genetic instability and increases in tumourigenesis associated with these syndromes [14]. The RecQ helicases also play roles in maintaining telomeres and cells derived from WS patients show premature entry into telomere dysfunction-induced senescence, which has been proposed to be a cause for the premature ageing phenotype of WS [15]. Similarly, in yeast, deletion of SGS1 induces rapid senescence in yeast cell lacking telomerase and Sgs1 is required for the generation of certain type of recombination-dependent ALT (Alternative Lengthening of Telomere) like survivors [16]–[18]. Interestingly, the RecQ helicase in fission yeast, Rqh1 acts on dysfunctional telomeres to promote telomere breakage and entanglement [19].
Several recent studies have identified important roles for the RecQ helicase family in resection of DNA double strand breaks (DSBs) to generate ssDNA, and in generating 3′ overhangs at shortened telomeres [20]–[24]. These resection activities were partially redundant with the 5′ to 3′ exonuclease, Exo1. Since cdc13-1 induced telomere uncapping leads to telomere resection controlled by Exo1 and other, as yet unidentified, nuclease activities [25], we speculated that Sgs1 might contribute to resection and response to telomere uncapping in cdc13-1 mutants.
Here we examined the role of Sgs1 in responding to telomere capping defects in cdc13-1 mutants. We show that although Sgs1 contributes to the good growth of cdc13-1 mutants, it strongly inhibits the growth of cdc13-1 exo1Δ rad9Δ mutants. Our experiments reveal important and yet complex functions of Sgs1 in regulating the growth of budding yeast cells with telomere capping defects. Our results have implications for the role of mammalian RecQ helicases at dysfunctional telomeres.
To begin to test the effect of Sgs1 on the response to telomere uncapping, we first deleted SGS1 in a cdc13-1 background. Cdc13-1 becomes increasingly non-functional at temperatures above 23°C, which leads to telomere uncapping, checkpoint activation and temperature sensitive growth [10]. We found that in marked contrast to deletion of EXO1, deletion of SGS1 made cdc13-1 cells more temperature sensitive at 26°C (Figure 1A). This suggests that Sgs1 contributes to the stability of uncapped telomeres in cdc13-1 mutants, and that the effect of Sgs1 differs from the effect of Exo1 which attacks uncapped telomeres, generates ssDNA and inhibits growth (Figure 1A) [25]. The role of Sgs1 in stabilising telomeres is consistent with other experiments in budding yeast and mammalian cells [16], [26].
Since Sgs1 functions redundantly with Exo1 to generate ssDNA at DSBs and shortened telomeres [20]–[22], [24], we examined the effect of deleting both Sgs1 and Exo1 activities on growth of cdc13-1 mutants. We found that sgs1Δ exo1Δ mutants grew slightly poorer at all temperatures in both CDC13 and cdc13-1 backgrounds (Figure 1A, Figure S1). The poor growth of CDC13 sgs1Δ exo1Δ mutants has been reported previously [20], [21]. Despite the poor growth of sgs1Δ exo1Δ mutants, deletion of SGS1 allowed cdc13-1 exo1Δ strains to grow slightly better at high temperature, since more cells appeared able to divide and grow at 30°C (Figure 1A). This suggests that in the absence of Exo1, Sgs1 may play a role inhibiting cell division when telomeres uncap.
Checkpoint pathways, as well as nucleases, inhibit the growth of cdc13-1 cells at high temperature [25], [27]. We have previously shown that simultaneous deletion of EXO1 and the checkpoint gene RAD9 allow better suppression of cdc13-1 temperature sensitivity than either single deletion [25]. Strikingly, deletion of SGS1 fully suppressed the temperature sensitivity of cdc13-1 rad9Δ exo1Δ strains and allowed cdc13-1 rad9Δ sgs1Δ exo1Δ strains to form colonies at 36°C, a temperature at which Cdc13 is expected to be completely non-functional (Figure 1A, Figure S2) [10]. This suggests that deletion of SGS1, EXO1 and RAD9 allows budding yeast cells to grow in the absence of any Cdc13 activity.
It was possible that growth of cdc13-1 rad9Δ sgs1Δ exo1Δ strains at 36°C was due to second site mutations/changes since we previously showed that at low rates (5×10−5) cdc13-1 exo1Δ rad9Δ strains could generate clones that can grow at 36°C [28]. Therefore to test the possibility that a second site mutation was responsible for the growth of cdc13-1 rad9Δ sgs1Δ exo1Δ strains at 36°C, we performed a backcross with an independent cdc13-1 strain. We found that all cdc13-1 rad9Δ sgs1Δ exo1Δ segregants derived from the cross were able to grow at 36°C (Figure S3). This demonstrated that no independently segregating mutation contributed to growth of cdc13-1 rad9Δ sgs1Δ exo1Δ strains at high temperature, and suggests that Sgs1, Exo1 and Rad9-dependent activities combine to inhibit the growth of telomere capping defective cdc13-1 mutants at 36°C.
At low rates a small fraction of telomerase-null ALT survivors that maintain telomeres by homologous recombination have also been shown to be able to survive without Cdc13 [29]. Therefore to test whether alterations in telomere structure were responsible for the good growth of cdc13-1 rad9Δ sgs1Δ exo1Δ strains at 36°C, we examined telomere structure by Southern blot. As positive controls, we examined both Type I and Type II telomerase-null ALT survivors which have amplified the Y' region and the TG repeats respectively. We found that cdc13-1 rad9Δ sgs1Δ exo1Δ strains had telomere structures that looked similar to wild-type strains at both 23°C and 36°C (Figure 1B, Figure 1C and data not shown). Furthermore, we found that cdc13-1 rad9Δ sgs1Δ exo1Δ strains can grow and maintain normal telomere structure at 36°C without Rad52, suggesting that homologous recombination is not required for the survival of these strains following telomere uncapping (Figure S4). Therefore we conclude that cdc13-1 rad9Δ sgs1Δ exo1Δ strains are not ALT survivors and the growth at 36°C is independent of major alterations in telomeric DNA structure.
In Schizosaccharomyces pombe, inactivation of the telomere capping protein Pot1 results in growth crisis and generation of survivors with circular chromosomes [30]. To test whether cdc13-1 rad9Δ sgs1Δ exo1Δ strains survive growth at 36°C by circularising their chromosomes, we used pulsed-field gel electrophoresis to examine chromosome structure. Circular chromosomes do not enter pulsed field gels. We found that at 23°C, cdc13-1 rad9Δ sgs1Δ exo1Δ strains contain linear chromosomes that were indistinguishable from wild-type or rad9Δ sgs1Δ exo1Δ strains (Figure 1D). Interestingly, the chromosomes of cdc13-1 rad9Δ sgs1Δ exo1Δ strain remain linear at 36°C (Figure 1D). We conclude that the ability of cdc13-1 rad9Δ sgs1Δ exo1Δ strains to grow without telomere capping by Cdc13 is not due to chromosome circularisation.
The Mre11-Rad50-Xrs2 (MRX) complex and the nuclease Sae2 can resect DSBs in an Sgs1 and Exo1-independent pathway [20], [22]. To test whether inhibition of this pathway also has an effect on response to telomere uncapping and ability of cdc13-1 cells to grow at 36°C, we examined the effect of deleting SAE2 on growth of cdc13-1 mutants. We found that like the MRX complex [31], Sae2 has a role in the protection of uncapped telomeres as deletion of SAE2 renders cdc13-1 cells slightly more temperature sensitive at 26°C (Figure S5). Furthermore, deletion of SAE2, in combination with deletion of SGS1 or EXO1 in the presence or absence of RAD9 does not allow cdc13-1 cells to grow at 36°C (Figure S5). We were unable to analyse sgs1Δ sae2Δ exo1Δ triple mutant cells as simultaneous deletion of these genes is lethal, as reported previously [22], [24] (Figure S6). We conclude that Sae2 has a protective role at uncapped telomeres and that unlike Sgs1, deletion of SAE2 does not allow cdc13-1 rad9Δ exo1Δ cells to grow at 36°C.
To further address the surprising ability of cdc13-1 rad9Δ sgs1Δ exo1Δ strains to grow at 36°C, we used synchronous cultures to examine two cellular responses to telomere uncapping: checkpoint activation and telomeric ssDNA generation. To examine checkpoint activation, we first quantified the fraction of cells arrested at medial nuclear division induced by cdc13-1 mutation at 36°C. A cdc15-2 mutation, which inhibits mitotic exit, was used to inhibit further cell cycle progression of any cdc13-1 cells that failed to arrest at medial nuclear division or escaped arrest and therefore progressed to reach the late nuclear division stage [27]. As expected, control cdc13-1 cells were largely arrested at medial nuclear division 80 minutes after release from G1 arrest to 36°C and the cells remained arrested for up to 4 hours (Figure 2A and 2B) [25]. A cdc13-1 sgs1Δ strain showed a similar cell cycle arrest to the cdc13-1 strain (Figure 2A and 2B). cdc13-1 exo1Δ strains showed no defect in the activation of checkpoint as most cdc13-1 exo1Δ cells were arrested at medial nuclear division after 80 minutes, but some cells escaped from the arrest at later time points, as previously reported (Figure 2A and 2B) [25]. We found that deletion of both SGS1 and EXO1 slightly reduced the number of cells arrested at medial nuclear division (to approximately 60%) after 80 minutes (Figure 2A). This is partly due to a slower cell cycle progression and also due to a mild defect in checkpoint activation as approximately 15% of cdc13-1 sgs1Δ exo1Δ cells had entered late nuclear division by 80 minutes (Figure 2B, Figure S7). At later time points, increasing fractions of cdc13-1 sgs1Δ exo1Δ cells passed through medial nuclear division and accumulated at the late nuclear division stage, presumably due to a defect in either activation or maintenance of checkpoint (Figure 2B). As expected, deletion of RAD9 completely eliminated checkpoint arrest in all these backgrounds (Figure 2A and 2B). We conclude that in the absence of both Sgs1 and Exo1, checkpoint activation/maintenance following telomere uncapping in cdc13-1 mutants is partially defective.
We next examined Rad53 mobility shifts, caused by checkpoint-kinase-dependent phosphorylation, as a biochemical marker for checkpoint activation in cdc13-1 strains [32]. It has been shown that Rad53 is strongly phosphorylated two hours following telomere uncapping in cdc13-1 mutants [32]. Consistent with this finding, we observed a strong Rad53 phosphorylation shift two hours after cdc13-1 cells were shifted to 36°C (Figure 2C). Deletion of SGS1 in cdc13-1 strains had little effect on this Rad53 phosphorylation but deletion of EXO1 mildly reduced Rad53 phosphorylation (Figure 2C). Interestingly, we found that deletion of both SGS1 and EXO1 caused a more severe defect on Rad53 phosphorylation, although some residual Rad53 phosphorylation could be detected (Figure 2C). As expected, Rad53 phosphorylation was abolished when RAD9 was deleted in all strains (Figure 2C). Taken together, the cell cycle arrest and Rad53 phosphorylation experiments suggest that Sgs1 and Exo1 control parallel pathways to activate the checkpoint induced by uncapped telomeres. However, the presence of residual Rad53 phosphorylation and checkpoint arrest in cdc13-1 sgs1Δ exo1Δ strains indicates that there is at least one additional pathway to activate the checkpoint cascade after telomere uncapping in the absence of Sgs1 and Exo1.
The major stimulus for checkpoint cascades after telomere uncapping is ssDNA accumulation caused by 5′ to 3′ resection, generating ssDNA specifically on the TG rich, 3′ strand at telomeres [25], [33]. To study the role of Sgs1 in resection, we used Quantitative amplification of ssDNA (QAOS) [25], [33] to examine ssDNA production at two repetitive loci in subtelomeric regions and three single copy loci near the right telomere of chromosome V. We first examined the Y' repeats which are found in two thirds of yeast telomeres. We measured ssDNA accumulation at two different Y' loci, Y'600 and Y'5000, located approximately 600 bp and 5000 bp from many chromosome ends. As previously reported, we detected accumulation of ssDNA specifically on the TG strands and this ss DNA was partly dependent on Exo1 (Figure 3C and 3D, Figure S8) [25]. This ssDNA was also partly dependent on Sgs1, this was particularly clear at the Y'5000 locus (Figure 3C). The difference between ssDNA accumulation in cdc13-1 and cdc13-1 sgs1Δ cells was perhaps not as significant at Y'600 as some standard deviations overlapped (Figure 3D). Interestingly, in the absence of both Sgs1 and Exo1, telomere resection was not detectable at Y'5000, but was still detectable at Y'600 (Figure 3C and 3D, Figure S9). This suggests that Sgs1 and Exo1 independently regulate 5′-3′ resection at cdc13-1-induced uncapped telomeres, especially further away from the telomere ends at Y'5000 loci. This resection defect likely provides an explanation to the weak checkpoint activation observed in cdc13-1 sgs1Δ exo1Δ strains (Figure 2, Figure S7). At Y'600, the ssDNA generated in cdc13-1 sgs1Δ exo1Δ cells was not significantly different from that seen in cdc13-1 exo1Δ cells (Figure 3D). The presence of ssDNA at Y'600 loci in cdc13-1 sgs1Δ exo1Δ strains suggests that there is a third pathway that can resect uncapped telomeres near the telomere ends (Figure 3D). The small residual amounts of telomeric ssDNA likely provide the stimulus for checkpoint activation in cdc13-1 sgs1Δ exo1Δ strains.
It has been shown that following telomere uncapping in cdc13-1 mutants, the resection of uncapped telomeres extends into the single copy regions of chromosomes [10]. To test whether Sgs1 affects accumulation of ssDNA in single copy regions, we examined ssDNA accumulation further away from the telomeres at three single copy loci, YER188W, YER186C and PDA1 located at 8.5 kb, 14.5 kb and 29.7 kb from the right end of Chromosome V (Figure 3A and 3B, Figure S8). As previously observed, we detected ssDNA in cdc13-1 cells at both the YER188W, YER186C loci, but not at the PDA1 locus which is located further away from the telomeres (Figure 3A and 3B, Figure S8). We found that the resection at these loci were highly dependent on EXO1 as previously reported [25], but considerably less dependent on Sgs1 (Figure 3A and 3B, Figure S8). Thus our data suggests that Sgs1 and Exo1 regulate different types of nuclease activities in cdc13-1 mutants.
We next examined resection in the absence of RAD9 because Rad9 plays an important role in the inhibition of nuclease activities at uncapped telomeres [25], [27], [34]. Specifically, deletion of RAD9 allows nuclease(s) activities to generate ssDNA further away from the telomere ends, even in the absence of Exo1, which results in the resection of DNA up to 30 kb away from the telomeres [25]. We first examined ssDNA production at the Y' repeats and found that as in RAD9+ background, DNA resection of uncapped telomeres at the Y'5000 loci was totally dependent on the combined activities of Sgs1 and Exo1 as little ssDNA was detected at these loci following telomere uncapping (Figure 3G, Figure S8). This observation again supports the idea that Sgs1 and Exo1 act in two alternative pathways to resect the DNA at these loci. Consistent with what was observed in RAD9+ background, there appears to be another nuclease that can resect the uncapped telomeres near the telomere end at Y'600 in the absence of both Sgs1 and Exo1 (Figure 3H). As previously reported, we found that deletion of RAD9 allows ssDNA generation at single copy loci up to 30 kb away from the telomere ends even in the absence of Exo1 (compare cdc13-1 exo1Δ in Figure 3A and 3B and cdc13-1 rad9Δ exo1Δ in Figure 3E and 3F). Interestingly, deleting SGS1 abolished this Exo1-independent ssDNA generation (compare cdc13-1 rad9Δ exo1Δ and cdc13-1 rad9Δ sgs1Δ exo1Δ) (Figure 3E and 3F). This suggests that Rad9 inhibits an Sgs1-dependent, but Exo1-independent pathway of generating ssDNA further away from the uncapped telomere ends.
The results described above suggest that Sgs1 and Exo1 contribute to two distinct pathways of resection near the telomeres of cdc13-1 strains. Our data suggests that attenuation of telomeric DNA resection pathways by deletion of both SGS1 and EXO1, coupled with inactivation of checkpoint machinery by deletion of RAD9, is sufficient to allow cells to grow in the absence of telomere capping by Cdc13.
Cdc13 plays at least two important roles at telomeres, one capping chromosome ends, the other recruiting telomerase [35]. cdc13-1 strains grown at 36°C would be expected to be defective in both activities. Therefore to test whether cdc13-1 rad9Δ sgs1Δ exo1Δ strains can maintain telomeres and/or tolerate long term absence of Cdc13, we grew cells for many passages at 36°C by repeatedly restreaking them onto plates which were incubated at 36°C (Figure 4A). We found that cdc13-1 rad9Δ sgs1Δ exo1Δ strains showed progressively slower growth with repeated passage at 36°C, but that some strains eventually accumulated faster growing colonies (Figure 4A). This pattern of growth is similar to telomerase deficient tlc1Δ cells which enter growth senescence due to shortened telomeres, and eventually accumulate post senescent survivors which survive by utilising recombination-dependent mechanisms to maintain their telomeres (Figure 4A) [36], [37].
To test whether cdc13-1 rad9Δ sgs1Δ exo1Δ strains growing at 36°C were behaving like telomerase deficient strains, we examined the telomere structure by Southern blot. At early passage, cdc13-1 rad9Δ sgs1Δ exo1Δ strains grown at 36°C had a comparatively normal telomeric DNA structure, but the terminal telomere fragments were weak (Figure 4B, Figure 1C). This confirms that growth of cdc13-1 rad9Δ sgs1Δ exo1Δ strains at 36°C does not depend on gross changes in telomere structure. By passage 2, the terminal telomere fragments of these strains had become shorter and weaker, similar to but distinct from the telomeres in a tlc1Δ strain, possibly explaining the poor growth of these strains at this passage at 36°C (Figure 4A and 4B). At passage 9, when some of the cdc13-1 rad9Δ sgs1Δ exo1Δ strains grew well they showed alternative telomere structures similar to that in a Type I tlc1Δ survivor and had amplified the 5.5 and 6.5 kb Y' repeats (Lanes 10-12, Figure 4B) [36]. Unlike the Type I tlc1Δ survivor shown, the terminal fragments in cdc13-1 rad9Δ sgs1Δ exo1Δ survivors remained very weak. We conclude that initial growth of cdc13-1 rad9Δ sgs1Δ exo1Δ strains at 36°C does not depend on gross alteration in telomere structure. However, continued growth without functional Cdc13 results in telomere shortening, cellular senescence and generation of post-senescence survivors with altered telomere structures.
The growth of cdc13-1 rad9Δ sgs1Δ exo1Δ strains at 36°C suggested that CDC13 might not be essential in cells lacking SGS1, EXO1 and RAD9. To directly test this hypothesis, we created diploid strains heterozygous for sgs1Δ, exo1Δ, rad9Δ and cdc13Δ and dissected tetrads following sporulation (Figure 5A, Figure S10). We found that cdc13Δ rad9Δ sgs1Δ exo1Δ spores were viable and formed colonies at a comparable frequency (average 89%) to CDC13 rad9Δ sgs1Δ exo1Δ spores, whereas other cdc13Δ genotypes were unviable (Figure 5A, Figure S10, Table S1). This result shows that deleting SGS1, EXO1 and RAD9 genes from yeast is sufficient to allow yeast cells to grow and divide in the absence of the usually essential telomere capping protein Cdc13. We found that the telomeres of cdc13Δ rad9Δ sgs1Δ exo1Δ strains shortened very quickly and the terminal fragments of fresh cdc13Δ rad9Δ sgs1Δ exo1Δ strains (marked with asterisk) were already shortened when first examined (Figure 5B). Similar to cdc13-1 rad9Δ sgs1Δ exo1Δ strains at 36°C, these terminal fragments of cdc13Δ rad9Δ sgs1Δ exo1Δ strains were very weak (Figure 5B). We found that these cdc13Δ rad9Δ sgs1Δ exo1Δ cells show variable rates of senescence, and faster growing ALT survivors that can grow for at least eight passages without Cdc13 eventually emerged (Figure S11, Figure S12). We conclude that Cdc13 is not an essential telomere capping protein in rad9Δ sgs1Δ exo1Δ strains, but that such strains eventually maintain telomeres like telomerase deficient cells.
Telomere capping protects the ends of linear chromosomes from the activation of DDR. However, cell division-induced telomere shortening in human somatic cells is believed to lead to progressive loss of telomere capping, and causes chromosome ends to be recognised as DNA breaks, which consequently activate a DDR and cell cycle arrest [4]. This telomere dysfunction-induced cell cycle arrest is thought to contribute to the establishment of replicative cellular senescence and the ageing process, and represent an important barrier to tumour formation [2], [38], [39]. Thus it is important to understand the cellular response to telomere uncapping.
Budding yeast temperature sensitive cdc13-1 mutants have been an informative model to understand the eukaryotic cell response to telomere uncapping. It was shown previously that at low rates cdc13-1 rad9Δ exo1Δ cells (5×10−5) or telomerase deficient survivors (8×10−5 to 4×10−2) were able to grow without telomere capping by Cdc13, but the nature of any other genetic or other change necessary to permit growth was unclear [28], [29]. Here we show that the simultaneous deletion of SGS1, EXO1 and RAD9 is sufficient to make Cdc13 dispensable for cell proliferation, as most cdc13Δ rad9Δ sgs1Δ exo1Δ cells are viable. This indicates that the essential function of telomere capping in yeast is to counteract the proliferative barrier exerted by specific components of the DDR. Our data suggests that attenuation of resection stimulating activities of Sgs1 and Exo1, together with DNA damage checkpoint inactivation is sufficient to allow yeast cells to grow in the absence of telomere capping by Cdc13. This suggests that in the absence of these proteins, either telomere capping is not required or that an alternative capping strategy can function in the absence of Cdc13. Intriguingly, the telomeres in cdc13-1 rad9Δ sgs1Δ exo1Δ strains (at 36°C) and cdc13Δ rad9Δ sgs1Δ exo1Δ strains look different to those in wild-type strains, as the telomeres have very weak terminal restriction fragments. Even though we showed that cdc13-1 rad9Δ sgs1Δ exo1Δ strains do not require Rad52 to survive telomere uncapping at 36°C, we cannot rule out a role for homologous recombination proteins at telomeres in cdc13Δ rad9Δ sgs1Δ exo1Δ strains.
Besides capping telomeres, Cdc13 is also required for the recruitment of telomerase [35], and as expected we found that in the absence of Cdc13, telomeres eventually shorten and are then maintained by the alternative lengthening of telomere (ALT) mechanisms, usually seen in telomerase deficient cells. Our work highlights the plasticity of the telomere cap and shows how modulation of the DDR can provide an important mechanism to overcome the proliferative barrier induced by dysfunctional telomeres. As telomere uncapping-induced cellular senescence represents an important proliferative barrier to prevent cancer formation, the mechanisms described here could be relevant to understanding the malignant transformation of human cells [2], [3].
The RecQ helicases, including Sgs1, are a family of highly conserved protein involved in the maintenance of genome stability and suppression of cancer formation in humans [14]. It is believed that in the absence of RecQ helicases, genome rearrangements that arise at stalled replication forks are a major cause for the increased tumorigenesis [14]. RecQ helicases also participate in the maintenance of telomeres, for example, deletion of SGS1 induces rapid senescence in yeast cells lacking telomerase and cells derived from WS patients show premature entry into telomere dysfunction-induced senescence [15], [16]. Here we report a similar role for Sgs1 in the protection of uncapped telomeres as deleting SGS1 makes cdc13-1 cells more temperature sensitive. However, paradoxically, we also show that Sgs1 plays a role in the resection of uncapped telomeres, an activity that is modulated by the checkpoint protein Rad9. This Sgs1-dependent resection activity leads to a critical role of Sgs1 in the inhibition of cell proliferation when telomeres uncap in the absence of Exo1 and Rad9. It is tempting to speculate that in certain genetic contexts RecQ helicase may also inhibit growth of mammalian cells with telomere capping defects, and a defect in this role could contribute to increased levels of tumorigenesis in BS, WS and RTS patients.
Previous studies have suggested that following telomere uncapping in cdc13-1 mutants, the telomeres were resected by three somewhat redundant nuclease activities, Exo1, ExoX and ExoY [25]. ExoX was defined as a Rad9-inhibited nuclease that was strongly dependent on Rad24 [25]. ExoY was a nuclease that acts on uncapped telomeres in the absence of Exo1 and Rad24 (ExoX). In this study, we have identified Sgs1 as a protein that shares similar properties with ExoX, because it is required for the resection of uncapped telomeres in an Exo1-independent pathway and its activity is inhibited by Rad9. Since the helicase Sgs1 acts with a nuclease Dna2 in the resection of DSBs and at shortened telomeres [20], [24], we speculate that Sgs1 and Dna2 might work together to regulate resection at uncapped telomeres (may contribute to ExoX). We also found that in the absence of Sgs1 and Exo1, ssDNA still accumulated at cdc13-1 induced uncapped telomeres, consistent with the existence of another nuclease (possibly ExoY). It is interesting to note that the Mre11-Rad50-Xrs2 (MRX) complex and the nuclease Sae2 can resect DSBs in an Sgs1 and Exo1-independent pathway [20], [22]. Thus a candidate for ExoY is MRX/Sae2. However, we have previously found that the MRX complex plays a capping role at telomeres, rather than contributing to resection [31]. Similarly, we found that Sae2 also has a protective role at uncapped telomeres. Therefore we believe that like the MRX complex, Sae2 is unlikely to be ExoY. Thus although uncapped telomeres share significant similarities with DSBs, there are also clearly significant differences. Future studies will be required to resolve the roles of the different types of nucleases and nuclease regulators at uncapped telomeres and DSBs.
All experiments were performed using Saccharomyces cerevisiae W303 strains as listed in Table S2. Gene disruptions of SGS1 and CDC13 were constructed by a one step PCR-mediated method using kanMX and hphMX cassettes respectively [40]. All the yeast strains were generated by standard genetic crosses.
Single colonies were inoculated into 2 ml YPD and incubated at 23 °C overnight until saturation. Five-fold serial dilution of the cultures were spotted onto agar plates using a 48 or 96-prong replica plating device. Plates were incubated for 2–4 days at different temperatures before being photographed using a SPimager (S&P Robotics).
Protein extracts were prepared by using TCA extraction method, as previously described [32]. Briefly, cells were collected, washed in water, resuspended in 10% TCA and broken using glass beads. Protein suspensions were then resuspended in Laemmli buffer. Samples were boiled for three minutes, centrifuged for 10 minutes and the supernatant used as the protein extract. For Western blotting, proteins were separated on 7.5% SDS-PAGE. Anti-Rad53 and anti-tubulin antibodies were from Dan Durocher, Toronto and Keith Gull, Oxford respectively.
DNA was purified from yeast strains following incubation in liquid culture for 24 or 48 hours. The DNA was cut with XhoI, run on a 0.8% (0.5xTBE) gel at 1V/cm overnight and transferred to a Magna nylon membrane. The membrane was then hybridised with a Y'+TG probe (synthesized using DIG-High Prime Labelling and Detection Kit (Roche)) as described previously [31]. The membrane was stripped and reprobed with a CDC15 probe as described previously [31].
Yeast strains with cdc13-1 cdc15-2 bar1 mutations were arrested in G1 at 23°C and released into 36°C to induce telomere uncapping, cell cycle position were scored using DAPI staining on a Nikon Eclipse 50i microscope. The amount of ssDNA at the TG and the AC strands at telomeric and single copy loci were measured by QAOS (quantitative amplification of single-stranded DNA) as described previously [41].
Pulsed-field gel electrophoresis were carried out as described previously with the following modification: gels were run at 6.0 V/cm with a switch time of 70 seconds for 15 hours and 120 seconds for 11 hours [42].
|
10.1371/journal.pbio.1000358 | Proteorhodopsin Phototrophy Promotes Survival of Marine Bacteria during Starvation | Proteorhodopsins are globally abundant photoproteins found in bacteria in the photic zone of the ocean. Although their function as proton pumps with energy-yielding potential has been demonstrated, the ecological role of proteorhodopsins remains largely unexplored. Here, we report the presence and function of proteorhodopsin in a member of the widespread genus Vibrio, uncovered through whole-genome analysis. Phylogenetic analysis suggests that the Vibrio strain AND4 obtained proteorhodopsin through lateral gene transfer, which could have modified the ecology of this marine bacterium. We demonstrate an increased long-term survival of AND4 when starved in seawater exposed to light rather than held in darkness. Furthermore, mutational analysis provides the first direct evidence, to our knowledge, linking the proteorhodopsin gene and its biological function in marine bacteria. Thus, proteorhodopsin phototrophy confers a fitness advantage to marine bacteria, representing a novel mechanism for bacterioplankton to endure frequent periods of resource deprivation at the ocean's surface.
| It is estimated that marine microscopic algae—phytoplankton—are responsible for half of the Earth's photosynthesis. As much as half of the surface ocean bacteria have proteorhodopsins, which are membrane proteins that allow harvesting of energy from sunlight, implying a potentially significant role of non–chlorophyll-based phototrophy in oceanic carbon cycling and energy flux. Functional evidence for specific roles for proteorhodopsins in native marine bacteria and the marine environment remains surprisingly scarce. One reason for this is the lack of marine bacteria (containing proteorhodopsin genes) that can be maintained in laboratory culture and that are tractable to genetic manipulation. In this study, we show that a proteorhodopsin-containing member of the widespread marine genus Vibrio displays light-enhanced survival during starvation in seawater. Furthermore, growth recovery experiments showed that bacteria starving in the light could more rapidly respond to improved growth conditions than those incubated in the dark. We generated a proteorhodopsin deficient Vibrio strain and used it to confirm that light-dependent survival of starvation was mediated by the proteorhodopsin. Proteorhodopsin phototrophy thus provides a physiological mechanism that allows surface ocean bacteria to manage an environment where resource availability fluctuates markedly.
| Proteorhodopsins (PRs) are membrane-embedded, light-driven proton pumps, which generate a chemiosmotic potential by translocating protons across an energy-transducing membrane [1]–[3]. This proton gradient can subsequently be used for production of biologically available energy in the form of adenosine triphosphate (ATP), the basic energy currency conserved among living beings, and/or for fueling motility and enhancing solute transport across the membrane [2],[3]. The discovery of PR in marine bacteria revealed a possible role of non–chlorophyll-based phototrophy in biogeochemical carbon cycling and energy fluxes in the ocean [1]. Consistent with their inferred ecological importance, PRs are highly abundant and exceedingly genetically diverse in aquatic environments [4]–[10]. The large genetic diversity of PRs suggests that they could potentially display an array of physiological and ecological functions [2],[11],[12]. However, there is a striking lack of knowledge concerning which biological function PRs fulfill and how they contribute to the success of PR-containing bacteria in the marine environment.
Survival and reproduction are the main components that determine fitness, a fundamental concept in ecology. For marine bacteria, the molecular mechanisms that contribute to the variation in these components are poorly understood. Previous studies have focused on the effects of PR on reproduction. In a flavobacterial strain containing PR, light-stimulated growth in seawater was observed [13]. However, similar experiments with other marine bacteria containing PR—Flavobacteria or members of the ubiquitous SAR11 or SAR92 clades—revealed no detectable effect of light on growth [10],[14],[15]. Nevertheless, Lami et al. [16] recently showed that PR expression in SAR11 and Flavobacteria was up-regulated in the presence of light and could be correlated with the abundance of PR genes. Taken together, these findings suggest that PR may involve fitness components other than growth. Thus, in this work, we explored the consequences of PR phototrophy for survival of marine bacteria.
Vibrio species are widespread marine bacteria and are frequently referred to as metabolically versatile heterotrophs [17]. Undoubtedly, the most well-known member of the genus is V. cholerae, the etiological agent of the disease cholera. Vibrios are typically found associated with detritus particles, algae or zooplankton, as commensals or pathogens on higher organisms, or as free-living populations in the water column [18],[19]. Nevertheless, the potential for phototrophy using PR or other light-harvesting mechanisms has not previously been reported for any member of the genus. We investigated the ecological response to light in a proteorhodopsin-containing member of the genus Vibrio. This representative of marine bacteria showed enhanced survival during starvation when exposed to light compared to darkness. Moreover, mutational analysis provided a direct link between the proteorhodopsin gene and the light response that conferred an increased ecological fitness.
A PR-encoding gene was identified from the whole-genome sequence of strain AND4, isolated from surface waters of the Andaman Sea. Phylogenetic analysis of the 16S rRNA gene as well as comparative genome analyses showed that AND4 is a member of the Gammaproteobacteria genus Vibrio (Figure 1 and Table 1). The AND4 PR shares a sequence similarity of 87% over 269 amino acid residues with the PR encoded in the publicly available genome sequence of V. harveyi strain BAA-1116. To our knowledge, the PR gene has as yet not been found in any other member of the genus Vibrio. PRs from the isolates AND4 and BAA-1116 both contain Leu in position 105, which fine tunes the PR light absorption peak towards green light (absorption maximum 535 nm, Figure S1), thereby adapting to the dominant light conditions prevailing in surface seawater [5],[20],[21]. All essential amino acid residues of the energy transducing rhodopsins are conserved (Figure S2), and the protein photocycle has a half-life of approximately 50 ms (Figure S1), as would be expected for a proton pump [22].
In AND4 and BAA-1116, the PR gene and the genes required for synthesis of the chromophore retinal, crtEIBY and blh [2],[23], were found at the same genetic locus (Figure 2). Phylogenetic analysis of the Vibrio PR amino acid sequences showed that, in contrast to the 16S rRNA gene placing AND4 and BAA-1116 among the Gammaproteobacteria, PRs in these bacteria clustered with PRs in Alphaproteobacteria (Figure 1). Moreover, the retinal biosynthesis genes have an ancestry that is divergent from the flanking genes (Table S1). This strongly suggests that the genes for PR and its chromophore have been acquired as a linked set of genes through lateral gene transfer from relatively distantly related bacteria, as has recently been suggested for other marine bacteria [2],. Lateral gene transfer may involve mobile genetic elements since transposase genes are found flanking the PR, crtEIBY, and blh genes in both BAA-1116 and AND4 (Figure 2). The transposase gene closest to the PR gene in AND4 was truncated and showed best matches to transposases in V. anguillarum 775, V. parahaemolyticus AQ3776 and V. cholerae 91, with percent similarities of 83%–87%. Several of the transposase genes in the genomes of AND4 and BAA-1116 are part of the IS903 subfamily of the IS5 family, which frequently are part of compound transposons in Vibrio species [26]. This indicates that AND4 and BAA-1116 share, or have shared, with other vibrios the mechanisms for lateral gene transfer. McCarren and DeLong [25] recently suggested that diverse marine bacteria may have acquired and retained the PR gene because it confers a competitive advantage in an otherwise resource-depleted surface ocean. However, there are no reported studies demonstrating how PR genes that have been putatively acquired through lateral gene transfer could impact on the life strategy of its carrier.
To explore this possibility, we investigated the growth and survival of AND4 in light and dark using a suite of approaches, where light refers to photosynthetically active radiation. Growth experiments with AND4 in rich medium showed no differences in cell yields for light (continuous light, 133 µmol photons m−2 s−1) and dark conditions (Figure 3A, inset). Also upon transfer of cells from rich medium to sterile and particle-free natural seawater with low concentrations of organic and inorganic nutrients, an increase of cell numbers within the first 2 d was observed (Figure 3A). Notably, epifluorescence microscopy images of AND4 cultures showed that most of the observed increase in cell numbers was due to reductive division rather than growth, i.e., cell numbers increased, but total biomass did not because each cell decreased in size (Figure 3B). This decrease in cell size is a well-described characteristic of vibrios (and many other bacteria) exposed to starvation, being an important strategy for optimizing cellular energetic efficiency when resources become limited [27]. After 10 d of incubation, bacterial numbers decreased in all cultures, but remained 2.5 times higher in the light compared to darkness. This finding strongly suggests that PR phototrophy can improve the survival of marine bacteria during periods of starvation in seawater.
Next, we monitored the development of optical densities (ODs) and bacterial numbers of AND4 grown in rich medium, washed, resuspended in sterile seawater, and exposed to four different light conditions (Figure 4A). In the dark, OD values steeply decreased during the first 13 d of starvation, whereas cultures exposed to light decreased more slowly, with the difference in ODs between the light treatments and darkness changing significantly over time (F = 5.81, df = 18, 24, p = 0.0079). After 7–13 d of starvation, OD values were 40%–60% higher for the treatments with continuous high light and with 16∶8 h light∶dark cycles (133 and 150 µmol photons m−2 s−1, respectively; corresponding to light intensities in oceanic upper mixed-surface layers), when compared with treatments maintained in the dark. Concomitantly, the effect of light in the low-light treatment (continuous light, at 6 µmol photons m−2 s−1; corresponding to light intensities at the lower limit of the photic zone) was less pronounced. In parallel with the initial decrease in OD, bacterial numbers peaked on day 3 in the high-light treatments, and thereafter decreased in all treatments. Nevertheless, bacterial numbers remained nearly twice as high for the high-light intensity treatments (Figure 4A, inset). These results again imply that PR can increase the survival rates of bacteria during starvation under light conditions corresponding to those found in the surface ocean.
To establish that the PR gene conveys this light-enhanced survival during starvation in AND4, we constructed a strain where the PR gene had been removed by an in-frame deletion of the near-complete PR gene (AND4 Δprd). In contrast with the wild-type behavior (Figures 4A, 5A, and 5B), starvation experiments with the Δprd strain showed no differences in ODs or bacterial numbers between light and dark conditions (Figures 4B, 6A, and 6B). However, when the mutant was complemented with the prd gene in trans, the wild-type phenotype was regained (Figure 4C; significant day-by-light treatment interaction; F = 13.21, df = 8, 24, p = 0.0083).
Moreover, we analyzed the ability of the wild-type and Δprd strains to recover growth during 5 h incubation in rich medium after increasing periods of starvation (Figures 5 and 6). Although little difference in growth recovery was observed after 1.5 d of starvation, after 5 d, the wild-type bacteria exposed to light during starvation grew to 3- to 6-fold higher densities when compared to bacteria starved in darkness (Figure 5). No differences in recovery were detected in the Δprd strain, irrespective of the history of light-exposed or dark incubations (Figure 6). These growth recovery experiments on wild-type and modified AND4 strains thus confirm that the increased rates of survival and the ability to actively respond to improved growth conditions are a direct consequence of having the light energy–harvesting potential of PR.
Genetic inventories of the world's ocean have revealed that the potential for harvesting light energy by means of PR phototrophy is found in a diverse variety of marine bacteria, encompassing organisms with very different life strategies and physiologies [14],[15],[28]. For example, members of the SAR11 clade are free-living bacteria with a range of cellular and physiological adaptations allowing them to minimize the consequences of starvation in oligotrophic waters. Their PRs are highly expressed, which may be a contributing factor to obtain positive net growth in seawater [14],[16]. For particle-attached bacteria or commensals/pathogens, PR phototrophy could also be important, but on a more irregular basis, during phases of starvation survival between particle or host colonization events. Irrespective of life strategy, the ability to survive starvation while maintaining the potential to proliferate is an essential trait for any evolutionarily successful organism [29].
Our results demonstrate that PR phototrophy represents a physiological mechanism that imparts an improved ability to survive when resources are scarce. This, thus, represents a substantial widening of the phototrophic properties known for marine bacteria in general, and vibrios in particular. Vibrio sp. AND4, studied here, is closely related to organisms that are known pathogens on higher organisms (e.g., V. harveyi BAA-1116 and V. parahaemolyticus) and requires nutrient-rich seawater for growth. AND4 and BAA-1116 are so far the only genome-sequenced members of the genus Vibrio that contain the PR gene. For BAA-1116, it is still unknown whether the gene is functional, although the high PR gene region synteny and PR amino acid sequence similarity may suggest that its function is similar to that in AND4. An important challenge for the future will be to unveil the physiological status and growth capacity of other PR-containing bacteria, both cultivated species, and major taxa in the marine environment, and to what extent PR phototrophy may alleviate starvation and/or contribute to other physiological processes in key species. Given the enhanced fitness observed in the present study, the acquisition and maintenance of the PR gene may be highly advantageous in the competitive marine environment, potentially influencing bacterioplankton community composition and population dynamics in the ocean's surface.
AND4 was isolated from surface water (2-m depth) in the Andaman Sea (7° 48′ 0″ N, 98° 12′ 36″ E) in December 1996 by spreading a 100-µl seawater aliquot on Marine Agar 2216 (Difco). After initial isolation and purification, the isolate was stored in glycerol (20% final concentration) at −70°C.
Whole-genome sequencing of AND4 was carried out by the J. Craig Venter Institute (JCVI) through the Gordon and Betty Moore Foundation initiative in Marine Microbiology. The draft genome sequence consists of 143 contigs representing ten scaffolds. The genome sequence was obtained using a Sanger/pyrosequencing hybrid method [30]. Our genome analysis is based on open reading frames predicted and annotated using JCVI's prokaryotic annotation pipeline (genome sequence available at https://moore.jcvi.org/moore/). All automatically annotated genes of interest were inspected and verified manually by BLAST, COG, PFAM, and TIGRFAM analyses.
The genome sequence (accession no. ABGR00000000; annotation added by the National Center for Biotechnology Information prokaryotic genomes automatic annotation pipeline group), 16S rRNA gene sequence (accession no. AF025960), and proteorhodopsin amino acid sequence (accession no. ZP_02194911) of Vibrio sp. AND4 are publicly available in the GenBank database.
For the phylogenetic tree of 16S rRNA genes shown in Figure 1A, a multiple alignment was generated using the software package ClustalW (Version 1.83). The alignment was edited with Gblocks (Version 0.91b) to identify conserved regions. The tree was constructed based on a Jukes-Cantor distance matrix and the Neighbor-Joining method using the PHYLIP package (Version 3.68). The sequence of Polaribacter sp. MED152 (DQ481463) served as outgroup. GenBank accession numbers are given in parentheses. The scale bar represents Jukes-Cantor distances (nucleotide substitutions per base position). Filling of circles at nodes represents matching topology with a maximum likelihood tree constructed with RAxML [31] and a neighbor-joining tree (default parameters) constructed with the ARB software package [32].
The proteorhodopsin amino acid sequence tree in Figure 1B was created from a multiple sequence alignment in ClustalW using the PHYLIP software package and the Kimura distance matrix and the neighbor-joining method. The alignment was edited with Gblocks (Version 0.91b) to identify conserved regions with a minimum block of five. The scale bar represents the Kimura distances (number of amino acid substitutions per base position). Circles at nodes represent matching topologies with a maximum likelihood tree constructed with RAxML [31] using the PROTGAMMABLOSUM62F amino acid model. The sequence of Polaribacter sp. MED152 (EAQ40925) served as the outgroup.
The number of transposase genes was obtained by BLASTP hits against the ISFinder dataset (http://www-is.biotoul.fr) and an E-value<10−10 and sequence identity values >35%.
To obtain measures of AND4 proteorhodopsin absorption maximum and photolysis rates, cell-free expression of AND4 prd was carried out by cloning from genomic DNA into the TOPO vector pEXP5NT (Invitrogen) using the sense and antisense primers 5′-ATGAAAAACCAAGTTGAAAAGATAACA-3′ and 5′-TTACGCATCCTGACTCTCGG-3′, respectively. This generated a prd construct with an N-terminal 6xhistidine tag followed by a Tev protease cleavage site preceding the AND4 coding sequence. prd was expressed with an in-house cell-free expression system based on Escherichia coli S12 extract, essentially according to a combination of protocols described in Kim et al. [33] and Torizawa et al. [34]. Expression was performed in batch format at 34°C in the presence of 0.1% Brij35 and 5 µg ml−1 all-trans retinal for 2 h at 800 rpm. The resulting reaction mix was centrifuged at 13,000×g for 5 min, and the supernatant was bound in batch mode to TALON resin (Clontech) pre-equilibrated with buffer A (phosphate-buffered saline containing 10 mM imidazole and 0.05% Brij35). After 1 h incubation at 4°C, the resin was transferred to a gravity-flow column, washed with 20 column volumes of buffer A, and eluted with buffer B (buffer A with 150 mM imidazole). The resulting eluate was passed through a PD10 column to change the buffer to phosphate-buffered saline containing 0.05% Brij35, and then concentrated with VivaSpin 6 ultrafiltration tubes with a MWCO of 30,000. For flash-photolysis experiments, the buffer was changed to either 25 mM CAPS (pH 10), 0.05% Brij35, or 25 mM MES (pH 5.5), 0.05% Brij35 by repeated dilution and concentration cycles in the ultrafiltration tube.
To create a null mutation in the PR gene, an in-frame deletion was made by allelic exchange using the suicide vector pDM4 as described by Milton et al. [35] with a few minor changes. Plasmid pDM4-rdp-AD, which carries a mutated allele of the PR gene that encodes the first seven amino acids fused to the last eight amino acids of the gene, was introduced into Vibrio sp. strain AND4 by conjugation. After the mating, selection for a Vibrio strain carrying the plasmid in the chromosome was done using Trypticase soy agar containing 1% sodium chloride, 200 µg ml−1 carbenicillin, and 15 µg ml−1 chloramphenicol. To complete the allelic exchange, direct selection on Trypticase soy agar containing 5% sucrose for a strain that had lost the sacB gene carried on pDM4-rdp-AD was done as described previously. The in-frame deletion was confirmed by sequencing a PCR-amplified DNA fragment of the deleted chromosomal locus. Primers used for the overlap PCR to create the mutated allele were as follows: PR-A-5′-GGACTAGTGGTTACTGGACACAA, PR-B-5′-AACCAAGTTGAAAAGGCAACCTCCGAGAGT, PR-C-5′-CTTTTCAACTTGGTTTTTCATAAT, and PR-D-5′-CTCGAGCTCCAGGGGAGATAGGTT.
To complement the deletion mutation, the wild-type prd gene was expressed in trans from the plasmid pMMB-prd-wt. To construct pMMB-prd-wt, the prd gene was amplified by PCR using KOD polymerase and the primers Rdp-5′-CTCGAGCTCCGTTAAAAGTGAGACTAT and Rdp-3′-CGCGGATCCTGGAAAGAGGGACAGAGA. An 890-bp fragment was gel purified, digested with SacI and BamHI, and ligated to pMMB207 [36], which was similarly digested. The resulting plasmid, pMMB-prd-wt, was mobilized into the Δprd mutant via conjugation.
For the rich-medium growth experiment (Figure 3, inset), ZoBell medium (5 g of peptone [Bacto Peptone; BD] and 1 g of yeast extract [Bacto Yeast Extract; Difco] in 800 ml of Skagerrak seawater and 200 ml MilliQ water) was 8-fold diluted in sterile-filtered and autoclaved Skagerrak seawater. Triplicate cultures were incubated at 16°C under an artificial light source of 133 µmol photons m−2 s−1. Triplicate dark control bottles were covered with aluminum foil. In all experiments, artificial light was provided by fluorescent lamps (L 36W/865, Lumilux, Osram) emitting photosynthetically active radiation, not the full spectrum of sunlight.
For the experiment with natural seawater (Figure 3), water was collected in the Skagerrak Sea and filter sterilized through 0.2 µm-pore-size membrane filters (Supor 200) and autoclaved. Each culture contained 250 ml of seawater (in 500-ml blue-cap glass bottles), and received a final concentration of 100, 2.1, and 0.3 µM dissolved organic carbon (in the form of ZoBell medium), nitrogen (NH4Cl), and phosphate (Na2HPO4), respectively. All material in contact with the samples was acid rinsed with 1 M HCl and extensively washed with MilliQ-water prior to use. Cultures were inoculated with AND4 bacteria previously grown overnight in rich medium (i.e., to early stationary phase). Duplicate light cultures were incubated at 16°C under an artificial light source of 133 µmol photons m−2 s−1, and duplicate dark controls were covered with aluminum foil.
For the starvation experiments, AND4 cells were grown overnight in 200 ml of rich medium (in 500-ml blue-cap glass bottles). Cells were harvested through centrifugation at 4,000 rpm for 10 min. Cell pellets were washed twice with sterile seawater (filter sterilized through 0.2 µm-pore-size membrane filters and autoclaved), resuspended in seawater, and distributed into Erlenmeyer flasks, with 75 ml of seawater–cells mix in each.
Experiments with the wild-type AND4 included duplicate flasks for the high-light intensity treatment at 133 µmol photons m−2 s−1, the 16∶8 h light∶dark cycle treatment at 150 µmol photons m−2 s−1, and the low-light treatment at 6 µmol photons m−2 s−1. Duplicate dark controls were completely covered with aluminum foil. The experiment was carried out at 16°C. Experiments with the Δprd strain and the Δprd strain complemented with the prd gene in trans included duplicate or triplicate flasks for each of the strains in the light (continuous light at 133 µmol photons m−2 s−1) and in dark controls. In a separate experiment, growth recovery experiments with the wild-type and Δprd AND4 strains were performed after 35, 131, and 203 h of starvation in light or darkness. Two hundred microliters of each starved culture were inoculated in 50-ml Falcon tubes that contained 25 ml of ZoBell medium based on Skagerrak seawater. Recovery cultures were incubated at room temperature in the dark and were monitored for 5 h.
Samples for optical density (OD) were measured at 600 nm using a bench top spectrophotometer (Beckman DU 640). Samples for bacterial numbers were fixed with 0.2 µm-pore-size filtered formaldehyde (4%, final concentration), stained with SYBR Gold (1∶100 dilution, Molecular Probes), filtered onto black 0.2 µm-pore-size polycarbonate filters (Poretics, Osmonics Inc.), and counted by epifluorescence microscopy within 48 h. Alternatively, samples for bacterial numbers were fixed with 0.2 µm-pore-size filtered formaldehyde (4%, final concentration), and stored frozen at −70°C until analysis by flow cytometry using a FACSCalibur flow cytometer after staining with Syto13 [37].
The effect of light on AND4 during starvation was analyzed by repeated-measures analysis of variance. Analyses were performed with PROC GLM in SAS 8.2, using type 3 sums of squares.
|
10.1371/journal.ppat.1003013 | Revised Phylogeny and Novel Horizontally Acquired Virulence Determinants of the Model Soft Rot Phytopathogen Pectobacterium wasabiae SCC3193 | Soft rot disease is economically one of the most devastating bacterial diseases affecting plants worldwide. In this study, we present novel insights into the phylogeny and virulence of the soft rot model Pectobacterium sp. SCC3193, which was isolated from a diseased potato stem in Finland in the early 1980s. Genomic approaches, including proteome and genome comparisons of all sequenced soft rot bacteria, revealed that SCC3193, previously included in the species Pectobacterium carotovorum, can now be more accurately classified as Pectobacterium wasabiae. Together with the recently revised phylogeny of a few P. carotovorum strains and an increasing number of studies on P. wasabiae, our work indicates that P. wasabiae has been unnoticed but present in potato fields worldwide. A combination of genomic approaches and in planta experiments identified features that separate SCC3193 and other P. wasabiae strains from the rest of soft rot bacteria, such as the absence of a type III secretion system that contributes to virulence of other soft rot species. Experimentally established virulence determinants include the putative transcriptional regulator SirB, two partially redundant type VI secretion systems and two horizontally acquired clusters (Vic1 and Vic2), which contain predicted virulence genes. Genome comparison also revealed other interesting traits that may be related to life in planta or other specific environmental conditions. These traits include a predicted benzoic acid/salicylic acid carboxyl methyltransferase of eukaryotic origin. The novelties found in this work indicate that soft rot bacteria have a reservoir of unknown traits that may be utilized in the poorly understood latent stage in planta. The genomic approaches and the comparison of the model strain SCC3193 to other sequenced Pectobacterium strains, including the type strain of P. wasabiae, provides a solid basis for further investigation of the virulence, distribution and phylogeny of soft rot bacteria and, potentially, other bacteria as well.
| The soft rot bacteria of the genus Pectobacterium are an economically important group of plant pathogens in the Enterobacteriaceae family. Pectobacteria are characterized by their massive production of plant cell wall-degrading enzymes, which are aggressively used to decay the living plant tissue for nutrient acquisition. Genome analysis of Pectobacterium sp. SCC3193, a genetically well characterized model in soft rot research, provided several surprises. Phylogenetic analysis demonstrated that SCC3193 belongs to the species Pectobacterium wasabiae, which has recently been reported from potato in increasing numbers worldwide. It is not clear whether P. wasabiae represents a true emerging pathogen in potato fields or whether traditional methods have been unable to differentiate between P. wasabiae and P. carotovorum. The genome of SCC3193 was found to harbor novel virulence determinants, some of which may have been acquired from distantly related bacterial species or even from other kingdoms, such as plants. The 25 years of SCC3193 research and the novelties found in this study will provide the basis for understanding the host pathogen interactions of P. wasabiae and improving strategies to combat bacterial soft rot.
| The soft rot bacteria Pectobacterium and Dickeya of the family Enterobacteriaceae are responsible for significant, global economic losses of crops and ornamental plants, both in the field and in storage. Although soft rot enterobacteria can infect a wide variety of plants, the main crop affected is potato (Solanum tuberosum L.). Potato is number five among food crops in the world and is a staple food in a number of countries (FAOSTAT 2009, http://faostat.fao.org/). Previously, the Pectobacterium and Dickeya species were classified into the Erwinia genus, and relatively recent phylogenetic analyses have elevated them to novel genera [1], [2]. In addition, some of the subspecies have been raised to the species level [3]. For historical reasons, three distinct potato diseases caused by soft rot enterobacteria are described: common soft rot (tuber symptoms), blackleg (tuber-born stem disease) and aerial stem rot (spread mechanically or via insects) [4]. Pectobacterium and Dickeya are characterized as opportunistic pathogens that switch from an asymptomatic latent phase into a virulent phase in suitable environmental conditions [5]. The virulent phase is thought to begin under anoxic conditions when oxygen radical-dependent plant defense mechanisms decline, allowing bacterial multiplication and the induction of plant cell wall-degrading enzymes (PCWDEs). The induction of PCWDEs occurs when the bacterial cell density reaches a quorum of 107 cfu/g of plant tissue [5].
The soft rot enterobacteria are necrotrophs and are generally considered to be brute-force pathogens relying on PCWDEs for pathogenicity. In fact, several regulatory mutants affecting enzyme production are essentially avirulent [6]–[9]. Studies of more fine-tuned virulence mechanisms may have been hampered by this massive production of PCWDEs [6]. Therefore, the latent stage preceding necrotrophy remains poorly understood. However, during the last 25 years, a number of other virulent lifestyle promoting determinants have been identified from Pectobacterium. These determinants include the following: flagella-based motility, cell membrane structures, such as enterobacterial common antigen (ECA) and lipopolysaccharide (LPS), type III secretion systems (T3SS), type IV secretion systems (T4SS), type VI secretion systems (T6SS), necrosis-inducing protein (Nip), a protein similar to an avirulence protein in Xanthomonas (Svx), coronafacic acid synthesis pathways (cfa genes), plant ferredoxin-like protein (FerE) and citrate uptake and 3-hydroxy-2-butanone pathways (bud) [10]–[19]. The virulence strategies of necrotrophic bacteria differ from hemibiotrophic and biotrophic phytopathogens, such as Erwinia amylovora, Pantoea sp. and Agrobacterium tumefaciens, which rely mainly on T3SS or T4SS for pathogenicity [20], [21]. The genomic era has provided novel tools for the identification of previously unknown virulence determinants without large-scale biological experiments. Several genomic studies of plant and animal pathogenic enterobacteria have been conducted, but only two have been published on soft rot bacteria. The first study compared P. atrosepticum and Salmonella, and the second study characterized P. atrosepticum, P. carotovorum and P. carotovorum subsp. brasiliensis [11], [22].
The soft rot pathogen Pectobacterium sp. SCC3193 was originally isolated from a diseased potato stem from a Finnish field in the early 1980s and was characterized as P. carotovorum (previously called Erwinia carotovora subsp. carotovora) [23]. Since its discovery, SCC3193 has been a model strain in soft rot research, and much is known about its virulence and molecular biology [7], [15], [17], [23]–[33]. In this work, we present a revised phylogeny, an analysis of virulence determinants and a characterization of genomic islands; we also discuss novel virulence strategies utilized throughout the lifestyle of Pectobacterium sp. SCC3193. We show that SCC3193 can be taxonomically classified as Pectobacterium wasabiae and that P. wasabiae has unique features when compared with other Pectobacterium strains; these features were most likely acquired via horizontal gene transfer. This work indicates that P. wasabiae has been present, though unnoticed, in European and maybe in American potato fields for a long time. Genome analysis is supplemented with experimental results to suggest novel virulence determinants of Pectobacterium; many of these determinants could be important during the poorly characterized latent stage of infection.
The species status of Pectobacterium sp. SCC3193 was questioned after an initial review of the recently sequenced genome of SCC3193 by our Pectobacterium sequencing consortium in Helsinki, Finland (CP003415, Koskinen et al., in press). We discovered distinctive sequence similarity of SCC3193 to the strain WPP163 (NC_013421.1), which was sequenced in 2009 by another group (Nicole Perna and coworkers, and US DOE Joint Genome Institute; unpublished). WPP163 was isolated from potato stem and classified as P. wasabiae prior to the genome sequencing [34]. Originally, Pectobacterium sp. SCC3193 was identified as Pectobacterium carotovorum based on disease symptoms in potato, the ability to produce PCWDEs, fatty acid composition and other biochemical properties. Subsequent studies suggested that SCC3193 may not be a typical P. carotovorum strain due to its LPS composition, sensitivity to T4 phage, decreased ability to macerate plant tissues and inability to grow at +37°C [23]. However, P. carotovorum has been recognized as a highly variable species that is composed of several subspecies. Thus, differences in phenotype have been accepted [3]. At the time of the isolation and characterization of SCC3193, the species Pectobacterium wasabiae (previously called Erwinia carotovora subsp. wasabiae) had not yet been described. The type strain (CFBP 3304T) of P. wasabiae was isolated from wasabi (Japanese horseradish) in 1987 [3], [35]. Reports of P. wasabiae isolates are limited when compared to the number of P. carotovorum and P. atrosepticum reports [35]–[40]. To evaluate the species status of SCC3193, we conducted a thorough phylogenetic analysis of SCC3193, which included biochemical and genomic methods. The genome of P. wasabiae CFPB 3304T was sequenced by our consortium to be used as a reference in this phylogenetic analysis.
To obtain an overall view of SCC3193 virulence factors, we mined the genome for known virulence determinants of Pectobacterium. Because we are working with an established model strain, many virulence-associated determinants of SCC3193 have already been identified in genetic studies, and some of these were even originally described in SCC3193. This may create a bias that is reflected in the number of published virulence-associated genes found in the genome of SCC3193.
Only a few previously described virulence-related genes in Pectobacterium were not found in SCC3193: namely, the genes for coronafacic acid synthesis (specific to P. atrosepticum) and the type III secretion system (T3SS) present in many P. atrosepticum and P. carotovorum strains [11], [13]. The T3SS is composed of the injection machinery encoded by the conserved hrp/hrc gene cluster and of a species/strain-specific collection of effectors required to suppress the basal defenses of the host [55]. Analysis of the SCC3193, P. wasabiae WPP163 and the type strain (CFBP 3304T) genomes failed to identify any traces of T3SS. It appears typical of P. wasabiae that it lacks this widespread and important virulence determinant. These results are in agreement with previous failed efforts to detect T3SS in SCC3193 or other P. wasabiae strains; for example, no signs of this injection machinery or the associated effectors (encoded by hrpN, dspE and hecB genes) have been found [34], [36], [56]. Although absent from P. wasabiae, T3SS contributes to the virulence of other soft rot species [57]–[59].
Contrary to hemibiotrophic pathogens such as Pseudomonas syringae, where T3SS is essential for pathogenicity [21], the role of T3SS in necrotrophic soft rot bacteria appears quite complex. T3SS contributes to the virulence of P. carotovorum, P. atrosepticum and Dickeya strains, but even strains naturally lacking T3SS, such as P. wasabiae, are still able to infect potatoes. A recent study showed no clear correlation between virulence and the presence of T3SS in P. carotovorum [34]. The relatively modest significance of T3SS to pathogenicity in pectobacteria is also reflected in the small number of T3 effectors found in these species compared, for example, to the dozens of known effectors of Pseudomonas syringae [60]. Thus, Pectobacterium may have alternative ways of modifying the host plant at the initiation of infection.
The lack of T3SS and numerous effectors may also benefit Pectobacterium by widening its host range. Soft rot bacteria, excluding the potato-specific P. atrosepticum, are often found to colonize a wide range of food crops and ornamental plants, while many T3SS-dependent plant pathogens, such as Erwinia, Pseudomonas, Xanthomonas and Xylella, have a very narrow host range. It is well documented that some T3 effectors may act as colonization-inhibiting avirulence proteins that are recognized by the plant [55]. Virulence assays to investigate the role of T3SS in soft rot bacteria in planta thus far indicate that in a complex natural niche, T3SS may have an important role under certain conditions but not in others.
Horizontal gene transfer plays an important role in the evolution of bacteria. It enables the rapid acquisition of beneficial traits in a single event. These gene clusters of probable horizontal origin are termed genomic islands (GIs/GEIs) or horizontally acquired islands (HAIs) [11], [61], [62]. To characterize the genome composition of SCC3193 and identify horizontally acquired virulence determinants or other adaptive traits, we determined putative GIs in the genome of SCC3193. We used two sequence composition-based GI prediction methods, SIGI-HMM and IslandPath-DIMOB [63], [64], which were found to have the highest overall accuracy of the six methods tested in a recent bioinformatic study [65]. In addition, we used one comparative genomic-based GI prediction method, IslandPick [65]. The three methods predicted a different number of islands and smaller islets: six for IslandPath-DIMOB, ten for IslandPick and over hundred for SIGI-HMM (63 consisted of five or more successive ORFs) (Figure 4). Automated predictions were subsequently manually curated, resulting in a total of 56 genomic islands (Table S2). The GIs comprise ∼0.86 Mb, which is 16.7% of the size of the genome and encompasses 21.6% of all ORFs (1040 of the 4804 ORFs). In comparison, 17.6% of the ORFs of Escherichia coli MG1655 were estimated to have been acquired horizontally [66]. In general, GIs are estimated to comprise between 1.6% and 32.6% of the ORFs in prokaryotic genomes [67].
A majority of the SCC3193 islands can be found identically or with slight permutations in P. wasabiae WPP163, whereas less than half of the islands are present in the genome of P. wasabiae CFBP 3304T (Table S2). Approximately 15 islands were specific to P. wasabiae among the soft rot group. Some Pectobacterium-specific islands (for example, GI_44) or islands partially present in other Pectobacterium strains but completely lacking from Dickeya (for example GI_13 and GI_36) were also discovered. A number of islands were specific to SCC3193 and could not be found in any other strains of soft rot bacteria (for example, GI_20, GI_40, GI_43 and GI_54). Furthermore, a few islands showed a significant similarity to the genomes of bacteria outside the soft rot species (GI_9, GI_10, GI_30, GI_40, GI_44 and GI_50). Functional predictions for the genes on the islands in SCC3193 suggested the presence of a high number of mobile element genes and genes with unknown functions. The islands were also found to contain several genes for known virulence determinants, such as Nip (GI_17), which is necessary for the full virulence of SCC3193 [15], and DsbA (GI_56), which is required for the correct conformation of many secreted virulence proteins in P. atrosepticum and P. carotovorum [68], [69]. Most of the islands were shown to carry genes with predicted functions that could often be potentially associated with plant colonization or virulence (Table 1, Table S2, Figure 4).
The characterization of the SCC3193 genome revealed novel genes that could contribute to the virulent lifestyle. We selected a group of interesting potential virulence determinants for experimental verification. The corresponding genes were inactivated by targeted mutagenesis, and their contribution to virulence was tested on axenic tobacco seedlings and potato tuber slices. A previously characterized phytase gene in P. wasabiae CFBP 3304T (Y17_1078) [106] is also present in P. wasabiae SCC3193 (W5S_4347) and P. wasabiae WPP163 (Pecwa_4189). In our virulence assays the phenotype of an SCC3193 phytase gene knock-out mutant was inconclusive and its potential role in virulence would require additional studies. However, the phytase gene is not unique to P. wasabiae as it can be found from all sequenced Pectobacterium strains and many other plant associated bacteria. The mutants that had phenotypes in planta were tested for their ability to grow in vitro and produce polygalacturonases and cellulases. No differences were observed between the mutants and the wild- type strain (data not shown), indicating that the phenotypes in planta are due to plant-microbe interactions and not general growth defects or a major reduction in PCWDE production.
The model strain Pectobacterium sp. SCC3193 has been one of the most intensively studied soft rot strains for over two decades. Therefore, the molecular-level information on its virulence has had a significant impact on the theory of virulence in Pectobacterium. However, SCC3193 was originally incorrectly identified as P. carotovorum at the time of isolation, and our extensive phylogenetic analysis in this report reveals that it belongs to the species P. wasabiae, which is a less characterized species in the soft rot group. It is likely that other Pectobacterium isolates could also be incorrectly classified, and P. wasabiae may be more common than previously thought. Our report will facilitate further studies of the distribution and molecular biology of P. wasabiae, as a well-studied model strain has been added to the species, and three genome sequences are now publicly available, including the Japanese type strain sequenced in this study.
The absence of the T3SS and T3 effectors is the most distinctive feature separating Pectobacterium wasabiae SCC3193 and other P. wasabiae strains from other soft rot bacteria, other plant pathogens and other animal pathogenic enterobacteria. However, the number of confirmed T3 effectors is limited in soft rot bacteria overall, compared with the dozens in hemibiotrophs, and the effect of T3SS on virulence is not central in Pectobacterium. Our genomic approach, which was supplemented with in planta experiments, revealed putative ways for Pectobacterium wasabiae and other pectobacteria to establish infection. Promising candidates for novel virulence determinants having an effect at the early stage of infection are Virulence cluster 2, which carries genes for a putative novel lipoprotein transport system; HopL1, which is an adjacent T3SS effector-like protein; and T6SS. The latter has similar features to T3SS, which injects effectors into the host cell. T6SS is found in a variety of bacterial species and is not confined to pathogens. On the nitrogen-fixing plant symbiont Rhizobium leguminosarum, T6SS is related to host specificity [121]. The P. wasabiae benzoic acid/salicylic acid methyltransferase may also represent a novel way to manipulate the host in the latent stage, but we have not yet found experimental evidence to support this hypothesis.
In the future, to learn more about the latent stage of Pectobacterium infection, we should also consider bacteria outside the pathogens relying on T3SS. Many endophytic bacteria lack T3SS, T4SS and/or the production of pectinolytic enzymes common in phytopathogens [122]–[125]. They are known to use a diverse set of attachment structures and flagella-based motility to colonize plants and to produce plant hormones and other compounds for the modification of plant metabolism. Altogether, endophytic bacteria and symbionts, which efficiently establish interactions with plants, could be viewed as an opportunity to learn more about colonization mechanisms, which may also be important for the pathogenic lifestyle of soft rot bacteria during the latent stage of infection.
In this work, we used Pectobacterium wasabiae SCC3193, its derivatives and Pectobacterium wasabiae CFBP 3304T for the sequencing and/or biological experiments (Table S3). Standard growth conditions included culturing bacteria on Luria Broth (L3522, Sigma-Aldrich, US) for 1 d at +28°C. The antibiotics chloramphenicol (20 µg/mL) or ampicillin (100 or 150 µg/mL) were added when appropriate.
Genomic DNA of Pectobacterium wasabiae CFBP 3304T was extracted from an overnight culture using phenol-ether purification and ethanol precipitation. The quality and quantity of the DNA was assessed using spectrophotometry and agarose gel electrophoresis. The DNA concentration used for the hybridization step was 8 pM. Template amplification/cluster generation was performed using the Cluster station and the Single-Read Cluster generation kit v. 4 (Catalog # GD-103-4001). The sequencing was performed with an Illumina Genome Analyzer IIx using a v. 5 sequencing kit (FC-104-5001). All operations were performed following the manufacturer's protocols (Cluster Station User Guide, Part # 15005236 Rev. B, November 2009, Sequencing Kit v5 Reagent Preparation Guide, Part # 15013595 Rev. A, May 2010). Data generated by Illumina Solexa GAIIx were analyzed with SCS-RTA v. 2.9. Matrix and phasing parameters that were estimated from the PhiX control were used for base-calling. The demultiplexing and conversion of bcl-files to FASTQ-files were performed using OLB v. 1.9.0 and CASAVA GERALD v. 1.7.0. The instrument vendor provided all the software. Adapter sequences were clipped from the reads using cutadapt-tool. Additionally, if an adapter was removed from one read, the other read was shortened to reflect the change when necessary. The read data were assembled with ABySS using the following parameters: −j = 2 k = 60 n = 10 q = 15 ABYSS_OPTIONS = ‘–illumina-quality -c 25 -e 25’.
This Whole Genome Shotgun project has been deposited at DDBJ/EMBL/GenBank under the accession AKVS00000000. The version described in this paper is the first version, AKVS01000000.
The ORF prediction was completed for CFBP 3304T using the Prodigal gene prediction program [126]. Systematic errors made by Prodigal were corrected, and the intergenic areas were double checked for missed gene predictions with the GenePRIMP program [127] and manual correction. Predicted and corrected protein sequences were then functionally annotated with descriptions (DE), Gene Ontologies (GO) and Enzyme Commission Numbers (EC) using the PANNZER tool (Koskinen et al., in preparation; method is unpublished). The COGnitor tool [128] was used to link sequences into COG database clusters. Finally, functional elements (for example, domains) were searched using the InterProScan tool [129].
The annotated ORFs of SCC3193 (Koskinen et al., submitted) were grouped into hypothetical operons using the ofs1.2 software [130]. The grouping was based on link probabilities, which represent the probability of a given ORF to be co-expressed with a downstream gene (see Dataset S2). Based on an E. coli benchmark (RegulonDB 6.7 release [131]), our estimates suggest that thresholding these link probabilities at 0.54 will recall approximately 83% of true operon links with 81% specificity. The OFS operon prediction is based on intergenic distances, similarities between annotations and the conservative clustering of similar genes in other species. The set of “other species” (referred to as informant species) used in this analysis was compiled using a single representative of each bacterial order enlisted in the NCBI bacteria taxonomy. At the time of compilation, there were 119 orders with at least one full genome sequenced; therefore, the list contains 119 entries (Dataset S3).
In addition to the predicted protein sequences of Pectobacterium wasabiae SCC3193 and Pectobacterium wasabiae CFBP 3304T, full RefSeq proteomes were fetched from NCBI for 52 additional species. One-to-one orthologs of SCC3193 proteins were determined using the RBH (reciprocal best hit) criterion. Protein X.G1 from proteome G1 and protein X.G2 from proteome G2 are reciprocal best hits, if there X.G2 is the best match of X.G1 in proteome G2 and X.G1 is the best match of X.G2 in proteome G1. The full proteomes of SCC3193 and 53 target species were compared using SANS [41] with window size of 100. For phylogenetic analysis, we selected 51 groups of orthologs of SCC3193 proteins present in each of the 53 other species. The ortholog groups were then aligned using Muscle v. 3.8.31 [132]. For all multiple alignments, 1000 bootstrap trees were created using RAxML v. 7.0.4 [133]. The settings used for RAxML were “raxmlHPC -m PROTGAMMAJTT -c 4 -f d -n %s -s %s -x 137 -N 1000”, where %s was replaced with the correct input and corresponding output file names. All the bootstrapped trees were merged using the Consense program v. 3.68 from the Phylip package [134]. The merged trees were visualized using iTOL webtools [135].
Biochemical tests for the ability to reduce sugars, phosphatase activity, indole production, growth on sorbitol, growth on melibiose, growth on raffinose, growth on lactose, utilization of keto-methyl glucoside, growth in 5% NaCl and growth at +37°C were conducted for P. wasabiae SCC3193 and the type strains P. wasabiae CFBP 3304T, P. carotovorum CFBP 2046T and P. atrosepticum HAMBI 1429T according to the previously described protocols [136], [137].
In the proteome comparison, we re-identified all the ORFs for the sequenced Pectobacterium and Dickeya species and for the outgroup Yersinia pestis CO92. This was performed using Prodigal for gene prediction. The proteomes were then aligned using the blastp program and clustered into orthologous groups with OrthoMCL program. OrthoMCL clusters were converted into an orthologs vs. species (OvsS) matrix. The obtained OvsS matrix can be altered for ease of interpretation by ordering similar columns and rows next to each other, which was accomplished by creating a hierarchical cluster tree from the rows and columns. The internal nodes of the tree were flipped to place more similar neighboring clusters next to each other [138]. The similarity of the columns was based on the Pearson correlation, while the similarity of the rows was based on the cosine similarity. These measures were selected by reviewing the obtained visualization from the different similarity measures. The ordered matrix was then visualized on a heat map. A distance matrix was created by calculating the Pearson correlation similarity scores between species in the OvsS matrix and then visualized as a heat map.
Pectobacterium wasabiae CFBP 3304T contigs were ordered by aligning them against a reference genome, Pectobacterium wasabiae SCC3193. This alignment was created with Mauve v.2.3.1 [139]. The reordered Pectobacterium wasabiae CFBP 3304T contigs were then aligned against the genomes of Pectobacterium wasabiae WPP163 and Pectobacterium atrosepticum SCRI1043. Mauve produced a genome content distance matrix as an output from the pairwise alignments, which was used to quantify the differences in the aligned sequences.
Genomic island predictions were performed using three computational tools: two that utilize sequence composition-based GI prediction methods, SIGI-HMM from the Colombo package [63] and IslandPath-DIMOB [64], and one that is based on a comparative genomic-based GI prediction, IslandPick [65]. Automated predictions were manually curated. The borders were adjusted such that genes known to be frequently associated with GIs or mobile genetic elements, such as integrase and phage genes, were added when these genes were located adjacent to automatically predicted islands. In addition, when necessary, the proximity of tRNA genes, similar functions encoded adjacent to the predicted island and the absence of the region in closely related strains based on blastn were used as criteria to modify the islands. Finally, we filtered out all islands consisting of less than five ORFs. To determine whether the SCC3193 genomic islands are also present in the genomes of other soft rot bacteria, blastn searches were performed with the nucleotide sequences of the islands against all available Pectobacterium and Dickeya genomes in GenBank. For P. carotovorum WPP14 and P. carotovorum subsp. brasiliensis PBR1692, whose genomes have not been completed, the combined query coverage of all contigs for each island was estimated based on a WGS-blastn search against these genomes. For P. wasabiae CFBP 3304T, the estimation was based on blastn pairwise alignment.
The comparison of orthologous proteins of Pectobacterium wasabiae SCC3193 was conducted against 41 selected soft rot bacteria, plant pathogens, animal pathogens and insect pathogens (Dataset S4) to identify novel virulence genes in SCC3193, in P. wasabiae or in soft rot bacteria. The comparison was performed manually, based on clusters created with OrthoMCL [140]. The genome and specific proteins identified were analyzed using the genome viewer Argo v. 1.0.31 [141], operon predictions, tblastn via Embster 2.0 beta launched from the CSC Chipster platform (http://chipster.csc.fi/embster/), annotation, blastn and blastp in GenBank NCBI [142], [143]. These analyses were conducted to identify known virulence determinants, detect missing virulence determinants and identify novel putative virulence determinants. The analysis of PCWDEs was performed by utilizing publications, sequence similarity and selected GO terms (Table S4) known to be associated with the enzymes of interest. The enzymes with the selected GO terms were mined using PANNZER (Koskinen et al., in preparation; method is unpublished), InterProScan [129] or BioMart [144]. The proteinases present in E. coli were discarded according to Glasner et al. [22], and they were not considered as putative enzymes targeting plants.
The inactivation of individual genes and gene clusters was performed by deleting target sequences and replacing them with an antibiotic cassette, according to Datsenko and Wanner [145]. The antibiotic cassette was amplified from a pKD3 template plasmid with primers carrying 50 bp of similar sequence to the genomic DNA of Pectobacterium wasabiae SCC3193 (Table S5). The cloning was conducted using the proofreading PCR enzyme Phusion according to the manufacturer's 3-step protocol (Finnzymes). The insert was gel purified. For the electrocompetent cells, bacteria were grown overnight, diluted 1∶50 and grown to OD600 0.4. The cells were cooled down and washed twice with sterile ice-cold ddH2O and once with sterile ice-cold 10% glycerol. The cells were resuspended in 1.5–2x volume of sterile ice-cold 10% glycerol. Electroporation was performed with the following settings: 2.5 kV, 25 µF and 200 Ω in 0.2 cm cuvettes (Bio-Rad Laboratories). The recovery times for pKD46 and the antibiotic cassette insertion were 15 min and 3.5 h, respectively. The mutations and the position of the insert were confirmed with two PCR reactions according to Datsenko and Wanner [145] in addition to sequencing (Table S5). For the double mutant of T6SS, the first antibiotic cassette was digested using flipase produced in trans from the pFLP2 plasmid [146].
For the complementation experiments, the genes of the sirB locus were amplified by PCR from wild-type SCC3193 genomic DNA using the proofreading PCR enzyme Phusion (Finnzymes). The following primers were utilized: SirB1_compl_F and SirB1_compl_R for sirB1; SirB2_compl_F and SirB2_compl_R for sirB2; and SirB2_compl_F and SirB1_compl_R for the complete sirB locus (Table S5). The PCR products were gel purified, digested with HindIII and SacI and ligated into pMW119 (Nippon Gene), which was digested with the corresponding enzymes. The constructs were confirmed via PCR and sequencing.
The bacterial virulence was tested on axenic tobacco seedlings (Nicotiana tabacum cv. ‘Samsun’) according to Pirhonen et al. [24]. The seedlings were propagated in 24-well tissue culture plates on ½ MS medium supplemented with vitamins (Duchefa) and 2% sucrose and solidified with 0.8% agar. The seedlings were grown for 16 days with a 16/8 h day/night cycle at 26/22°C. The bacterial strains were grown overnight, washed and diluted in 10 mM MgSO4. A single leaf from each of 48 plants was wounded with a needle and inoculated with 1.5 µl of bacterial solution (OD600 0.05). The inoculated plants were kept in the dark at room temperature, and the development of disease symptoms (leaf maceration) was scored visually at 24 and 48 h after inoculation on a scale of 0 to 6 according to the severity of symptoms. To determine bacterial growth in planta (cfu/plant), the inoculated plants were kept in the dark at 26°C and homogenized into 10 mM MgSO4 after 24 and 44 h; serial dilutions of bacteria were plated. All experiments were repeated three times, and the data of each replicate were analyzed statistically utilizing a Mann-Whitney significance test for the pairwise comparison of two independent samples using PASW Statistics 18.
For the tuber slice assay [17], bacterial strains were grown overnight, washed and resuspended into 10 mM MgSO4. Tubers (cv. Van Gogh, H&H Tuominen, Finland) were washed with tap water and surface sterilized with 10% Na-hypochlorite for 5 min. The tubers were sliced 0.6 cm thick, surface sterilized again with flaming and placed on wet paper tissue in a Petri dish. Ten to 12 tuber slices were inoculated with 10 µl of bacterial solution containing 106 cfu/ml and incubated at room temperature in a shady place for three days. The macerated area was assessed, and the results were classified into seven groups: 0%, ∼5%, ∼20%, ∼50%, ∼75%, ∼90% and 100% rotten tissue per tuber slice surface area. The experiments were repeated five times, and data from all the replicates were combined for statistical analyses utilizing the Mann-Whitney significance test for the pairwise comparison of two independent samples using PASW Statistics 18.
Cellulase (Cel) and polygalacturonase (PehA) activities were assayed from 10 µl of supernatant of liquid cultures grown overnight and from appropriate dilutions of the supernatant on corresponding enzyme indicator plates [7]. The growth of the sirB, Vic1, Vic2 and T6SS-double mutants were compared with the SCC3193 wild-type strain in an hrp-inducing minimal medium [147] using 0.4% polygalacturonic acid (Sigma P3850) as a sole carbon source. The bacteria were washed once with 1x minimal medium, diluted to OD600 0.1 and then grown for 20 h at +28°C with shaking (200 rpm).
|
10.1371/journal.ppat.1004551 | Hypercytotoxicity and Rapid Loss of NKp44+ Innate Lymphoid Cells during Acute SIV Infection | HIV/SIV infections break down the integrity of the gastrointestinal mucosa and lead to chronic immune activation and associated disease progression. Innate lymphoid cells (ILCs), distinguishable by high expression of NKp44 and RORγt, play key roles in mucosal defense and homeostasis, but are depleted from gastrointestinal (GI) tract large bowel during chronic SIV infection. However, less is known about the kinetics of ILC loss, or if it occurs systemically. In acute SIV infection, we found a massive, up to 8-fold, loss of NKp44+ILCs in all mucosae as early as day 6 post-infection, which was sustained through chronic disease. Interestingly, no loss of ILCs was observed in mucosa-draining lymph nodes. In contrast, classical NK cells were not depleted either from gut or draining lymph nodes. Both ILCs and NK cells exhibited significantly increased levels of apoptosis as measured by increased Annexin-V expression, but while classical NK cells also showed increased proliferation, ILCs did not. Interestingly, ILCs, which are normally noncytolytic, dramatically upregulated cytotoxic functions in acute and chronic infection and acquired a polyfunctional phenotype secreting IFN-γ, MIP1-β, and TNF-α, but decreased production of the prototypical cytokine, IL-17. Classical NK cells had less dramatic functional change, but upregulated perforin expression and increased cytotoxic potential. Finally, we show that numerical and functional loss of ILCs was due to increased apoptosis and ROR γt suppression induced by inflammatory cytokines in the gut milieu. Herein we demonstrate the first evidence for acute, systemic, and permanent loss of mucosal ILCs during SIV infection associated with reduction of IL-17. The massive reduction of ILCs involves apoptosis without compensatory de novo development/proliferation, but the full mechanism of depletion and the impact of functional change so early in infection remain unclear.
| HIV-1 has long been shown to deplete CD4+ T cells and disrupt barrier integrity in the gastrointestinal tract, but effects on other subpopulations of lymphocytes are less well described. A recently identified subpopulation of mucosa-restricted cells, termed innate lymphoid cells (ILCs) is thought to play critical roles in maintaining homeostasis in the gastrointestinal tract and mucosal pathogen defense. Although previous work from our laboratory and others have shown SIV infection of rhesus macaques can deplete ILCs in some parts of the gastrointestinal tract, systemic as well as kinetic effects were unclear. In this report we show that ILCs, but not classical NK cells are systemically depleted during infection and also acquire cytotoxic capabilities. Furthermore, our data is the first to indicate that this important subset of innate cells is depleted acutely, permanently, and systemically during SIV infection of rhesus macaques as a model for HIV-1 infection. Given the important role of ILCs in maintaining gut homeostasis these findings could have significant implications for the understanding and treatment of HIV-induced disease.
| During acute infection, the gastrointestinal (GI) tract is a primary target site for HIV-1 and SIV replication [1]–[4]. CD4+T cells are rapidly infected and depleted and the mucosal epithelial barrier is compromised. These early events after infection generally set the pace of disease progression, and while subsequent microbial translocation and immune activation drive ongoing disease, the early events in the mucosae following infection remain incompletely understood [2], [3], [5]–[7].
A growing number of reports indicate that innate lymphoid cells (ILCs) play critical roles in maintaining mucosal epithelial integrity, tissue remodeling and repair, and defense against intestinal pathogens [8]–[12]. ILCs are a heterogeneous group of the lymphoid lineage, but depend on the helix-loop-helix transcription factor inhibitor of DNA binding 2 (Id2), the common γ-chain receptor and IL-7 for their development [13]–[17]. ILCs are divided into three groups in mice and humans, based on their expression of cell surface markers, functional characteristics and transcriptional regulation. Group 1 ILCs (ILC1) contain natural killer (NK) cells, which are cytotoxic, produce IFN-γ and depend on T-bet for their development; group 2 ILCs (ILC2) are innate IL-5- and IL-13-producing cells and depend on transcription factor GATA-3 for lineage commitment; group 3 ILCs (ILC3) produce IL-22 and/or IL-17 and depend on RORγt for development [18]–[22]. Interestingly, development of both ILC1 and ILC3 require IL-7, but additive IL-β drives differentiation to ILC3. In contrast, addition of IL-12, IL-15, or IL-18 in combination with IL-7 drives differentiation toward ILC1. Although the general features of ILCs are conserved in mice and humans, no specific uniform nomenclature for ILCs has been ascribed in rhesus macaques, due to a lack of identification of each lineage. Previously, we identified NKp44+ILCs from rhesus macaques and found them to be restricted to mucosal tissues, express high levels of RORγt, and produce IL-17 [23], making them most likely analogous to ILC3. Furthermore, during chronic SIV infection, others and we have shown that NKp44+ILCs are reduced in the GI tract and IL-17 production is suppressed [23]–[27]. However, these studies were performed primarily in limited tissues and in chronically SIV-infected animals. The systemic effects of SIV infection and kinetics of loss are unknown.
The role(s) of NK cells in HIV pathogenesis and disease remains controversial. Studies on highly exposed individuals who remain seronegative (HESN), including intravenous drug users and heterosexual partners of HIV-positive individuals, suggest that increased NK cell activity may be associated with resistance to HIV infection [28]–[32]. Epidemiological and genetic studies have also shown that expression of immunoglobulin like receptor KIR3DS1 on NK cells, and its ligand HLA-Bw4-80I, are associated with slower disease progression [33], [34]. Furthermore, NK cells expressing KIR3DS1 can strongly inhibit HIV-1 replication in vitro [34]. However, contradictory reports show that NK cell activity may have no effect on controlling virus replication [35]. Recently, another report compared the relative importance of NK cells, CD8+T cells, B cells and target cell limitation in controlling acute SIV infection in rhesus macaques and suggested that NK cells have little impact on the death rate of infected CD4+ cells and that their net impact may increase viral load [36]. However, these studies typically use samples collected from peripheral blood or lymph nodes and overlook the roles of mucosae-resident NK cells in limiting HIV replication during the initial stages of infection. In this study, we used the rhesus macaque model to evaluate the quantitative and qualitative effects of acute and chronic SIV infection on mucosae resident NKp44+ILCs and NK cells.
All animals were housed at the New England Primate Research Center of Harvard Medical School in accordance with the rules and regulations of the Committee on the Care and Use of Laboratory Animal Resources. Animals were fed standard monkey chow diet supplemented daily with fruit and vegetables and water ad libitum. Social enrichment was delivered and overseen by veterinary staff and overall animal health was monitored daily. Animals showing significant signs of weight loss, disease or distress were evaluated clinically and then provided dietary supplementation, analgesics and/or therapeutics as necessary. Animals were humanely euthanized using an overdose of barbiturates according to the guidelines of the American Veterinary Medical Association. All studies reported here were performed under IACUC protocol #04637 which was reviewed and approved by the Harvard University IACUC.
A total of twenty-six Indian rhesus macaques were analyzed in this study, including six SIV-naïve, twelve acutely and eight chronically infected with SIVmac239. For some tissues (i.e., blood, colorectal biopsy tissue), pre-infection data are grouped with naïve samples for cross-sectional comparisons. Most animals were infected intravenously except for 6 animals sacrificed at days 6 or 7, which were infected intravaginally. Infection was verified by plasma virus quantification by RT-PCR (see Fig. 1A) or by immunohistochemistry to SIV antigens (used in day-6/7 infection sacrifices, to be published elsewhere). Chronically SIV-infected macaques were infected between 162 and 707 days, with a median duration of 308 days. Chronic viral loads at time of necropsy were between 4.3 and 6.2 log10 copies of viral RNA/ml, plasma, with a median of 5.1 log10 copies of viral RNA/ml, plasma. CD4+ T cell frequencies in blood were between 54% and 22% of T cells, with a median of 39%. All animals were free of simian retrovirus type D and simian T-lymphotropic virus type 1, and were housed at the New England Primate Research Center or at the National Cancer Institute, National Institutes of Health. All animals were housed and cared for in accordance with the American Association for Accreditation of Laboratory Animal Care standards. All animal procedures were performed according to protocols approved by the Institutional Animal Care and Use Committee of Harvard Medical School and the National Institute of Allergy and Infectious Diseases, National Institutes of Health.
Macaques were humanely euthanized at indicated time points and tissues were collected from colon, jejunum, mesenteric lymph node (MLN) and pararectal/paracolonic lymph node (PaLN). In a sub-group of the acutely infected animals, pre-infection (day -14) lymph node and colorectal biopsies were taken. Mucosal tissues were collected and lymphocytes isolated by mechanical and enzymatic disruption as described previously [23], [37], [38]. Total peripheral blood mononuclear cells were isolated from EDTA-treated venous blood by density gradient centrifugation over lymphocyte separation media (MP Biomedicals, Solon, OH) and a hypotonic ammonium chloride solution was used to lyse contaminating red blood cells.
Antibodies to the following antigens were included in this study and except where noted, all were obtained from BD Biosciences: α4β7-APC (clone A4B7, NHP reagent resource), active-caspase-3-Alexa647 (clone C92-605), CCR7-Alexa700 (clone 150503, R&D Systems), CD3-APC-Cy7 (clone SP34.2), CD4-FITC (clone L-200), CD16-Alexa-700 (clone 3G8), CD45-FITC (clone D058-1283), CD45-PerCp-Cy5.5 (clone Tu116), CD56-PE-Cy7 (clone NCAM16.2), CD62L-FITC (clone SK11), CXCR3-PE-Cy5 (clone 1C6), HLA-DR-PE-Texas Red (clone Immu-357, Beckman-Coulter), NKG2A-PE (clone Z199, Beckman-Coulter), NKG2A-Pacific Blue (clone Z199, in-house custom conjugate, Beckman-Coulter), NKp44-PerCp-Cy5.5 (clone Z231, Beckman-Coulter), Ki67-FITC (clone B56), Perforin-Pacific Blue (in-house custom conjugate, clone Pf-344, Mabtech). Flow cytometry acquisitions were performed on an LSR II (BD Biosciences, La Jolla, CA) and FlowJo software (version 9.6.4, Tree Star Inc., Ashland, OR) was used for all analyses. Pestle (Version 1.6.2) and SPICE (Version 5.1) were used for multi-parametric analyses.
TruCount flow cytometric assays for absolute CD4+ T cell counts were performed as previously described [39].
Cytokine concentrations in plasma were determined in a custom luminex assay as previously described [40]. Quantification of cytokines in mucosal washes was performed using a modified assay Millipore 23-plex non-human primate luminex kit platform. Briefly, colon or jejunum tissues collected at necropsy were diced into 3 mm pieces in R10 collection media. Aliquots of the cell-free media and plasma were snap-frozen for subsequent assays. Plates were read on a Bio-Rad 200 Bio-Plex system according to the manufacturer's suggested protocol. One hundred beads per regions were collected and results were optimized according to Bio-Rad software presets.
Mononuclear cells were stimulated with phorbol myristate acetate (PMA, 50 ng/mL) and ionomycin (1 ug/mL) or cultured in medium (RPMI 1640 containing 10% FBS) alone. Anti–CD107a (PerCp-Cy5, clone H4A3) was added directly to each of the tubes at a concentration of 20 µl/ml, and Golgiplug (brefeldin A) and Golgistop (monensin) were added at final concentrations of 6 µg/ml. After culture for 12 hours at 37°C in 5% CO2, cells were surface stained then permeabilized (Caltag Fix & Perm) and stained intracellularly with anti-IL-17 (APC conjugate, clone eBio64DEC17, eBioscience), anti-IFN-γ (PE-Cy7 conjugate, clone B27; Invitrogen), and anti-TNF-α (Alexa700 conjugate, clone Mab11).
NKp44+ ILCs from MLN of SIV-naïve rhesus macaques were stimulated overnight with either rhesus macaque IL-12 (50 ng/ml), human IL-1β (50 ng/ml), human IL-2 (1000 IU/ml) human IL-15 (50 ng/ml), or human IL-23 (50 ng/ml) (all from R&D Systems). After culture, cells were analyzed for intracellular expression of caspase-3 or RORγt (clone AFKJS-9, eBioscience).
Plasma SIV RNA copy numbers were determined using a standard quantitative real-time RT-PCR assay based on amplification of conserved gag sequences as described previously [41]. Tissue vDNA and vRNA quantifications were performed only on animals sacrificed at day-6/7 post-infection as described previously [42]. Briefly, 1-3 separate sections from duodenum, jejunum, ileum, cecum, colon, and rectum were examined and only 1 of 6 animals had detectable virus. Infection was confirmed by immunohistochemistry to SIV antigens in spleen or mucosal tissues (to be published elsewhere), but the undetectable vDNA and vRNA suggest infection in the GI tract was likely still focal at this early time point.
All statistical and graphic analyses were performed using GraphPad Prism 6.0 software (GraphPad Software Inc., La Jolla, CA). Nonparametric Mann-Whitney U tests were used where indicated, and a P value of <0.05 (by a 2-tailed test) was considered statistically significant.
Acute lentivirus infections are characterized by high levels of viremia coupled to a dramatic loss of CD4+ T cells in the GI tract [3], [4]. In our cohort of acutely SIV-infected macaques, plasma viremia peaked between days 10 to 14 (Fig. 1A), and mucosal CD4+ T cells were significantly depleted by day 14 in jejunum, colon, and MLN tissues when compared to naïve controls and remained reduced in chronic infection (Fig. 1C-1E). The frequencies and absolute numbers of peripheral CD4+ T cell numbers were also significantly reduced by day 14 post-infection (Fig. 1B, S1 Figure).
Unlike T cells, innate immune responses in the GI tract during acute SIV infection are poorly understood. We have previously reported that chronic SIV infection resulted in depletion of NKp44+ILCs and classic NKG2A+NK cells in GI tract large bowel tissues [23]. However, little is currently known about the effects of SIV infection on ILCs and NK cells in other sites of the GI tract, and nothing is known about the effects of acute infection. In the present planned-euthanasia study, we found, as early as one week after infection, there was up to a 3-fold decrease of NKp44+ILCs (Fig. 2A) in colons from acutely infected compared to naïve macaques (Fig. 2B). Using pre-infection colorectal biopsy samples, we also analyzed changes in NKp44+ILCs in individual animals sacrificed at day 14 post-infection and found that all 6 macaques exhibited NKp44+ILC depletion in colorectal tissue (S2A Figure). Furthermore, NKp44+ILCs were profoundly and persistently lost from jejunum in both acute and chronic SIV infections with 4- and 9-fold decreases, respectively (Fig. 2C). Frequencies of NKp44+ILCs in MLN and PaLN in either acute or chronic infection were unchanged compared to naïve macaques (Fig. 2C, 2D), suggesting loss of NKp44+ILCs may be compartmentalized. Furthermore, loss of ILCs did not correlate with viral load.
Although the percentage of NK cells in circulation was significantly decreased by day 14 post-infection (S2C Figure), in stark comparison to the massive loss of mucosal NKp44+ILCs there were no significant changes in the frequency of NK cells throughout the GI tract (Fig. 2E–2G). Furthermore, no change in NK cell numbers in colorectal tissues from individual animals pre- and post-infection was observed (S2B Figure). There was also no change in the distribution of CD56+CD16- (CD56+); CD56-CD16+(CD16+); and CD56+CD16+, and CD56-CD16- NK cell subsets during acute infection (S3 Figure).
To begin to address the mechanism of depletion of NKp44+ILCs in intestinal mucosae during acute SIV infection, we analyzed the expression of the active form of the apoptotic molecule casapase-3 and the proliferation marker Ki67. As shown in Fig. 3A, NKp44+ILCs from GI tissues had little to no expression of active-caspase-3 in naïve macaques, but increased greater than 100-fold in macaques sacrificed by day 14 post-infection. Interestingly, NKp44+ ILCs had no detectable change in Ki67 expression. By comparison NK cells had significant increases in both active-caspase-3 and Ki67 expression following infection (Fig. 3B). We have previously reported similar findings for ILCs in chronic SIV infection [23], but no correlation was found between caspase-3 levels and ILC frequencies. These disparate changes in turnover could account for the massive loss of NKp44+ILCs while NK cells were maintained in tissues.
Although NKp44+ILCs are mucosae-resident, trafficking in and out of the mucosae could influence the numbers of NKG2A+ NK cells. We next investigated whether turnover of NK cells in tissues after infection could be affected by trafficking to and within the GI tract. We have previously demonstrated that chronic SIV infection induces NK cell trafficking to the gut mucosae, characteristically by significant up-regulation of the gut-homing marker α4β7 on peripheral NK cells with concomitant down-regulation of the lymph node-homing markers, CCR7 and CD62L [23], [43]. Here, we compared the expression of each of these trafficking markers on NK cells in blood and mucosal tissues from naïve, acute and chronically SIV-infected macaques. As shown in S4 Figure, compared with uninfected macaques NK cells in both the circulation and tissues had no statistically significant differences in α4β7 expression except for PaLN. Similarly, no significant change in expression of CCR7, CD62L, or CXCR3 was detectable on NK cells from any GI tissue during acute infection. However, during chronic infection we observed a dramatic down-regulation of CXCR3 in colon, jejunum, and MLN. These data combined with previous observations from our laboratory and others [23]–[25] suggest that while SIV infection may alter NK cell trafficking to the mucosae, it likely occurs in chronic rather than acute infection (S4 Figure).
Our data show that acute SIV infection induces a massive depletion of NKp44+ILCs from the intestinal mucosae. We next asked whether there is any impact on the functionality of ILCs or NK cells during acute infection. We first analyzed a surrogate marker of cytotoxic potential, intracellular expression of the cytolytic granule, perforin. Our previous research suggested that while NKp44+ILCs are generally noncytolytic, under inflammatory conditions they can acquire killing activity [23]. As shown in Fig. 4A, mucosal NKp44+ILCs from acutely SIV-infected macaques had expression of intracellular perforin at low levels not significantly different from naïve animals. However, during chronic SIV infection intracellular perforin increased 4-fold in NKp44+ILCs within colon, jejunum, and MLN.
We then investigated cytokine production and degranulation by NKp44+ILCs as true measures of functionality. As shown in Fig. 4B–D, following mitogen stimulation, NKp44+ILCs in colon, MLN, and PaLN from acutely SIV-infected macaques had more than 3-fold increases in expression of CD107a+ and IFN-γ compared to naïve animals. Furthermore, multiparametric analysis revealed NKp44+ILCs in the GI tract from either acute or chronic SIV infected macaques have increased multifunctional capacity – upregulating CD107a and producing increased IFN-γ and TNF-α compared with naïve macaques (Fig. 4C, 4D). Interestingly, NKp44+ ILCs maintained their ability to produce IL-17 in acute, but not chronic SIV infection (S5A Figure). Furthermore, the acute loss of ILCs appeared to be associated with an acute loss of IL-17 in plasma and overall reduction in systemic and mucosal IL-17 in chronic disease (S5B Figure and S5C Figure), well in line with previous observations [23], [25], [26].
Thus far, it is still unclear what role NK cells play in controlling SIV replication in GI tract. NK cells can directly inhibit virus infectivity and lyse infected cells by releasing perforin and granzyme. Here, we investigated the expression of intracellular perforin in mucosal NK cells ex vivo. We found that mucosal NK cells in all tissues significantly upregulated perforin expression by two weeks post-infection (S6 Figure). Furthermore, perforin expression remained upregulated in chronic infection, suggesting heightened cytolytic potential throughout the disease course. Interestingly, perforin expression was increased in all subpopulations of NK cells except CD16+NK cells, which already had the highest levels of expression (S4 Figure). However, the level of perforin expression in mucosae resident NK cell did not have a statistically significant relationship with plasma viral load.
Lastly, we tested the functionality of mucosae-resident NK cells. As shown in Fig. 5B, following stimulation with PMA/ionomycin acutely SIV-infected macaques had 2-fold increases in CD107a and 1.5-fold higher IFN-γ secreting NK cell in colon. Multi-parametric analysis showed these mucosae resident NK cells from both acute and chronic SIV-infected macaques have increased percentages of multifunctional cells which were positive for CD107a, IFN-γ, and TNF-α, compared with naïve macaques (Fig. 5C, 5D).
Our data demonstrate that not only are NKp44+ ILCs depleted in SIV infection, but they also have altered phenotypes and functions. NKp44+ ILCs in macaques are most likely analogous to ILC3, but in infection their functional repertoire resembles that of ILC1. Moreover, ILCs stem from a common precursor are thought to exhibit varying degrees of plasticity, whereby environmental cues can convert ILC3 to ILC1 and vice versa. We next evaluated whether changes in the inflammatory environment due to infection might favor ILC1 development and thus explain the loss of ILC3. Interestingly we found that IL-7, which is necessary for both ILC1 and ILC3 development was elevated in acute infection and remained so during chronic disease (Fig. 6). IL- β which favors ILC3 development was either undetectable or remained unchanged. In contrast, IL-2, IL-12, IL-15, all of which favor ILC1 development, were all elevated at various points in disease. Indeed, in in vitro experiments culturing NKp44+ ILCs with various cytokines, IL-23 and IL- β both promoted RORγt expression, IL-2, IL-12, IL-15 suppressed expression and increased apoptosis (Fig. 7A, 7B). Overall these data suggest that the inflammatory environment found in both acute and chronic SIV infection depletes the ILC3 population by simultaneously inducing apoptosis and favoring development of ILC1.
In this study, we sought to explore the impact of acute SIV infection on ILCs and NK cells in the GI tract and draining lymph nodes. We observed a rapid and massive depletion of NKp44+ILCs, but not NK cells, as early as one week following SIV infection, at least partially attributable to increased apoptosis. Furthermore, we found both NKp44+ILCs and mucosae-resident NK cells had altered functional repertoires characteristic of heighted cytotoxicity. To the best of our knowledge, these data are the first to demonstrate a SIV-induced rapid and permanent loss of NKp44+ ILCs in the gut. The full implications of this novel aspect of lentiviral pathogenesis remain unclear.
Previously, based on a limited study focusing on colorectal tissue only, we reported that NKp44+ILCs were depleted during chronic SIV infection [23]. Soon after other groups corroborated our findings by reporting that that IL-17-secreting ILCs were lost from jejunum and in SIV-infected rhesus macaques [24], [25]. However, prior to the current study it was unknown kinetically when NKp44+ILCs were depleted and whether this phenomenon occurred in other mucosal tissues. In this new study we report that NKp44+ILCs were massive depleted from GI tissues as early as the first week after SIV infection, and remained suppressed during chronic disease. We also found that NKp44+ILCs had dramatically increased levels of apoptosis, without any change in proliferation rate, a plausible explanation for the net loss of cells. Our in vitro analyses clarified that apoptosis was due, at least in part, to increased inflammatory cytokines IL-2, IL-12, and IL-15 in the gut (Fig. 7). Furthermore, these new data demonstrate that the loss of NKp44+ ILCs is compartmentalized, since we observed no depletion in pararectal/paracolonic- and mesenteric-draining lymph nodes. Because we have demonstrated that NKp44+ ILC loss in the colorectum is partially due to increased inflammation (Fig. 3, Fig. 6, & Fig. 7) [23], [44], it will be of interest in future studies to determine if these mediators are not increased in lymph nodes. While a specific mechanism is unclear, ILC loss is unlikely due to infection as previous studies from our laboratory have shown [23].
NKp44+ILCs are most likely analogous to RORγt+ ILC3 [23] that play a significant role in mucosal homeostasis and intestinal integrity. The rapid and significant depletion of NKp44+ILCs after acute SIV infection suggests ILC loss might be directly or indirectly related to the breakdown of the gut epithelium, a hallmark of HIV/SIV disease [45], [46]. Indeed, a recent study in mice suggests ILC3 promote anatomical containment of lymphoid-resident bacteria through induction of antimicrobial peptides, and depletion of ILCs results in peripheral dissemination of commensal bacteria and systemic inflammation [11]. Thus, loss of ILCs might also have a direct role in gut breakdown and subsequent microbial translocation. Indeed, Klatt and colleagues [25] have previously shown that loss of IL-17/22 production by ILCs in SIV-infected macaques is associated with loss of epithelial integrity in the gut. ILC3 also exhibit significant regulation of myeloid and T regulatory cells in the gut through production of GM-CSF, and gastrointestinal dysfunction in SIV infection could be partially attributed to this mechanism [47]. NKp44+ILCs in the GI tract had further alterations shifting to multifunctional cells, including production of IFN-γ and TNF-α and gaining cytotoxic potential. Some human and mouse studies have actually shown IL-17- and IL-22- secreting ILC3 can become IFN-γ-secreting cells with T-bet acquisition induced by bacterial infection [48]–[50]. Our data demonstrating suppression of RORγt related to increased inflammatory mediators in the gut could explain this overall change in ILC phenotype and functional conversion of NKp44+ILCs to a proinflammatory and/or hypercytotoxic repertoire could exacerbate the pathogenesis of SIV infection.
Increased NK cell activity has been reported during HIV infections [51], [52] and we demonstrate a similar effect in both acute and chronic SIV infections. However, it has been difficult to ascribe a specific role for NK cells in either control of virus replication, or, alternatively, pathogenesis. Partly due to the fact that current strategies for NK cell depletion are incomplete [53], [54]. NK cells are known to lyse both HIV- and SIV-infected cells, but to what extent this occurs in vivo is unclear [34], [51], [52], [55]–[59]. Thus far, most studies have also focused on circulating NK cells, but in this study we investigated mucosae-resident NK cell responses during early SIV infection. The broad activation of NK cells in acute disease that persisted into chronic infection suggests NK cells are highly responsive to ongoing virus replication, a notion supported by their decreased activation during antiretroviral therapy [60]–[62]. Recently it was reported that NK cells from individuals carrying the KIR3DL1 receptor had greater tri-functional responses (TNF-α, CD107a and IFN-γ) [63], and linked these responses to greater control of viremia. Similarly, we found that during acute SIV infection, NK cells acquired a similar multifunctional phenotype. It will be of interest in future studies to determine if such a functional repertoire could be associated with greater control of focal virus replication in the gut or abortive infections.
In summary, we show acute SIV infection has significant, yet somewhat disparate, effects on two populations of mucosal innate lymphocytes. Although the precise niche of NKp44+ ILCs in primates is yet to be elucidated, it is tempting to speculate that the early and massive loss of these cells constitutively producing IL-17 and IL-22 could have a significant impact on mucosal homeostasis. Furthermore, the increases in cytotoxicity and inflammatory cytokine production by both ILCs and NK cells during acute infection are likely to contribute to the massive apoptosis and dysregulation in the gut. Regardless, given the profound impact of acute SIV infection on both cell types, further study into both the underlying mechanisms and clinical consequences are warranted.
|
10.1371/journal.pcbi.1007030 | Disentangling juxtacrine from paracrine signalling in dynamic tissue | Prolactin is a major hormone product of the pituitary gland, the central endocrine regulator. Despite its physiological importance, the cell-level mechanisms of prolactin production are not well understood. Having significantly improved the resolution of real-time-single-cell-GFP-imaging, the authors recently revealed that prolactin gene transcription is highly dynamic and stochastic yet shows space-time coordination in an intact tissue slice. However, it still remains an open question as to what kind of cellular communication mediates the observed space-time organization. To determine the type of interaction between cells we developed a statistical model. The degree of similarity between two expression time series was studied in terms of two distance measures, Euclidean and geodesic, the latter being a network-theoretic distance defined to be the minimal number of edges between nodes, and this was used to discriminate between juxtacrine from paracrine signalling. The analysis presented here suggests that juxtacrine signalling dominates. To further determine whether the coupling is coordinating transcription or post-transcriptional activities we used stochastic switch modelling to infer the transcriptional profiles of cells and estimated their similarity measures to deduce that their spatial cellular coordination involves coupling of transcription via juxtacrine signalling. We developed a computational model that involves an inter-cell juxtacrine coupling, yielding simulation results that show space-time coordination in the transcription level that is in agreement with the above analysis. The developed model is expected to serve as the prototype for the further study of tissue-level organised gene expression for epigenetically regulated genes, such as prolactin.
| Prolactin is a major hormone product of the pituitary gland, the central endocrine regulator found underneath the brain. In mammals it is crucial for milk production and reproductive function. Production of such an important protein needs to be regulated tightly, and therefore one might imagine that its gene expression is largely static. However, recent experiments using real-time imaging techniques at a single-cell resolution have revealed prolactin gene transcription to be highly dynamic and stochastic in nature, while displaying clear tissue-scale space-time coordination. This discovery raised a new question, namely, what kind of cellular communication mediates such a space-time organization? In this study, by developing a statistical method that involves network theory, we show that such unexpected behaviour involves contact-driven inter-cell signalling. The study develops a mathematical model that can reproduce realistic levels of space-time coordinated gene expression. The method and model developed here are generic and can be used in the study of other signalling systems that show space-time coordinated behaviour.
| Gene expression at a single-cell level is highly dynamic in time, and the processes involved in gene activation and inactivation are now well-known to be highly stochastic [1–7]. The use of single-cell live imaging techniques, which employ luciferase or light emitting proteins such as destabilised EGFP as a reporter (Fig 1A, [1, 2]) aided by statistical models that infer the gene transcription process from reporter signals, has been critical for this development in the understanding of transcriptional dynamics. While they were thought to be smoothly changing graded processes, mathematical modelling has indicated that they can be well explained by discrete time stochastic switch functions [1, 3–6]. This might be a binary switching between on and off states [1], or a more complex process [7–9]. While much progress has been made in understanding the processes involved in transcriptional control in single cells, an important challenge is to translate this to understanding the transcriptional dynamics in multicellular tissues. In the context of the behaviour of intact tissue, overall gene expression levels should be accurately controllable and predictable, but it is still unclear how overall coordinated tissue function emerges from the switching transcriptional dynamics of individual cells.
Space-time coordination requires communication between cells. At the tissue scale we consider such signalling to be either juxtacrine or paracrine ([11], Ch. 15). Juxtacrine signalling is a type of cellular communication between contacting cells, for example by means of gap junctions that allow for signalling molecules to pass from cell to cell. This type of interaction can be transitive, allowing distant cells to communicate with each other by successive cellular contacts. In contrast to this, paracrine signalling does not rely on any cellular contact but depends on the intercellular diffusion of signalling molecules. It is possible that these two signalling mechanisms coexist.
The mammalian pituitary gland, which secretes a series of key hormones including prolactin, has proved to be an excellent model system for the study of dynamic gene expression in vivo ([1, 8, 12]). In contrast to tissues with deterministic developmental programmes, the pituitary gland generates complex function from an assortment of intermingled cells that require both short- and long-term adaptive processes. The different cell types are arranged as interdigitated networks, linked by specific cell-contacts, including gap junctions [13]. The cell networks adapt structurally and functionally to ensure temporal optimisation of hormone expression in response to different reversible physiological adaptations such as pregnancy and lactation [14].
In the present paper, we propose a novel statistical methodology to assess the existence of juxtacrine and paracrine signalling mechanisms between cells in the pituitary gland by analysing quantitative single-cell live imaging data obtained from rat pituitary tissue slices. We apply this method to the transcriptional dynamics of the prolactin gene after inferring dynamical transcription profiles from imaging using Stochastic Switch Modelling (SSM, Fig 1A, [7]). By developing a stochastic simulation model, we test if the signalling mechanisms identified above are sufficient to reproduce the observed space-time structure.
While we studied three adult tissue portions, labelled D1, D2, and D3, to ensure reproducibility, we present mainly the results obtained for D1, which correspond to the data studied in [8]. The results for D2 and D3 are consistent with the findings for D1, and are summarised in S1–S3 Tables.
Fig 1B shows a snapshot during the GFP-imaging of one of the tissue slices studied here, with a white enclosure indicating the analysed area. The tissue areas consist of about 100 cells, whose spatial locations are found to be random (Fig 1C, examined in detail in [8]), while their transcriptional behaviour appears quite dynamic (Fig 1D). Some cell pairs display high correlation in their GFP time series, but others do not (Fig 1E). In [8], we showed that the expression of the prolactin gene was coordinated in space and time at a tissue scale in adult male rats, while no such coordination was observed at an embryonic or a neonatal stage. However, this is not merely a consequence of a purely spatial organisation of cells growing in development. While such structures have been documented for the pituitary on a relatively large scale [12], we showed in [8] that the cellular locations were statistically random in space in our adult tissues while the space-time coordinated behaviour was underlain by the cellular network defined by the cellular contacts. The importance of gap junctions for this coordination was suggested because the application of the gap-junction inhibitor, AGA (Alpha-glycyrrhetinic acid), significantly reduced the correlation between transcriptional time-profiles of individual cells. This motivated our development of the method proposed here to estimate the extent of juxtacrine and paracrine signalling.
While we have revealed the presence of space-time structure in the set of GFP signals on prolactin in our previous work [8], here we examine this important property in more detail, and later provide a mathematical model for it. Fig 2A shows ten randomly chosen GFP time series, for visual clarity, from all 101 cells in the tissue of interest. These are carefully reconstructed profiles by combining the signals from two channels of different light sensitivity such that the linearity holds between light intensity and the GFP population (Fig 2B, see also Materials and Methods). Pearson’s correlation coefficients are calculated between all cell pairs, and plotted against the corresponding Euclidean distances in Fig 2C. To analyse a trend in this kind of scatter plot we generally used the method of quantile regression [15], which is robust to the non-normality arising from the boundedness of the correlation coefficient. In particular, we focused on the median (red). A statistically significant rightward decline in this median regression line indicates the presence of a space-time correlation structure (see S1 Table for the corresponding p-value). Here, and throughout this paper, we use the one-sided t-test.
While the results in Fig 2C establish the presence of a space-time correlation, they do not reveal underlying mechanisms. To identify the contributions of juxtacrine and/or paracrine signalling, we introduce another distance measure, namely geodesic distance. Geodesic distance between two cells is associated with their shortest paths defined on a cellular network, which in turn is defined here by cellular contacts [16]. It is the smallest number of cellular contacts between two cells, and therefore takes an integer value if they reside on a common cluster, or is set to infinity if there is no path between them.
The geodesic distances are estimated as detailed in Materials and Methods. Although we cannot directly observe contacts between cells, we can use cell positions and sizes to statistically estimate geodesic distance. It is therefore important to identify the distribution of cell sizes. Fig 3A shows these distributions inferred from two experimental methods–light-sheet microscopy, and confocal microscopy. They both result in similar values for the mean and standard deviation of the cell diameters, namely (12.0, 2.2) μm and (13.1, 1.9) μm, respectively. Because light-sheet microscopy allows for measurement in the depth as well as the horizontal directions, it can be assumed to give more precise measurements, and is therefore used in our further analysis, while the cell-size sensitivities are investigated in S3 Table. We also considered electron microscopic measurement ([8], Fig 4), but this resulted in apparently small sizes (9.0, 1.2) μm due to some cell shrinkage during the fixation process [17, 18]. Electron microscopic estimates of cell sizes were therefore not used in this study.
In addition to the above investigations, cell sizes are also estimated computationally from the spatial distributions of cell centroids (see Fig 1C and Materials and Methods). In Fig 3A, the estimated mean cell diameters (green) in three datasets, D1, D2 and D3, are compared to the other experimentally determined values. A good degree of coincidence between them suggests that experimental estimates are reasonable. It also suggests that cells are densely packed, as the computational cell size estimation is based on a model that assumes a 2-D dense packing of circles that accounts for 79% of the tissue area (see Materials and Methods).
Using the cell size distribution obtained from the light-sheet microscopy, geodesic distances between all cell pairs are statistically estimated, and compared to their corresponding Euclidean distances in Fig 3B. While there is an overall linear relationship between these two distance pairs, there is still some variation, which allows us to distinguish between the two types of signalling as follows. All cell pairs, with associated GFP time series, are now characterised by three quantities (r, dE, dG), with r the correlation coefficient, dE Euclidean distance, and dG geodesic distance. In a collection of cell pairs within a given dE range, if r decreases with dG, it suggests juxtacrine signalling (Fig 3C, left). On the other hand, in a collection of cell pairs within a given dG range, if r decreases with dE, it suggests paracrine signalling (Fig 3C, right). These two types of signalling can co-exist.
Before moving on to the analysis of signalling mechanisms, we provide further evidence that the proposed method of estimating geodesic distances in a GFP data results in realistic values. Fig 4A shows a montage of x800 EM images of anterior pituitary tissue, which is different from those GFP imaged (D1, D2, D3). In this the cellular contacts between lactotrophes are clearly visible. Although there might have been some shrinkage of the cells [17, 18], since geodesic distance is a scale-invariant property it is expected that such an image will give a realistic estimate of the distribution of geodesic distances between cell pairs. Each lactotroph is shaded in a different colour to distinguish individual cells. There are 18 cells that are fully contained in this montage, and are labelled by a *. For these cells, geodesic distances are calculated, and their distribution is shown in Fig 4B as a histogram. In comparison, geodesic distances in the GFP dataset are calculated for the same number of 18 cells nearest to the tissue centre (see Fig 1C). Their histogram is shown in Fig 4C. Amongst those 18 cells, one was found to be isolated from the other 17 cells, causing the histogram y-axis limit smaller than that in Fig 4B. Otherwise, however, geodesic-distance distributions in Fig 4B and 4C appear to be similar, endorsing the use of the present method of geodesic distance calculations in the GFP datasets. This figure also confirms the tight packing of the cells.
The presence of juxtacrine and paracrine signalling is investigated in the cell-pairs within a limited range of Euclidean (dE) and geodesic (dG) distances, respectively, where dE and dG are set small. This is because small dE (dG) is generally associated with large δdG/dG (resp. δdE/dE, where δdG denotes variation in δdG while δdE variation in dE), allowing better identification of juxtacrine (resp. paracrine) signalling. Fig 5 shows the results of this analysis in dataset D1 using quantile regressions, for juxtacrine (a–c) and paracrine (e, f) signalling mechanisms. Statistical test results are summarised in S1 Table for all three datasets. The trend lines strongly suggest the presence of juxtacrine signalling, with no strong evidence of paracrine signalling (notice that the trend lines in (e) and (f) are much flatter than that in (d), with slopes of (d) -0.25, (e) -0.028, and (f) -0.0018). However, the latter cannot be entirely ruled out as one may conclude the presence of some, albeit weak, paracrine signalling mechanisms if the cell-pair samples in each distance-limited test are pooled across the three datasets (see S1 Table).
Note that in panel (d), unlike (e) and (f), the cell pairs of geodesic distance 1 imply that cells are physically contacting. Hence, the clear trend of decreasing correlation coefficient with Euclidean distance suggests that the more tightly packed the cells the stronger the correlation. This could, for example, be facilitated by gap junctions on the cell surface.
Our conclusions stated above remain consistent when we repeat our analysis based on the cell size distribution identified by confocal microscopy instead (S3 Table).
Having identified juxtacrine signalling in a population of cells in a tissue at the level of expression of GFP reporter signal, we now attempt to approach its origin more directly in the expression of the transcription rate of the reporter gene, as a more direct measure of the promoter activity of the prolactin gene regulatory elements contained in the prolactin-d2EGFP transgene. If successful, this analysis will show that the spatial correlation in transcription is mostly due to juxtacrine signalling and therefore provides a more fundamentally mechanistic insight. This approach removes the influence of post-transcriptional activities in GFP expression (Fig 1A).
As explained in [8] the Stochastic Switch Model (SSM, [7]) enables the inference of different levels of transcription rate as well as the timing of switches between different levels of activity. The model uses a reversible jump Markov chain Monte Carlo algorithm [19] to produce a posterior probability distribution over all possible transcriptional profiles for each cell. Post-processing of this distribution enables the extraction of a number of accepted candidate transcriptional profiles, each with an associated probability of occurrence (See figure 2A in [8])). Consequently, the transcriptional analysis for each cell provides a weighted analysis of all possible transcriptional profiles taking into account the probability of occurrence of each profile (see Materials and Methods for more details). Fig 6A shows the estimated transcription profiles for the ten example cells considered in Fig 2A, which are showing characteristic temporal changes in transcription at a few distinct switch time points [1]. We then perform a space-time analysis, that is analogous to the one applied to the GFP signals (Fig 5), to the distribution of reconstructed transcriptional profiles.
To measure and quantify the correlation between such discrete profiles as in Fig 6A, there are, in theory, a range of possible scoring functions. Here we define two score functions, Sc1 and Sc2, by using Eq 1 in Materials and Methods. To compute Sc1, we take the transcription profiles for each pair of cells weighted by their probabilities and compute the Pearson correlation coefficient. Thus, this score function considers the correlation in the transcription levels over time. To compute the second score, Sc2, we focus on the times of the transitions. In this we replace the transcription profiles by the sum of normal distributions placed at the transition points as illustrated in Fig 6C and measure the Pearson correlation between these. The standard deviation of these is set to 3-hours in accordance with the typical spacing found in the profiles (see [8], Fig 2Aii).
Analysis using both types of score function shows significant space-time correlation at the transcriptional level (Fig 6B and 6D). The analytical approach using Euclidean and geodesic distances is analogous to the approach we developed for the GFP signal above and is summarised in S2 Table. Although it is less clear in individual datasets, when cell-pair samples are pooled across all three datasets, juxtacrine signalling appears to be the dominant mechanism of cellular interaction, with only a weak indication of paracrine signalling. These results are consistent with those found for the GFP signals (S1 Table). The results in S2 Table are for score function Sc1 but use of Sc2 yields qualitatively the same results for the pooled data as shown in S2 Table.
More details about the model are given in S1 Text. We proceed by proposing a stochastic model for prolactin gene expression that incorporates juxtacrine signalling (Fig 7). We address the question of whether such a model with parameters estimated from experimental data can reproduce the observed space-time structure. We assume a three-state model for the prolactin gene as suggested by the analysis in [1], which assumes that each allele can be in one of three states: on, off or primed (Fig 7). When a gene is off or primed none or very little mRNA is transcribed, while when it is on, mRNA is produced at a faster rate. Moreover, the gene transitions from the off-state to the on-state only occur via the primed state and the times in each of these states are exponentially distributed with half-lives toff, tprimed and ton, respectively. The rates and the transcription rates corresponding to the three states are estimated from the data (see Materials and Methods for more details). To model coupling between the cells we assume that a cell senses the states of its neighbours and changes its gene cycle dynamics accordingly. More specifically, here we assume that for a cell the rates of transition between the off and primed states and between the primed and on-state are influenced by the number of on genes in the connected cells as described in the Materials and Methods.
An example of mRNA profiles of all the cells in one simulation is shown in Fig 8A, where a characteristic mean convex shape is reproduced as observed in the real experiments (S1 Fig). The corresponding correlation analysis is shown in Fig 8B, where, for all cell pairs, correlation coefficients at the mRNA level are plotted against Euclidean distances, with a trend line resulting from the median regression. The negative slope of this trend line suggests space-time correlation as seen above in the analysis of experimental data. The correlation analysis can also be done by using the transcription-rate profiles (β(t), defined as Eq 10 in Materials and Methods, see S2 Fig for the transcription-rate counterparts of Fig 8A and 8B). The results are comparable to the results in the experimental data at the level of the SSM-inferred transcription-rate profiles (see Fig 6B for dataset D1), while the results in mRNA profiles may indirectly be compared to those of the experimental data at the GFP level. When such a set of simulation and analysis are repeated 200 times, we see that negative values dominate, both at the transcription-rate and mRNA levels, as is shown in the distributions in Fig 8C (signalling and its associated dynamics are highly stochastic, and therefore the slopes could take positive or negative values. The role of signalling is to shift the distribution in the negative direction. S4 Fig shows the slope distribution when cells are not communicating. We also see that the distribution of the slope of the trend line at the transcription-rate level matches nicely the slopes determined in the three experimental datasets. Gradients are greater when we restrict to the short range (<24 μm, or equivalently twice the cell diameter) than in the longer range. This is also true in the experimental data. However, as already implicated in the analysis of the experimental data, the spatial coupling is more likely to be in action also in the processes downstream to the on-off-primed gene cycle up to the GFP profiles. If we assume, for example, an additional mechanism in the transcription dynamics defined by Eq (6), the slopes, on average, become larger as shown by the distribution of the grey dots in Fig 8C. This is one possible mechanism leading to high space-time correlations at the GFP level, while others will be discussed in the Discussion section.
Space-time coordination in the expression of the prolactin gene was reported for the first time by [8] and the importance of gap junctions for this coordination was discussed. The present study strongly endorses such a picture and uncovers the role of juxtacrine signalling in a robust fashion. In addition, the present study shows a weak contribution of paracrine signalling.
While we have seen that GFP-visible cells were densely packed (Figs 3A and 4 and [20]), it is still a possibility that the network structure was not fully revealed. A particular possibility is that cells are connected in 3D-space, resulting in an overestimation of geodesic distance dG when dG >2. Geodesic distances can indeed only be overestimated when viewed in 2-D. However, this does not challenge our conclusion that juxtacrine signalling is functioning because if geodesic distances are overestimated the trend line in the plot of correlation coefficients v. geodesic distance, will show an even steeper decline. Note also that the influence of juxtacrine signalling is also evident for touching cells, characterised by dG = 1, which is free from this bias (Fig 5D). On the other hand, in the test of paracrine signalling, any overestimation of geodesic distance (dG) would influence the membership of a cell-pair subset, in which the correlation coefficient is regressed on Euclidean distance. In the present study, tests were performed in two overlapping dG ranges, 2 or 3 and 3 or 4, so that the general lack of a negative trend suggests possible corrections in the estimation of geodesic distances are less likely to change the conclusion (see S1 and S2 Tables, remember dG was set small (see Results)). Nonetheless, a precise 3-D estimation of the cellular network might be able to rule out paracrine signalling with more clarity, and in general this remains an important task for future work for prolactin and other systems. Imperfect cellular segmentation might be happening during image analysis. Overlap of the cells due to boundary errors could cause errors in the estimated, but this is only relevant for touching pairs. Therefore, the influence of this on the trend line in the plot of correlation coefficient against distance should be minor.
In the present work, we used the linear correlation coefficient as a measure of temporal coordination in a pair of time series. This uses a zero time lag between the two time series. Using other time lags tends to decrease correlations (S5 Fig). While this is a natural choice, which appears successful in distinguishing between signalling types, this is not the sole candidate. While we tested two score functions with qualitatively similar results, the use of other score functions may capture different aspects of the system, in particular the temporally discrete nature of the back-calculated transcription activities as a step function (see Fig 6A).
We have so far established that prolactin gene promoter activity shows space-time correlations on the spatial scale of 100 μm and the temporal scale of 48 hours. Coordination of behaviour across a tissue allows for more accurately controllable hormone production by the tissue as a whole, and may be important for coordination of major episodes of hormone production, such as during the physiological demands of lactation.
We have also established that the observed space-time correlations are mediated by juxtacrine signalling occurring at the level of gene transcription. Stronger evidence for the juxtacrine signalling in the GFP signals than in the inferred transcription rates suggests the involvement of in-/post-transcription juxtacrine signalling. The next question is to reveal the molecular mechanisms that underlie juxtacrine signalling, which may be associated with multiple molecules or channels. Regarding this question, our preceding work has shown that a gap-junction blocker, AGA (Alpha-glycyrrhetinic acid), diminishes the space-time correlation [8]. Application of the methods developed in the present paper to AGA-treated tissues may reveal the loss of juxtacrine signalling in those tissues. However, no cell-diameter data is available to us for the AGA-treated tissue. Such an investigation will need careful determination of cell sizes in a pharmacological environment. Moreover, it should be noted that gap-junction blockers, such as AGA, often have non-specific effects and these may disrupt the network structure. There is a need to specifically delete Cx43 (connexin 43) in PRL cells. Post-transcriptional juxtacrine signalling may also be by mRNA trafficking between cells. A future mathematical model that involves prolactin and other proteins is likely to involve such post-transcription mechanism(s).
While juxtacrine signalling was found to be predominant, it may be too early to rule out the presence of paracrine signalling in vivo. This is because paracrine signalling may be impaired in tissue slices, where blood supply and other mechanisms that could influence transport factor are not present.
We have shown in [8] that space-time correlation arises in the course of development, and that such a structure is much less visible in embryonic/neonatal rat tissues. The application of the present analysis method will require careful determination of cell sizes at each developmental stage. If the proportion of GFP-visible cells at embryonic stage is much smaller than at adult stage, cellular network estimation may only be possible by the use of a cellular membrane marker, which makes all cells visible. Including also the case of adults, determining the degree of influence of non-GFP-visible cells remains an important question in future work.
As reviewed briefly in the Introduction, the discovery that the gene expression is stochastic and heterogeneous, while showing some degree of space time coordination was made very recently [8]. The mathematical model developed in the present study is the first model that addresses all three properties, stochasticity, heterogeneity and space-time coordination, with a proper consideration of the refractory period in gene transcription. We have parameterised this model by fitting it to experimental data and shown that it can reproduce the observed space-time correlation. In future this model may be used to guide biological experiments, serving as the base model, with modifications such as the inclusion of more juxtacrine and paracrine signalling channels.
It could be objected that our analysis has not ruled out a situation where a very slowly diffusing molecule that on the relevant time-scale only reached cells neighbouring that which produced. However, this would need a molecule that diffuses less than a few hundred μm over a time scale of a few hours which seems unlikely and in any case such signalling would likely function as though it was true juxtacrine signalling.
In [8] we discussed a global picture where transcription is highly stochastic but has some coordination of bursting at distances up to approximately 35 μm in adult pituitary tissue, but not at greater distances. We hypothesised that short-range signalling allowed uncorrelated behaviour of well-separated cells combined with short range cell-to-cell communication that may stabilise longer term changes in the expression level of the tissue, such as those associated with the oestrus cycle or lactation. Dynamic control is necessary (e.g. acute suppression or activation by hypothalamic regulators) and while the heterogeneity of transcription across cells may well guard against overshooting or oscillation, when making rapid changes local coordination may be necessary to overcome the damping effect of this and allow a quicker response.
The overall method developed in the present research is generic and can therefore be applied to a wide range of other systems. For example, they may be applicable to other dynamic processes that may be important for prolactin and other endocrine cell function (e.g. Ca2+, voltage etc.). The good agreement between the results obtained in the GFP signal (S1 Table) and the results obtained at the transcription level (S2 Table) suggests that juxtacrine signalling is predominant.
Animal studies were performed under UK Home Office licence (PPL: 40/3296 and PPL:70/8082) following local ethical review by the University of Manchester's Animal Welfare and Ethical Review Body.
As mentioned in [8], the fit of the SSM model was tested through calculation of recursive residuals as a way of comparing the prediction from the model and the observed data (for more information see ([7], appendix G). These showed no departure from the model assumptions, indicating that the switch model fitted the data well.
The parameter values were estimated by analysing the transcription profiles statistically inferred by SSM to two GFP-time-course datasets (male rats with no stimulation; maleCont, maleCont16), and shown in S4 Table. The maximum likelihood estimates of refractory and off periods, T1 and T2, were obtained by fitting the sum of two exponential distributions to the durations after a down switch with the assumption of T1 < T2, while on period, T0, was estimated by fitting an exponential distribution to the durations after an up switch. Both durations in the SSM results were analysed as right-censored ones. These period estimations are detailed below. Low and high rates, βL and βH, are the respective transcription rates after a down and up switch.
|
10.1371/journal.pbio.1002018 | Millisecond-Scale Motor Encoding in a Cortical Vocal Area | Studies of motor control have almost universally examined firing rates to investigate how the brain shapes behavior. In principle, however, neurons could encode information through the precise temporal patterning of their spike trains as well as (or instead of) through their firing rates. Although the importance of spike timing has been demonstrated in sensory systems, it is largely unknown whether timing differences in motor areas could affect behavior. We tested the hypothesis that significant information about trial-by-trial variations in behavior is represented by spike timing in the songbird vocal motor system. We found that neurons in motor cortex convey information via spike timing far more often than via spike rate and that the amount of information conveyed at the millisecond timescale greatly exceeds the information available from spike counts. These results demonstrate that information can be represented by spike timing in motor circuits and suggest that timing variations evoke differences in behavior.
| A central question in neuroscience is how neurons use patterns of electrical events to represent sensory information and control behavior. Neurons might use two different codes to transmit information. First, signals might be conveyed by the total number of electrical events (called “action potentials”) that a neuron produces. Alternately, the timing pattern of action potentials, as distinct from the total number of action potentials produced, might be used to transmit information. Although many studies have shown that timing can convey information about sensory inputs, such as visual scenery or sound waveforms, the role of action potential timing in the control of complex, learned behaviors is largely unknown. Here, by analyzing the pattern of action potentials produced in a songbird's brain as it precisely controls vocal behavior, we demonstrate that far more information about upcoming behavior is present in spike timing than in the total number of spikes fired. This work suggests that timing can be equally (or more) important in motor systems as in sensory systems.
| The relationship between patterns of neural activity and the behaviorally relevant parameters they encode is a fundamental problem in neuroscience. Broadly speaking, a neuron might encode information in its spike rate (the total number of action potentials produced per unit time) or in the fine temporal pattern of its spikes. In sensory systems as diverse as vision, audition, somatosensation, and taste, prior work has demonstrated that information about stimuli can be encoded by fine temporal patterns, in some cases where no information can be detected in a rate code [1]–[11]. This information present in fine temporal patterns might be decoded by downstream areas to produce meaningful differences in perception or behavior.
However, in contrast to the extensive work on temporal coding in sensory systems, the timescale of encoding in forebrain motor networks has not been explored. It is therefore unknown whether the precise temporal coding of sensory feedback could influence spike timing in motor circuits during sensorimotor learning or whether millisecond-scale spike timing differences in motor networks could result in differences in behavior. Although many studies have shown that firing rates can predict variations in motor output [12]–[14], to our knowledge no studies have examined whether different spiking patterns in cortical neurons evoke different behavioral outputs even if the firing rate remains the same.
The songbird provides an excellent model system for testing the hypothesis that fine temporal patterns in cortical motor systems can encode behavioral output. Song acoustics are modulated on a broad range of time scales, including fast modulations on the order of 10 ms [15],[16]. Vocal patterns are organized by premotor neurons in vocal motor cortex (the robust nucleus of the arcopallium [RA]) (Figure 1a), which directly synapse with motor neurons innervating the vocal muscles [14],[15],[17]. Bursts of action potentials in RA (Figure 1b) are precisely locked in time to production of vocal gestures (“song syllables”), with millisecond-scale precision, suggesting that the timing of bursts is tightly controlled [18]. Similarly, the ensemble activity of populations of RA neurons can be used to estimate the time during song with approximately 10 ms uncertainty [15]. However, although these prior studies demonstrate that the timing of bursts is tightly aligned to the timing of song syllables, it is unknown how variations in the temporal pattern of spikes might encode the trial-by-trial modulations in syllable acoustics known to underlie vocal plasticity [19]. Significantly, biomechanical studies have shown that vocal muscles in birds initiate and complete their force production within a few milliseconds of activation (far faster than in most mammalian skeletal muscles), suggesting that RA's downstream targets can transduce fine temporal spike patterns into meaningful differences in behavior [20],[21]. However, while it is clear that trial-by-trial variation in spike counts within a 40 ms time window can predict variations in the acoustics of individual song syllables [14],[22], it is unknown whether the precise timing of spikes within bursts might be an even better predictor of vocal motor output than spike counts.
To quantify the temporal scale of encoding in the vocal motor system, we adapted well-established mathematical tools that have previously been applied to measure information transfer in sensory systems. First, we used a spike train distance metric to quantify the differences between pairs of spike trains produced during different renditions of individual song syllables and a classification scheme to quantify whether distance metrics based on rate or timing yielded the best prediction of acoustic output [23],[24]. Second, we used model-independent information theoretic methods to compute the mutual information between spike trains and acoustic features of vocal behavior [8],[10]. Crucially, both techniques measure information present in the neural activity at different timescales, allowing us to quantify the extent to which spike timing in motor cortex predicts upcoming behavior.
We collected extracellular recordings from projection neurons in vocal motor area RA in songbirds (Figure 1a). In total, we analyzed 34 single-unit cases and 91 multiunit cases, where a “case” is defined as a neural recording being active prior to the production of a syllable (Figure 1b), as explained in Materials and Methods. The number of trials (syllable iterations) recorded in each case varied from 51 to 1,003 (median 115, mean 192.4). Iterations of each song syllable were divided into groups based on acoustic similarity (“behavioral groups”) (Figure 1c and 1d), and information-theoretic analyses were used to quantify whether spike timing conveys significant information about upcoming motor output, as schematized in Figure 1e.
We first used a version of the metric-space analysis established by Victor and Purpura to compare the information conveyed by spike rate and spike timing [24],[25]. As described in Materials and Methods, this analysis quantifies how mutual information between neural activity and motor output depends on a cost parameter q, which quantifies the extent to which spike timing (as opposed to spike number) contributes to the dissimilarity, or “distance,” between spike trains (Figure 2a). The distance between two spike trains is computed by quantifying the cost of transforming one spike train into the other. Here, parameter q, measured in ms−1, quantifies the relative cost of changing spike timing by 1 ms, as compared to the fixed cost of 1.0 for adding or subtracting a spike. Spike train distances are then used to classify iterations of each song syllable into behavioral groups, and the performance of the classifier I(GP,G) is used to quantify the mutual information between neural activity and vocal output. Figure 2b shows a representative “rate case,” where qmax = 0 (that is, information is maximized at q = 0, where spike train distances are computed based solely on spike counts). As q increases, the performance of the classifier decreases from its maximal value. This means that the best discrimination between behavioral groups (Figure 1c and 1d) occurs when only spike counts are used in calculating the distances between pairs of spike trains. In contrast, Figure 2c illustrates a “temporal case.” In temporal cases, mutual information between neural activity and vocal motor output reaches its peak when q>0. This indicates that there is better discrimination when spike timings are taken into consideration. Note that in the case shown in Figure 2c, the rate code does not provide significant information about behavioral output (empty symbol at q = 0).
Across all analyses in cases where information was significant at any value of q, including cases where qmax = 0, the median value of qmax was 0.3, suggesting a high prevalence of temporal cases. Figure 2d shows the prevalence of rate cases and temporal cases in our dataset. As described in Materials and Methods, we assigned the iterations of each song syllable to behavioral groups based either on a single acoustic parameter (e.g., pitch, Figure 1c) or using multidimensional clustering (3D acoustics) (Figure 1d). The different grouping techniques yielded similar results. When syllable acoustics were grouped by clustering in a three-dimensional parameter space (Figure 2d, blue bars), the fraction of temporal cases was significantly greater than the fraction of rate cases (blue asterisk; p<10−8, z-test for proportions). Similarly, temporal cases significantly outnumbered rate cases when acoustics were grouped using only a single parameter (pitch, amplitude, or spectral entropy, shown by green, yellow, and red asterisks respectively; p<10−8). Note that in approximately 25% of cases (between 31 and 36 out of 125 cases across the four analyses shown in Figure 2d) these analyses did not yield a significant value of I(GP,G;q) for any value of q; the fractions in Figure 2d therefore do not sum to unity. In such cases variations in a neuron's pattern of neural activity during a particular syllable are not predictive of variations in the particular acoustic parameter being analyzed. Furthermore, note that an alternate version of this analysis in which the spike train distance measurement was not normalized by the total number of spikes (see Materials and Methods) yielded nearly identical results, as shown in Figure S1. Additionally, we asked whether the proportions of temporal cases shown in Figure 2d were significantly greater than chance by randomizing the spike times in each trial (Poisson test; Materials and Methods). This analysis revealed a significant proportion of temporal cases when vocal acoustics were measured by multidimensional clustering (3D acoustics, p<0.05 after Bonferroni correction for multiple comparisons indicated by cross in Figure 2d), but the same measure fell short of significance when the three acoustic parameters were considered individually (p = 0.06–0.24 after Bonferroni correction).
To measure the maximum information available from the metric-space analysis, we computed Īmax, the average peak information available across all cases (see Materials and Methods). Across all metric-space analyses Īmax was 0.10 bits out of a possible 1.0 bit. As discussed below, this value suggests that additional information might be available in higher-level spike train features that cannot be captured by metric-space analyses. Additionally, since the proportion of rate and temporal cases did not differ significantly when computed from single- or multiunit data (p>0.07 in all cases; z-tests for proportions), we combined data from both types of recording in this as well as subsequent analyses. The similarity between the single- and multiunit datasets likely results from multiunit recordings in this paradigm only reflecting the activity of a single or a very small number of neurons, as discussed previously [14]. Finally, the results of the metric-space analysis were not sensitive to the number of behavioral groups used to classify the iterations of each song syllable. Although our primary analysis uses two behavioral groups (Figures 1c and 2), as shown in Table 1 we found a similar prevalence of rate and temporal cases when the trials were divided into three, five (Figure 1d), or eight groups.
Our metric-space analysis therefore indicates that in most RA neurons, taking the fine temporal structure of spike trains into account provides better predictions of trial-by-trial variations in behavior than an analysis of spike rate alone (asterisks, Figure 2d). Furthermore, at least when vocal outputs are grouped in three-dimensional acoustic space, spike timing can predict vocal acoustics significantly more frequently than would be expected from chance (cross, Figure 2d). However, it remains unclear whether spike timing can provide information about single acoustic parameters (beyond the 3-D features). Finally, since not all cases were temporal, it is also unclear how important this information is on average, rather than in special cases. Answering these questions necessitates the direct method of calculating information, as described below.
In the metric-space analysis, not all cases were classified as temporal. Further, when behavior was grouped by a single acoustic parameter rather than in multidimensional acoustic space, the number of temporal cases was not significantly larger than by chance (Figure 2d, green, yellow, and red plots). Thus it still remains unclear to what extent spike timing is important to this system overall, rather than in particular instances. Additionally, a drawback of metric-space analyses is that they assume that a particular model (metric) of neural activity is the correct description of neural encoding. As discussed more fully in Materials and Methods, metric-space approaches therefore provide only a lower bound on mutual information [23],[25]. Put another way, metric-space analyses assume that the differences between spike trains can be fully represented by a particular set of parameters, which in our case include the temporal separation between nearest-neighbor spike times (Figure 2a). However, if information is contained in higher order aspects of the spike trains that cannot be captured by these parameters (e.g., patterns that extend over multiple spikes), then metric-space analyses can significantly underestimate the true information contained in the neural code. We therefore estimated the amount of information that can be learned about the acoustic group by directly observing the spiking pattern at different temporal resolutions (Figure 3a), without assuming a metric structure, similar to prior approaches in sensory systems [8],[10]. We used the Nemenman-Shafee-Bialek (NSB) estimator to quantify the mutual information [26],[27]. As described in Materials and Methods, this technique provides minimally biased information estimates, quantifies the uncertainty of the calculated information, and typically requires substantially less data for estimation than many other direct estimation methods [26]. Nevertheless, the NSB technique requires significantly larger datasets than metric-space methods. We therefore directly computed mutual information using the subset (41/125) of cases where the recordings were long enough to gather sufficient data to be analyzed with this method.
We found that the mutual information between neural activity and vocal behavior rose dramatically as temporal resolution increased. As shown in Figure 3b, when averaged across all 41 cases analyzed using the NSB technique, mutual information was relatively low when only spike counts were considered (i.e., for ms). Across the four methods of grouping trials based on syllable acoustics, mutual information between spike counts and acoustic output ranged from 0.007 to 0.017 bits (with standard deviations of ∼0.012), which is not significantly different from zero given that all mean information estimates were within ∼1 SD of zero bits. If information about motor output were represented only in spike counts within the 40 ms premotor window, then mutual information at dt<40 would be equal to that found at dt = 40 (dashed lines in Figure 3b); note that this is true despite the increase in word length at smaller dt [8],[10]. However, in all analyses mutual information increased as time bin size dt decreased and reached a maximum value at dt = 1 ms, the smallest bin size (and thus greatest temporal resolution) we could reliably analyze. At 1 ms resolution, mutual information ranged from 0.131 to 0.188 (with standard deviations of ∼0.06) bits across the four analyses performed. These values of mutual information correspond to d′ values near zero at dt = 40 ms and to d′ values between 0.9 and 1.1 at 1-ms resolution (Figure 3b, right-hand axis). Additionally, although Figure 3b shows the results of analyzing data from single-unit and multiunit recordings together, we found very similar results when single- and multiunit data were considered separately (see Figure S2 and Data S1 and S2). These results indicate that far more information about upcoming vocal behavior is available at millisecond timescales and suggest that small differences in spike timing can significantly influence motor output. Therefore, although in some individual cases more information may be available from a rate code in the metric-space analysis (Figure 2d, empty bars), across the population of RA neurons much more information is present in millisecond-scale spike timing. Similarly, note that although in the direct information calculations (Figure 3) mutual information is averaged across all neural recordings, the information at different timescales varied across different neurons (e.g., information at dt = 5 ms in some recordings was greater than information at dt = 1 ms in other recordings). The low mutual information present at dt = 40 ms, for example, therefore reflects the fact that datasets with higher (relative to other datasets) information in the spike count are greatly outnumbered by cases with very low information at dt = 40.
We performed further analysis to investigate whether the information present at dt = 1 ms reflects differences in burst onset times or differences in the pattern of spikes within bursts (either of which could account for our results). To do so, we performed an alternate analysis (see Materials and Methods, “Inter-Spike Interval Analysis”) in which we calculated the mutual information conveyed by sequences of inter-spike intervals (at a temporal resolution of 1 ms) rather than by absolute spike times. Representing neural activity in this way removes all information about burst onset time and instead quantifies only the mutual information between inter-spike intervals and the acoustic output. In this alternate analysis, information estimates ranged from 0.148 to 0.172 (with standard deviations of ∼0.06) bits across the four different behavioral groupings, and thus were nearly identical to those obtained at dt = 1 ms in our primary analysis. This finding suggests that the information contained in spike timing patterns is carried primarily through the structure of spike timing within bursts rather than by burst onset times.
The results shown in Figure 3 demonstrate that millisecond-scale differences in spike timing can encode differences in behavior. To highlight these timing differences, we examined particular “words” (spike patterns) and considered how different timing patterns could predict vocal acoustics. Figure 4a and 4b each show eight different words from two different single-unit recordings, color-coded according to the behavioral group in which each word appears most frequently. All words shown in Figure 4 contain the same number of spikes, and thus are identical at the time resolution of dt = 40 ms (Figure 3a). In the example shown in Figure 4a, a distinct set of spike timing patterns predicts the occurrence of low-pitched (group 1) or high-pitched (group 2) syllable renditions. In Figure 4b, behavioral groupings are performed in the three-dimensional acoustic space and similarly show that distinct spike timing patterns can predict vocal acoustics. In some cases, the timing patterns associated with behavioral groups share intuitive features. For example, the words associated with higher pitch in Figure 4a (blue boxes in grid) have shorter inter-spike intervals (but the same total number of spikes) compared with words associated with lower pitch (Figure 4a, red boxes), suggesting that fine-grained interval differences drive pitch variation. However, in other cases (e.g., Figure 4b) no such common features were apparent. Future studies incorporating realistic models of motor neuron and muscle dynamics are therefore required to understand how the precise timing patterns in RA can evoke differences in vocal behavior.
We compared the maximum information available from the metric-space analysis (see Materials and Methods), which is Īmax = 0.10, to the information available at the smallest dt = 1 ms in the direct information calculation, MINSB = 0.16 bits. Reassuringly, the peak information available from the direct method is of the same order of magnitude but somewhat larger than that computed independently in the metric-space analysis. This finding points at consistency between the methods and yet suggests that additional information may be present in higher order spike patterns that cannot be accounted for by a metric-space analysis, namely in temporal arrangements of three or more spikes. (Note, however, that in a small number of cases we were unable to compute mutual information at the finest timescales using the direct method, possibly leading to small biases in our estimates of mutual information at dt = 1 ms; see Materials and Methods). Similarly, a common technique in metric-space analysis is to estimate the “optimal time scale” of encoding as 1/qmax (although other authors suggest that such estimates may be highly imprecise [25]). In our dataset, the median value of qmax was 0.3 ms−1, suggesting that spike timing precision is important down to 1/qmax∼1 ms, which is again in agreement with the direct estimation technique.
We computed the mutual information between premotor neural activity and vocal behavior using two well-established computational techniques. A metric-space analysis demonstrated that spike timing provides a better prediction of vocal output than spike rate in a significant majority of cases (Figure 2). A direct computation of mutual information, which was only possible in the subset of recordings that yielded relatively large datasets, revealed that the amount of information encoded by neural activity was maximal at a 1 ms timescale, while the average information available from a rate code was insignificant (Figure 3). It also suggested that information in the spike trains may be encoded in higher order spike patterns.
Although previous studies have shown that bursts in RA projection neurons are aligned in time to the occurrence of particular song syllables [15],[18], ours is the first demonstration to our knowledge that variations in spike timing within these bursts can predict trial-by-trial variations in vocal acoustics. These acoustic variations are thought to underlie vocal learning ability in songbirds. A number of studies have demonstrated that nucleus LMAN (the lateral magnocellular nucleus of the anterior nidopallium), the output nucleus of the anterior forebrain pathway (AFP) and an input to RA (Figure 1a), both generates a significant fraction of vocal variability and is required for adaptive vocal plasticity in adult birds [28]–[30]. A significant question raised by our results therefore concerns the extent to which LMAN inputs can alter the timing of spikes in RA. Recent work has shown that spike timing patterns in LMAN neurons encode the time during song [31]. Future studies might address whether the observed patterns in LMAN spiking can also predict acoustic variations, and lesion or inactivation experiments could quantify changes in the distribution of firing patterns in RA after the removal of LMAN inputs [32].
Our results indicate that spike timing in cortical motor networks can carry significantly more information than spike rates. Equivalently, these findings suggest that limiting the analysis of motor activity to spike counts can lead to drastic underestimates of information. This contrast is illustrated by a comparison of the present analysis and our prior study examining correlations between premotor spike counts and the acoustics of song syllables [14]. In that earlier study, we found that spike rate predicted vocal output in ∼24% of cases, a prevalence similar to the proportion of rate cases observed in the metric-space analysis and far smaller than the prevalence of temporal cases (Figure 2). Similarly, direct computations of mutual information (Figure 3) show that a purely rate-based analysis would detect only a small fraction of the information present in millisecond-scale timing. Therefore our central finding—that taking spike timing into account greatly increases the mutual information between neural activity and behavior—suggests that correlation and other rate-based approaches to motor encoding might in some cases fail to detect the influence of neural activity on behavior.
As shown in Figure 3, we found that spike timing at the 1 ms timescale provides an average of ∼0.16 bits out of a possible 1.0 bit of information when discriminating between two behavioral groups. While this value is of course less than the maximum possible information, it is important to note that this quantity represents the average information available from a single neuron. A number of studies in sensory systems have demonstrated that ensembles of neurons can convey greater information than can be obtained from single neurons [33]. While our dataset did not include sufficient numbers of simultaneous recordings to address this issue, future analyses of ensemble recordings could test the limits of precise temporal encoding in the motor system.
Temporal encoding in the motor system could also provide a link between sensory processing and motor output. Prior studies have shown that different auditory stimuli can be discriminated on the basis of spike timing in auditory responses [11],[34],[35], including those in area HVC, one of RA's upstream inputs [36]. Our results demonstrate that in songbirds, temporally precise encoding is present at the motor end of the sensorimotor loop. Propagating sensory-dependent changes in spike timing into motor circuits during behavior might therefore underlie online changes in motor output in response to sensory feedback [37],[38] or serve as a substrate for long-term changes in motor output resulting from spike timing-dependent changes in synaptic strength [19],[39]–[41].
While the existence of precise spike timing is strongly supported for a variety of sensory systems, a lingering question is how downstream neural networks could use the information that is present at such short timescales, and hence whether the animal's behavior could be affected by the details of spike timing. Although theoretical studies have suggested how downstream neural circuits could decode timing-based spike patterns in sensory systems [42], the general question of whether the high spiking precision in sensing, if present, is an artifact of neuronal biophysics or a deliberate adaptation remains unsettled [43].
In motor systems, in contrast, spike timing differences could be “decoded” via the biomechanics of the motor plant, thereby transforming differences in spike timing into measureable differences in behavior. In a wide range of species [44]–[47], the amplitude of muscle contraction can be strongly modulated by spike timing differences in motor neurons (i.e., neurons that directly innervate the muscles) owing to strong nonlinearities in the transformation between spiking input and force production in muscle fibers. Furthermore, biomechanical studies have shown that vocal muscles in birds have extraordinarily fast twitch kinetics and can reach peak force production in less than 4 ms after activation [20],[21], suggesting that the motor effectors can transduce millisecond-scale differences in spike arrival into significant differences in acoustic output. Finally, in vitro and modeling studies have quantified the nonlinear properties in the songbird vocal organ, demonstrating that small differences in control parameters can evoke dramatic and rapid transitions between oscillatory states, suggesting again that small differences in the timing of motor unit activation could dramatically affect the acoustics of the song [48],[49]. Future studies that model the dynamics of brainstem networks downstream of RA as well as the mechanics of the vocal organ could address how particular spiking patterns in RA (such as those shown in Figure 4) might drive variations in acoustic output.
Our results demonstrate that the temporal details of spike timing, down to 1 ms resolution, carry about ten times as much information about upcoming motor output compared to what is available from a rate code. This is in marked contrast to sensory coding [8],[10], where the information from spike patterns at millisecond resolution is often about double that available from the rate alone. For this reason, the most striking result of our analysis might be that precise spike timing in at least some motor control systems appears to be even more important than in sensory systems. In summary, although future work in both sensory and motor dynamics is needed to fully explicate how differences in spike timing are mapped to behavioral changes, our findings, in combination with previous results from sensory systems, represent the first evidence, to our knowledge, for the importance of millisecond-level spiking precision in shaping behavior throughout the sensorimotor loop.
All procedures were approved by the Emory University Institutional Animal Care and Use Committee.
To measure the information about vocal output conveyed by motor cortical activity at different timescales, we recorded the songs of Bengalese finches while simultaneously collecting physiological data from neurons in RA. We then quantified the acoustics of individual song syllables and divided the iterations of each syllable into “behavioral groups” on the basis of acoustic features such as pitch, amplitude, and spectral entropy. Mutual information was then computed using two complementary techniques. First, we used a metric-space analysis [23] to quantify how well the distance between pairs of spike trains can be used to classify syllable iterations into behavioral groups. Second, we used a direct calculation of mutual information [10],[26],[27],[50] to produce a minimally biased estimate of the information available at different timescales.
Single-unit and multiunit recordings of RA neurons were collected from four adult (>140 days old) male Bengalese finches using techniques described previously [14]. All procedures were approved by the Emory University Institutional Animal Care and Use Committee. Briefly, an array of four or five high-impedance microelectrodes was implanted above RA nucleus. We advanced the electrodes through RA using a miniaturized microdrive to record extracellular voltage traces as birds produced undirected song (i.e., no female bird was present). We used a previously-described spike sorting algorithm [14] to classify individual recordings as single-unit or multiunit. In total, we collected 53 RA recordings (19 single-unit, 34 multiunit), which yielded 34 single-unit and 91 multiunit “cases,” as defined below. Based on the spike waveforms and response properties of the recordings, all RA recordings were classified as putative projection neurons that send their axons to motor nuclei in the brainstem [14],[15],[51]. A subset of these recordings has been presented previously as part of a separate analysis [14].
We quantified the acoustics of each song syllable as described in detail previously [14]. Briefly, for each syllable we measured vocal acoustics at a particular time (relative to syllable onset) when spectral features were well defined (Figure 1b, red line). Syllable onsets were defined on the basis of amplitude threshold crossings after smoothing the acoustic waveform with a square filter of width 2 ms; therefore the measurement error of our quantification of syllable onset time is on the order of milliseconds. Note that this uncertainty cannot account for our results, since millisecond-scale jitter in syllable onset (and thus burst timing) will decrease, rather than increase, the amount of information present at fine timescales. We quantified the fundamental frequency (which we refer to here as “pitch”), amplitude, and spectral entropy by analyzing the acoustic power spectrum at the specified measurement time during each iteration of a song syllable. We selected these three acoustic features because they capture a large percentage of the acoustic variation in Bengalese finch song [14]. In the example syllable illustrated in Figure 1b (top), the band of power at ∼4 kHz is the fundamental frequency. Furthermore, each sound recording was inspected for acoustic artifacts unrelated to vocal production and these trials, which constituted less than 1% of the total data, were discarded to minimize potential measurement error. For each iteration of each syllable, we analyzed spikes within a temporal window prior to the time at which acoustic features were measured. The width of this window was selected to reflect the latency with which RA activity controls vocal acoustics. Although studies employing electrical stimulation have produced varying estimates of this latency [52],[53], a single stimulation pulse within RA modulates vocal acoustics with a delay of 15–20 ms [54]. We therefore set the premotor window to begin 40 ms prior to the time when acoustic features were measured and to extend until the measurement time (Figure 1b, red box). This window therefore includes RA's premotor latency [14],[22] and allows for the possibility that different vocal parameters have different latencies.
While grouping spike trains is straightforward in many sensory studies, where different stimuli are considered distinct groups, we face the problem of continuous behavioral output in motor systems. We took two approaches to binning continuous motor output into discrete classes. First, we considered only a single acoustic parameter and divided the trials into equally sized groups using all of the data. For example, Figure 1c shows trials divided into two behavioral groups based on one parameter (pitch). In addition to pitch, separate analyses also used sound amplitude or spectral entropy to divide trials into groups. In the second approach (which we term “3D acoustics”) (Figure 1d), we used k-means clustering to divide trials into groups. Clustering was performed in the three-dimensional space defined by pitch, amplitude, and entropy, with raw values transformed into z-scores prior to clustering. Note that both approaches allow us to divide the dataset into an arbitrary number of groups (parameter N, see “Discrimination analysis” below). Our primary analysis divided trials into N = 2 groups since a smaller N increases statistical power by increasing the number of data points in each group. However, alternate analyses using greater N yielded similar conclusions (see Results).
In previous studies, metric-space analysis has been used to probe how neurons encode sensory stimuli (for a review, see [55]). The fundamental idea underlying this approach is that spike trains from different groups (e.g., spikes evoked by different sensory stimuli) should be less similar to each other than spike trains from the same group (spikes evoked by the same sensory stimulus). In the present study, we adapt this technique for use in the vocal motor system to ask how neurons encode trial-by-trial variations in the acoustic structure of individual song syllables. To do so, we divide the iterations of a song syllable into “behavioral groups” based on variations in acoustic structure (Figure 1c). We then construct a “classifier” to ask how accurately each spike train can be assigned to the correct behavioral group using a distance metric that quantifies the dissimilarity between pairs of spike trains [24]. As described in detail below, the classifier attempts to assign each trial to the correct behavioral group on the basis of the distances between that trial's spike train and the spike trains drawn from each behavioral group. Crucially, the distance metric is parameterized by q, which reflects the importance of spike timing to the distance between two spike trains. This method therefore allows us to evaluate the contribution of spike timing to the performance of the classifier, and thus to the information contained in the spike train about the behavioral group.
In addition to the metric-space analysis described above, we also directly calculated the mutual information between song acoustics and neural activity [10]. Whereas metric-space analysis makes strong assumptions about the structure of the neural code, the direct approach is model-independent [10],[56]. Specifically, spike train distance metrics assume that spike trains that have spike timings closer to each other are linearly more similar than spike trains whose timings are more different. As with all assumptions, the methods gain extra statistical power if they are satisfied, but they may fail if the assumptions do not hold. The direct method simply considers distinct patterns of spikes at each timescale (which can vary in total spike number, the pattern of inter-spike intervals, burst onset time, etc.), without assigning importance to specific differences. Crucially, direct methods allow us to estimate the true mutual information, whereas the mutual information computed from a metric-space analysis represents only a lower bound on this quantity [3]. However, because the direct method is a model-independent approach that does not make strong assumptions about the neural code, it requires larger datasets to achieve statistical power.
To determine whether there is information about acoustics in the precise timing of spikes, we compared the information between neural activity and behavioral group following discretization of the spike trains at different time resolutions. For a time bin of size dt, each ms-long spike train was transformed into a “word” with 40/dt symbols, where different symbols represent the number of spikes per bin. The mutual information is simply the difference between the entropy of the total distribution of words and the average entropy of the words given the behavioral group :(3)
could be quantified exactly if the true probability distributions , , and were known:(4)
However, estimating these distributions from finite datasets introduces a systematic error (“limited sample bias” or “undersampling bias”) that must be corrected [57]. There are several methods to correct for this bias, but most assume that there is enough data to be in the asymptotic sampling regime, where each typical response has been sampled multiple times. As we increase the time resolution of the binning of the spike train, the number of possible neural responses increases exponentially, and we quickly enter the severely undersampled regime where not every “word” is seen many times, and, in fact, only a few words happen more than once (which we term a “coincidence” in the data). We therefore employed the NSB entropy estimation technique [26],[27], which can produce unbiased estimates of the entropies in Equation 3 even for very undersampled datasets.
The NSB technique uses a Bayesian approach to estimate entropy. However, instead of using a classical prior, for which all values of the probability of spiking are equally likely, NSB starts with the a priori hypothesis that all values of the entropy are equally likely. This approach has been shown to reliably estimate entropy in the severely undersampled regime (where the number of trials per group is much less than the cardinality of the response distribution) provided that the number of coincidences in that data is significantly greater than one. This typically happens when the number of samples is only about a square root of what would be required to be in the well-sampled regime [26],[50].
This method often results in unbiased estimates of the entropy, along with the posterior standard deviation of the estimate, which can be used as an error bar on the estimate [50]. On the other hand, we know that no method can be universally unbiased for every underlying probability distribution in the severely undersampled, square-root, regime [58]. Thus there are many underlying distributions of spike trains for which NSB would be biased. Correspondingly, the absence of bias cannot be assumed and must instead be verified for every estimate, which we do as described below.
A priori, we restricted our analysis to cases in which the number of trials was large enough (>200) so that the number of coincidences would likely be significantly greater than 1. Of our 125 datasets, 41 passed this size criterion. We emphasize that no additional selection beyond the length of recording was done. Since recording length is unrelated to the neural dynamics, we expect that this selection did not bias our estimates in any way. The NSB analysis was performed using N = 2 behavioral groups, since increasing the number of groups greatly decreased the number of coincidences and increased the uncertainty of the entropy estimates (not shown). Additionally, because NSB entropy estimation assumes that the words are independent samples, we verified that temporal correlations in the data are low. To do this, we used NSB to calculate the entropy of four different halves of each dataset: the first half of all trials, the second half, and the two sets of every other trial, where the second set is offset from the first set by one trial. We found that the difference in mean entropy between the first half and second half data was very similar to the difference between the two latter sets. Any temporal correlations in our neural data are therefore very low and thus unlikely to affect entropy estimation, and the information at high spiking precision that we observe cannot just be attributed to modulation occurring on a longer time scale.
To make sure that the NSB estimator is unbiased for our data, we estimated each conditional and unconditional entropy from all available N samples, and then from αN, α<1, samples. Twenty-five random subsamples of size αN were taken and then averaged to produce . We plotted versus 1/α and checked whether all estimates for 1/α→1 agreed among themselves within error bars, indicating no empirical sample-size dependent bias [8],[10]. At temporal resolutions in the tens of milliseconds, virtually all cases showed no sample size-dependent drift in the entropy estimates, and hence the estimates from full data were treated as unbiased. As the temporal resolution increased to dt = 1, bias was visible in some cases. However, the bias could often be traced to the rank-ordered distribution of words not matching the expectations of the NSB algorithm. Specifically, some of the most common words occurred much more often than expected from the statistics of the rest of the words. Since NSB uses frequencies of common, well-sampled words to extrapolate to undersampled words, such uncommonly frequent outliers can bias entropy estimation [27]. To alleviate the problem, we followed [8] and partitioned the response distribution in a way such that the most common word was separated from the rest when it was too frequent (with “too frequent” defined as >2% of all words). We use to denote the frequency of the most common word and to denote the frequency of all other words. We then used the additivity of entropy,(5)to compute the total entropy by first estimating the entropy of the choice between the most common word and all others, , and the entropy of all of the data excluding the most common word, S2, independently using the NSB method (the entropy of the single most common word, S1, is zero). The error bars were computed by summing the individual error bars in Equation 5 in quadratures. Once the most common word was isolated in the cases in which it was “too frequent,” the resulting entropies were checked again for bias using the subsampling procedure explained above. In the majority of cases in which sample-size dependent bias was detected, this bias was removed through the partition-based entropy correction and the resulting entropy estimates were therefore empirically verified as being unbiased. In other biased cases, there were no uncharacteristically common words, and we could not perform the partition-based entropy correction. Furthermore, at the highest temporal resolutions, there were a few cases that did not have enough coincidences for the NSB algorithm to produce an estimate of entropy at all. These cases were excluded from the following steps. The total fraction of entropy estimates found to be unbiased (i.e., was either unbiased to begin with or bias was successfully corrected using partitioning) was 100% at dt = 40 ms and decreased monotonically to 72% at dt = 2 ms and 58% at dt = 1 ms. Thus, we were unable to correct entropy biases in a minority of entropy measurements at fine timescales. However, as described below (and illustrated in Figure S2), our results were nearly identical when we removed all cases with biased entropy from our dataset, demonstrating that any residual biases cannot account for our results.
We then averaged the mutual information between the spike train and the acoustic group over all cases, weighing contribution of each case by the inverse of its respective posterior variance. The variance of the mean was similarly estimated. As an example, the ten biased cases at dt = 1 for syllables grouped by pitch that could not be corrected had few coincidences and hence large error bars, so that the average value of the inverse variance for these ten cases was 14% of the average inverse variance for the 20 cases that were originally unbiased, or corrected to unbiased. Therefore, since 14% of ten cases is approximately one case, these ten biased cases together contributed about as much as one of the 20 unbiased or bias-corrected cases in the final calculation of the average mutual information. These biased cases thus cannot contribute significantly to bias in the average mutual information. This is true for the other behavioral groupings we performed as well. To confirm this, we further performed an alternate analysis (Figure S2d) in which we excluded the biased cases that could not be corrected from the final calculation of mutual information. This alternate analysis yielded very similar results as the original analysis, with a large increase in mutual information at finer temporal resolution. Our results therefore cannot be ascribed to the inclusion of some cases in which entropy biases could not be corrected. Finally, note that as described above we of necessity excluded from our analysis cases in which no coincidences occurred (since the NSB algorithm estimates entropy by counting the number of coincidences). This occurred in a small fraction of cases (13% of entropy estimates at dt = 1 ms, 5% of estimates at dt = 2 ms, <1% of estimates at dt = 5 ms, and 0% at dt = 10 ms or greater). Since cases in which no coincidences occur will likely be those with low mutual information, excluding these cases may have introduced a slight upward bias to our information estimates at fine timescales. However, information in such cases would be likely to have large uncertainties and thus would make insignificant contributions to the weighted average, and furthermore the number of excluded cases at timescales greater than 1 ms (crucially, including dt = 2 ms) is negligible. In any event, these excluded cases cannot account for either the non-zero information at fine timescales or for the dramatic increase in information observed as temporal resolution increases.
As discussed above, the metric-space and direct methods of computing mutual information differ in their underlying assumptions about the statistical structure of the neural code, and the metric-space method can only produce lower bounds on the signal-response mutual information. Therefore, comparing the values of information computed by the two methods is prone to various problems of interpretation. It is nevertheless instructive to ask whether the direct method estimates greater mutual information than the metric-space analysis, and thus if patterns of multiple spikes carry additional information beyond that in spike pairs, which is discoverable by the metric-space method. To answer this, we calculated the peak metric space information Īmax which is the mean of Imax across all cases. This is the upper bound on the information detectable through the metric-space method, as the information is maximized for each case independently, rather than finding a single optimal q for all cases.
|
10.1371/journal.pntd.0002257 | Dengue Virus Activates Membrane TRAIL Relocalization and IFN-α Production by Human Plasmacytoid Dendritic Cells In Vitro and In Vivo | Dengue displays a broad spectrum of clinical manifestations that may vary from asymptomatic to severe and even fatal features. Plasma leakage/hemorrhages can be caused by a cytokine storm induced by monocytes and dendritic cells during dengue virus (DENV) replication. Plasmacytoid dendritic cells (pDCs) are innate immune cells and in response to virus exposure secrete IFN-α and express membrane TRAIL (mTRAIL). We aimed to characterize pDC activation in dengue patients and their function under DENV-2 stimulation in vitro.
Flow cytometry analysis (FCA) revealed that pDCs of mild dengue patients exhibit significantly higher frequencies of mTRAIL compared to severe cases or healthy controls. Plasma levels of IFN-α and soluble TRAIL are increased in mild compared to severe dengue patients, positively correlating with pDC activation. FCA experiments showed that in vitro exposure to DENV-2 induced mTRAIL expression on pDC. Furthermore, three dimension microscopy highlighted that TRAIL was relocalized from intracellular compartment to plasma membrane. Chloroquine treatment inhibited DENV-2-induced mTRAIL relocalization and IFN-α production by pDC. Endosomal viral degradation blockade by chloroquine allowed viral antigens detection inside pDCs. All those data are in favor of endocytosis pathway activation by DENV-2 in pDC. Coculture of pDC/DENV-2-infected monocytes revealed a dramatic decrease of antigen detection by FCA. This viral antigens reduction in monocytes was also observed after exogenous IFN-α treatment. Thus, pDC effect on viral load reduction was mainly dependent on IFN-α production
This investigation characterizes, during DENV-2 infection, activation of pDCs in vivo and their antiviral role in vitro. Thus, we propose TRAIL-expressing pDCs may have an important role in the outcome of disease.
| Dengue is an important endemic tropical disease to which there are no specific therapeutics or approved vaccines. Currently several aspects of pathophysiology remain incompletely understood. A crucial cellular population for viral infections, the plasmacytoid dendritic cells (pDCs) was analyzed in this study. The authors found an in vivo association between the activation state of pDCs and the disease outcome. Membrane TNF-related apoptosis inducing ligand (TRAIL) expressing pDCs, representing activated pDCs, were found in higher frequency in milder cases of dengue than severe cases or healthy individuals. Detection of antiviral cytokine interferon-alpha (IFN-α) and soluble TRAIL positively correlated with pDC activation. Dengue virus (DENV) serotype-2 was able to directly activate pDCs in vitro. Under DENV stimulation TRAIL was relocalized from intracellular to pDC plasma membrane and IFN-α was highly produced. The authors suggest an endocytosis-dependent pathway for DENV-induced pDC activation. It is also highlighted here a role for exogenous IFN-α and pDCs in reducing viral replication in monocytes, one of DENV main target cells. These findings may contribute in the future to the establishment of good prognostic immune responses together with clinical manifestations/warning signs.
| Dengue is the most important arthropod-borne emerging viral disease in tropical countries due to its high morbidity and risk of mortality [1]. For example, in Brazil, dengue is a major public health problem and about two million cases were reported during 2010–2012 [2]. Dengue virus (DENV) is a single-stranded RNA virus belonging to genus Flavivirus [3], [4]. All DENV serotypes (DENV-1 to -4) may induce a broad spectrum of clinical manifestations from asymptomatic to severe clinical features, characterized by hemorrhagic manifestations and a shock syndrome [5], [6], [7]. High viral load may cause an exacerbated cytokine production that plays a key role in the generation of important physiopathological processes [8], [9]. Human monocytes/macrophages and dendritic cells are susceptible to viral replication [10], [11], [12], [13] and can release soluble mediators involved in vascular permeability and plasma leakage besides coagulation disorders [14], [15], [16], [17].
Dendritic cells link innate and adaptive immunity and play a key role in shaping effective immune responses. Two major subpopulations are described: myeloid or conventional dendritic cells (cDCs) and plasmacytoid dendritic cells (pDCs) [18], [19], [20], [21]. In contrast to cDCs, pDC are not found in homeostatic tissues but mainly in circulating blood and in lymphoid tissues [21], [22], [23]. Despite being rare cells, pDCs produce up to 1,000-fold more IFN-α than other cell types in response to virus exposure [24]. Viral activation of pDCs can be regulated by either one of the two Toll-like receptors (TLR), TLR-7 or TLR-9 [25], which are considered to be the pattern recognition receptors (PRR) for RNA [26] and DNA [27], respectively. It has been shown that cDC are efficiently infected by DENV and that viral replication blocked cDC maturation [28], [29]. However, unlike cDCs, it has been reported that pDCs are not supporting productive DENV replication [30]. Indeed, DENV can activate pDCs through cell endosomal activity and TLR-7 pathway [31]. Furthermore, dengue-infected patients had impaired pDC activation features. Indeed, absolute numbers of blood pDC were decreased [32], [33] and low levels of serum IFN-α [34] were reported.
TNF-related apoptosis-inducing ligand (TRAIL) is a pro-apoptotic molecule, which induces death of cells that express its death receptors (DR), DR4 and DR5 [35], [36]. Furthermore, IFN-α regulates TRAIL expression by several cell types [37]. Soluble or membrane TRAIL mediates apoptosis on cells that are selectively expressing DR4 and DR5, mainly killing virus-infected cells and leaving intact normal cells [38], [39]. Additionally an antiviral role was proposed for TRAIL. DENV-infected monocytes and dendritic cells display reduced viral replication when TRAIL is exogenously administered [40]. Soluble TRAIL (sTRAIL) was found in sera from dengue patients [41], but mTRAIL role and expression by DENV-2 exposed pDC to has not been investigated yet.
In this report we studied pDC activation by DENV and its consequences on viral infection. The clinical study showed that during acute phase of DF, pDCs are activated characterized by TRAIL and IFN-α markers. Indeed, the more pDC are activated the less the disease is severe. We found that DENV-2 efficiently activated TRAIL expression and IFN-α production by pDC. The microscopy study revealed that TRAIL was intracellularly stocked in resting pDC and was relocalized to plasma membrane when pDC were exposed to DENV-2. Furthermore, we showed that pDC could decrease DENV infection in monocytes mainly due to the effects of IFN-α produced. Thus pDC activation constitutes a host defense against DENV-2 infection strongly suggesting that these cells are likely beneficiating the disease outcome.
Experimental procedures with human blood have been approved by Necker Hospital Ethical Committees for human research and were done according to the European Union guidelines and the Declaration of Helsinki. Procedures were also approved by the ethical committee at Instituto de Pesquisas Clinicas Evandro Chagas, FIOCRUZ (CAAE 3723.0.000.009-08). All patients were informed of procedures and gave written consent.
Blood from HIV-1-seronegative blood bank donors was obtained anonymously from “Etablissement Français du Sang” (convention # 07/CABANEL/106), Paris, France. Forty three patients with confirmed dengue fever (Table 1) from two Brazilian Health Centers at Campo Grande, MS and Campos de Goytacases, RJ, Brazil were studied. All patients presented clinical diagnosis of dengue infection.
Dengue fever was considered mild when no warning signs (WS) or severe clinical manifestations were observed as follows. Dengue fever with WS was considered if patients presented any of the following warnings: (1) abdominal pain or tenderness; (2) persistent vomiting; (3) Clinical fluid accumulation; (4) mucosal bleeding; (5) lethargy; (6) liver enlargement more than 2 cm associated to laboratory parameters as increase in hematocrit (HCT) concurrent with rapid decrease in platelet counts (hemoconcentration or significant increase in hematocrit together with platelet counts bellow 50,000/mm3). Severe DF was considered if patient displayed fever of 2–7 days plus any of the following: (1) Evidence of plasma leakage, such as high or progressively rising hematocrit evidenced by hemoconcentration; pleural effusions or ascites; circulatory compromise or shock (tachycardia, cold and clammy extremities, capillary refill time greater than three seconds, weak or undetectable pulse, narrow pulse pressure or, in late shock, unrecordable blood pressure); (2) Significant (internal) bleeding. [42], [43]. Dengue virus infection was confirmed either by anti-dengue-IgM ELISA, serotype specific reverse transcription-polymerase chain reaction (RT-PCR) or by virus isolation as described earlier [44]. Predominant serotypes was Dengue-2 identified in DF±WS (N = 10) and Severe DF (N = 3) but Dengue-1 was also identified in DF±WS patients (N = 6).
Dengue virus type 2 (strain Thailand/16681/1984) [45] was used for virus stock preparation as described elsewhere [46]. Briefly, Aedes albopictus cell clone C6/36 (CRL-1660, ATCC) were maintained at 28°C in Dulbecco's modified Eagle Medium (Gibco/Life Technologies, Foster City, CA, USA) with sodium bicarbonate (Sigma-Aldrich, St. Louis, MO, USA) and supplemented with 5% fetal bovine serum (Hyclone, Logan, UT, USA), 1% penicillin-streptomycin-glutamine (Gibco), 0,5% non-essential amino acids (Gibco) and 10% tryptose phosphate broth (Sigma). C6/36 cell monolayers were infected with DENV-2 and cell culture supernatants were harvested 8 days later when cytopathic effect was observed. A purified DENV-2 stock was obtained by ultracentrifugation at 100,000 g for 1 h and set to a final volume 20 times smaller than initial (see also Fig. S1) [47], [48]. Titration was performed in C6/36 cells using a standard TCID50 (50% tissue culture infective dose) assay as described elsewhere [49]. Uninfected flasks were maintained, also purified and used as negative control (MOCK). Infectivity of ultracentrifuged virus inoculum (UC) was comparable with the original C6/36 supernatant (SNDT) because infection rates obtained with the dilution 1/100 (UC) is similar to the dilution 1/5 (SNDT) as shown in Fig. S1.
Cryopreserved peripheral blood mononuclear cells (PBMC) from patients or healthy donors were obtained from density gradient centrifugation of heparinized blood with lymphocyte separation medium (StemCell Technologies, Grenoble, FR). In vitro experiments were performed using fresh PBMC, which were obtained from blood bank donors and isolated as mentioned above. PDCs and monocytes were purified using Human plasmacytoid DC Negative Isolation Kit and Human CD14+ monocytes Isolation Kit, respectively (StemCell Technologies). Cells were cultured in RPMI 1640 (Invitrogen, Gaithersburg, MD, USA) containing 10% fetal bovine serum (Hyclone) and 1% penicillin-streptomycin-glutamine (Gibco) at 37°C in a humidified 5% CO2 chamber according to protocol.
Freshly purified pDCs were cultured with DENV-2 at approximately MOI 4 to 20, mock for 18 hours (overnight). Chloroquine (Sigma-Aldrich) was used at 5 µM/well and added before viral stimulation. Cells were harvested and assessed for pDC cell markers and membrane TRAIL expression or plated on coated slides for 3D microscopy. Supernatant was stored at −70°C for cytokine detection. Monocyte infection was performed as already described [46]. Briefly, freshly isolated monocytes were plated overnight followed by infection with DENV-2 at MOI 10, mock or not infected for 48 hours. Soluble human recombinant IFN-α (PBL International, Piscataway, NJ, USA) was added 18 hours before viral infection at 100 IU/mL. For autologous coculture assay, monocytes were cultured overnight in media, meanwhile pDCs were chloroquine-treated or not and then stimulated overnight with CpG A 2216 (InvivoGen, San Diego, CA, USA) at 5 µM or DENV-2 at MOI 20, or not stimulated. Monocytes were then infected with DENV-2 (MOI 10) and pDCs were added at ratio 1∶5 pDC/monocytes as explained in Fig. S2. Cells were harvested and assessed for intracellular DENV antigens.
Antibodies for fluorescein isothiocyanate (FITC)-conjugated anti-CD123 or BDCA-2 (Miltenyi Biotec, Auburn, CA), Phycoerythrin (PE)- conjugated CD11c (IOTest/Beckman Coulter, Marseille, FR), Allophycocyanin (APC)-conjugated anti-BDCA-4 (Miltenyi Biotec) and Allophycocyanin-Cy7 (APC-Cy7)-conjugated anti-CD14 (BD Biosciences, San Jose, CA), Vioblue-conjugated anti-CD4 (Miltenyi Biotec), V500 anti-CD3 (BD Biosciences) or with appropriate isotype-matched control antibodies (at 5 mg/mL each) in PBS containing 2% fetal bovine serum (Hyclone) and 2 mM EDTA (Gibco). Human PBMCs or isolated monocytes/pDCs were incubated for 20 min at 4°C with antibody cocktails. Cells were washed twice in ice-cold PBS and flow cytometry acquisition was performed on FACSCanto 7 colors or FACS Aria 13 colors flow cytometers using FACSDiva software (BD Biosciences). CD3− CD4+ CD14− CD123+/BDCA-2+ BDCA-4+ gated cells were then tested for the expression of surface markers using PE-labeled anti-TRAIL (BD Biosciences). Mosquito C6/36 cell line monolayers were washed with PBS-1% bovine serum albumin (Sigma) and incubated for 60 min at 4°C with purified anti-DENV-complex (Millipore, Billerica, MA, USA) then 30 min with goat anti-mouse Alexafluor647 (Molecular Probes/Life Technologies) and fixed. Intracellular antigen staining for C6/36 or cocultures was performed using 2% paraformaldehyde (Sigma) followed by antibodies staining steps with 0,1% saponin (Sigma) buffer. Cells were analyzed by C6 Cytometer (Accuri/BD Biosciences). FlowJo software (Treestar, Ashland, OR, USA) was used to analyze flow cytometry data.
Cells were plated on poly-L-lysine (Sigma)-coated slides and then fixed in 4% paraformaldehyde (Sigma), quenched with 0.1 M glycine (Sigma). Cells were blocked and incubated in permeabilizing buffer containing 0.1% saponin (Sigma) with mouse anti-TRAIL (clone RIK-2, eBioscience, San Diego, CA) or mouse anti-DENV (clone D3-2H2-9-21, Millipore). TRAIL and DENV staining were revealed using a secondary donkey anti-mouse IgG-Cy3 (Jackson ImmunoResearch, West Grove, PA, USA). Nucleus was stained using DAPI (Molecular Probes/Life Technologies). Mounted slides were scanned with a Nikon Eclipse 90i Upright microscope (Nikon Instruments Europe, Badhoevedorp, The Netherlands) using a 100× Plan Apo VC piezo objective (NA 1.4) and Chroma bloc filters (ET-DAPI, ET-Cy3) and were subsequently deconvoluted with a Meinel algorithm and 8 iterations and analyzed using Metamorph (MDS Analytical Technologies, Winnersh, UK). Overlays were: TRAIL or DENV/DAPI/Trans. ImageJ (NIH, Bethesda, MD, USA) plugin 3D interactive surface plot was used on overlay stack on pDC stained with TRAIL or DENV/DAPI. Quantity of TRAIL and DENV-2 were determined using the measure and label plugin (ImageJ). C6/36 mosquito cell line were plated on slides and fixed with cold acetone. Mosquito cells were stained with mouse anti-DENV complex (Millipore) in PBS-1%bovine serum albumin (Sigma), washed twice with PBS. DENV E protein was revealed with goat anti-mouse Alexafluor488 (Molecular Probes/Life Technologies). Slides were mounted with ProLong Gold with DAPI (Molecular Probes/Life Technologies) and visualized at Evosfl Microscope (AMG, Bothell, WA, USA).
Supernatants of pDCs/monocytes or cocultures in presence of DENV-2 or negative controls as well as acute phase plasma from dengue patients were tested for multispecies soluble IFN-α by ELISA (PBL International) according to the manufacturer's instructions. Plasma samples were also tested for soluble TRAIL by ELISA (R&D Systems, Minneapolis, MN, USA)
Experiments were repeated at least four times. P values (P) were determined using a two-tailed Student's t test for in vitro data and nonparametric Mann-Whitney test for patient data. P<0.05 was considered statistically significant. Univariate distributions of flow cytometric data were performed by probability binning, in 300 bins using FlowJo software [50].
We studied a cohort of DENV infected patients and classified them regarding the severity of the disease. Detailed demographic, clinical, and laboratorial data from dengue patients are summarized in Table 1. From 43 patients enrolled, 10 were classified as severe DF and the remaining as DF including those with warning signs for severity (WS), according to latest WHO classification [42], [43]. In order to explore pDC activation by DENV infection, we first characterized the CD4+/CD14−/BDCA-2/4+/CD123+ pDC frequency/profile in 40 patients compared to 20 healthy controls (figure 1A).
As described by others [51], pDC frequencies in healthy individuals range from 0.2% to 0.8% of peripheral blood mononuclear cells (PBMCs). We observed no significant differences in pDC frequencies among healthy donors, DF±WS patients or Severe DF patients (figure 1B). We then observed that mTRAIL expression on pDC was increased in DF±WS patients compared to healthy controls or severe DF cases (figure 1C). Therefore, pDCs become activated in dengue patients with regard to mTRAIL expression.
Although pDCs are not the only IFN-α producers, activated pDCs can support a 1000-fold greater production of this factor than other cell types. We next sought a correlation of IFN-α with severity. Soluble IFN-α level in plasma samples from the studied population was determined by ELISA. Similarly to TRAIL+ pDC frequency, we found that DF patients exhibit higher levels of IFN-α compared to healthy controls or Severe DF patients (figure 1D). Indeed, we found a positive correlation between IFN-α levels and TRAIL+ pDCs (Spearman r = 0.36, p<0.05). To further determine the IFN-α role, we quantified soluble TRAIL (sTRAIL) levels that is produced by immune cells and is induced by type I IFN. Similarly to previous data, DF±WS patients displayed elevated sTRAIL in contrast to healthy controls or severe DF patients (figure 1E). Moreover, a strong positive correlation between TRAIL+ pDCs and sTRAIL was determined (Spearman r = 0.60 p<0.005). PDC activation during dengue fever, elevated IFN-α and TRAIL levels is therefore associated with mild dengue fever.
PDC activation by DENV-2 was shown to occur by TLR-7 stimulation after endocytosis [31] and this pathway was crucial for IKpDC transformation by HTLV-1 [48]. To assess pDC activation by DENV-2, peripheral blood mononuclear cells (PBMCs) from healthy donors were stimulated overnight with virus. Initially, we observed that DENV-2 from mosquito cell line supernatant (SNT) promoted a trend, however not statistically significant, in TRAIL detection on pDC surface after viral stimulation in PBMCs, compared to unstimulated or mock-stimulated pDCs (figure 2A and B). Thereafter, an ultracentrifugation of DENV-2 viral stock was performed in order to concentrate viral particles, increasing MOI (figure S1). The DENV-2 infectivity was assayed for both viral stocks by infecting the mosquito cell line C6/36 and comparing them in serial dilutions. Viral antigens were detected inside cells inoculated with concentrated DENV-2 (UC) as early as 48 hours and at higher frequencies than the non-concentrated supernatant indicating that the concentrated virus had enhanced replication rates and it was intracellularly present as detected by immunofluorescence microscopy and flow cytometry (figure 2C and D). This viral stock (DENV-2 UC) was therefore adopted for assessing DENV-2 induced pDC activation in all experiments described in the present work.
Therefore, using purified virus in PBMC cultures we observed an increase of mTRAIL detection (figure 2E and F) in 41%±6% of pDCs (CD4+ CD14− BDCA4+ CD123+) compared with less than 10% TRAIL+ pDCs on mock or unstimulated conditions (p<0.05). To exclude pDC bystander activation and to confirm that DENV-2 is directly inducing mTRAIL on pDCs, we assayed purified pDC for TRAIL and IFN-α production. Purified pDC were exposed to different multiplicities of infection (MOI) for DENV-2 and we observed an increased inoculum-dependency of mTRAIL detection by virus-activated pDCs (figure 2G). The mTRAIL displayed on cell surface was mostly blocked when pDCs were pre-treated with chloroquine, an endosomal blocker of TLR activation, supporting the concept of an endocytosis-TLR-dependent TRAIL activation (figure 2H). To further characterize DENV-2-induced activation of pDCs, we measured IFN-α production in purified pDC cultures supernatants. DENV-2-stimulated pDCs produced approximately 10,000-fold more IFN-α than mock-treated or not stimulated pDCs. Chloroquine pre-incubation abrogated most DENV-2-induced IFN-α production (figure 2I). These results confirm that DENV-2 is able to activate pDCs in vitro through endocytosis pathway, responding by TRAIL expression and IFN-α production.
To better characterize DENV-2-activated pDCs, we analyzed them by 3-dimension (3D) microscopy. Focal plane analysis revealed the presence of intracellular TRAIL expression in unstimulated pDC (figure 3A, upper panels), confirming our cytometry data and our previous study [48]. Images also revealed some ‘peripheral’ TRAIL expression that did not seem to be localized in the cytoplasm but rather on the membrane (figure 3A, middle panels). TRAIL expression profile in DENV-2-stimulated pDC did not seem to differ from unstimulated cells, even if TRAIL appeared to be decreased in the cytoplasm at the expense of “peripheral” TRAIL (figure 3A, middle panels). However, it remained hard to distinguish between intracellular and membrane TRAIL profile expression in both conditions without the use of a membrane marker. The blocking of endosomal acidification by chloroquine use revealed the same profile as mock-stimulated pDCs.
Thus, to better characterize TRAIL localization in pDCs, 3D reconstruction (focal plan, XZ and YZ-stacks) analysis was performed (figure 3B–D). 3D interactive surface plot plugin of ImageJ software combined with phase contrast acquisition allowed us to visualize with precision internal or external localization of TRAIL (membrane delimitation) (figure 3B). This combined analysis clearly showed intracytoplasmic TRAIL repartition of mock stimulated pDC (figure 3B and C upper panel). DENV-2-stimulated pDC (figure 3C, middle panel) mainly harbored membrane TRAIL localization in contrast to the restrictive intracellular TRAIL expression of pDC from mock stimulated cells (figure 3A–B, right panels). The addition of the endocytosis-TLR pathway inhibitor chloroquine induced an intracellular blocking of TRAIL by DENV-2 exposed pDC. Quantification of membrane vs. intracellular TRAIL in pDC by 3D microscopy in independent assay sets demonstrates a clear shift from intracellular to membrane TRAIL location under DENV-2 exposure (figure 3D). We observed that almost all mock stimulated pDCs express only intracellular TRAIL. Chloroquine-treatment of DENV-2-stimulated pDC cultures prevented most TRAIL membrane co-localization on pDCs. Considered together, these results demonstrate that DENV-2 induces TRAIL relocalization from intracellular compartment to pDC plasma membrane.
We also attempted to detect virus inside pDCs by 3D microscopy. Because virus is rapidly degraded in endosomes by acid-activated proteases, we analyzed DENV-2 localization as early as 2 hours of viral stimulation. Focal plane images revealed that DENV-2 envelope protein was detected in close proximity to pDC periphery. In contrast, DENV-2 seemed to be intracellular in chloroquine treated pDC (figure 4A, upper panel). However, after overnight culture, DENV-2 labeling was exclusively detected in chloroquine-treated cells (figure 4A, lower panel). As described above for TRAIL detection, a 3D interactive surface plot analysis was performed and clearly showed that DENV-2 was co-localized in the cell membrane after 2 hours of stimulation (figure 4B, upper panel). We did not detect any virus in pDCs, suggesting a complete viral degradation within lysosomes either overnight (figure 4B, middle panel) or after a 2 hour-stimulation. However, DENV-2 particles were detected inside chloroquine-treated pDCs, indicating that chloroquine would probably neutralize acid proteases allowing viral antigen detection within most pDCs (figure 4B, lower panel and 4C). Therefore, within the same stimulus, pDCs exhibit TRAIL relocalization at the time point when no virus was detected, supporting our data for endosomal activation of TRAIL pathway.
Because viral load is considered to be an important factor in dengue severity [8], we next studied the role of pDCs in viral replication. For that purpose, we used primary autologous human monocytes that allow efficient DENV-2 replication in order to assess whether pDC could inhibit viral replication or not. Analysis of purified monocytes infected for 48 hours revealed in 2D microscopy a robust intracellular but not nuclear staining of DENV proteins (figure 5A, lower panel), consistent with flavivirus replication cycle [4]. Considering that pDCs produce high levels of IFN-α upon DENV-2 stimulation, we evaluated its antiviral effect. Monocytes that were pre-treated with IFN-α 24 hours before DENV-2 incubation showed a great reduction in viral antigen detection compared to untreated cells (figure 5B). Quantification by microscopy of DENV-2 positive/negative cells showed that IFN-α treatment reduced by 80% (p<0.001) the number of DENV-2 positive cells (figure 5C) and the same reduction was observed by flow cytometry (figure 5D). We also observed a low production of IFN-α by DENV-infected monocytes and confirmed IFN-α on supernatants of monocytes pre-treated with the cytokine (figure S2). These data are supporting that IFN-α has a restricting antiviral role during DENV infection.
Thus, to determine the potential effect of pDC on DENV infection within monocytes, we cocultured pDCs with infected monocytes. First, we confirmed that viral antigens were only detected in monocytes during cocultures, as only CD14+CD11c+ DENV-2-infected monocytes display DENV antigens compared to CD14−CD11c−CD123+ DENV-2-stimulated pDCs (figure 5E). To activate pDC in a non-viral way, we stimulated them with the TLR-9 agonist CpG, which was reported to induce IFN-α production and TRAIL expression by pDC [48]. DENV-2 detection on monocytes was significantly diminished when cells were incubated with CpG-stimulated pDCs including supernatants (figure 5F). Importantly, non-pre-activated pDCs also diminished DENV-2 infection in monocytes, although to lesser extent. Chloroquine, which is an inhibitor of TLR-9 pathway, blocked pDC activation and partially restored DENV+ cell detection within monocytes. Indeed, IFN-α was highly detected in the cocultures of infected monocytes with CpG-activated pDCs compared to non-pre-activated pDCs (figure 5G). Chloroquine completely blocked IFN-α secretion, suggesting the more IFN-α produced by activated pDCs the less viral antigens are detected. These findings are strongly supportive for an important role for pDC on DENV replication in monocytes.
Finally, we also tested whether IKpDC-mediated apoptosis was involved in reduced DENV-2 detection in monocyte-pDC coculture. Monocytes were infected with DENV-2 or mock and then co-cultured with or without CpG-stimulated with or without pDCs (figure S3). After 48 hours of infection, cultures were collected and stained for AnnexinV and TOPRO3. We observed that the addition of unstimulated pDC to DENV-infected monocytes had no major impact on cell death during co-culture, remarkably, when compared to DENV only. Furthermore, CpG-activated pDCs addition caused an increase survival of monocyte during co-cultures, meanwhile reducing viral antigens. Therefore, we could rule out the killing effect of IKpDCs and once more attribute an antiviral role for IFN-α (and/or TRAIL) in the supernatant.
The present work describes features of pDC activation during DENV-2 infection and discusses its importance for disease outcome. We characterized, for the first time, an activated profile of pDCs from dengue patients using membrane TRAIL expression as a marker. Moreover, we observed that, in vitro, activated pDCs exerted an antiviral activity in infected human primary monocytes. Thus, pDCs may contribute to the control of viral clearance and to diminish the severity of the disease.
Upon challenge by viral particles, pDC activation takes place, characterized by upregulation of co-stimulatory markers, and by very high levels of IFN-α secretion [52]. Simultaneously to IFN production, we previously demonstrated that viral-activated pDC also expressed the pro-apoptotic ligand TRAIL on their membrane, which transforms them into IFN-producing Killer pDC (IKpDCs) [48], [53], [54]. For instance, in HIV-1 infection, the number of IKpDCs was correlated to CD4 depletion and disease progression [53]. However, the pDC function depends on the etiology of viral infection. During dengue disease, pDC activation by membrane TRAIL expression was found associated with less severe clinical manifestations. Other studies have also assessed blood pDCs from DENV-infected patients. Reduced absolute numbers of these cells were associated with a poor outcome, because severe cases of dengue disease exhibit a lower number of blood pDCs [32] and low levels of blood pDCs were correlated with high viral loads [33]. Nevertheless, treatment with TLR-3 and -7/8 agonists enhanced pDC activation and reduced viral replication in non-human primate model during DENV infection [55]. Supposedly, a blunted pDC response would allow viral replication to take place. Therefore, we also decided to characterize plasma levels of pDC-related cytokines.
Viral activation of PDC leads to production of IFN-α. Although pDC does not produce sTRAIL, pDC-produced IFN-α leads to production of soluble or membrane bound TRAIL by several cell types including monocytes [56]. Because IFN-α and TRAIL were reported to be antiviral in vitro for DENV, we analyzed the soluble levels in dengue patients. Indeed, plasma levels of both factors were statistically correlated with pDC activation in our cohort. Regarding blood cytokine levels in dengue patients, we find discrepancies in literature [9]. Inflammatory cytokines are increased in severe cases compared to mild forms. Even though, IFN-α was reported in DF and DF severe cases [57], Chen et al. detected higher IFN-α levels in DF compared to severe cases [34], supporting our data. Soluble TRAIL levels were not associated to severe forms but to febrile period and to primary infections [41]; however, we found a negative association between soluble TRAIL and severity. Because TRAIL is a downstream IFN stimulated gene, reduced IFN-α levels could explain low levels of soluble TRAIL in severe patients. Moreover, a weak type I IFN response in severe DF patients could represent a viral escape pathway. Others have reported that some viruses can evade TLR-induced IFN-α production, by inhibiting pDC function through the binding to BDCA-2, a cell surface molecule that functions as IFN secretion inhibitor [58], [59]. Indeed, BDCA-2 attachment can also abolish TRAIL-mediated cytotoxicity of pDCs [60]. It remains to be investigated whether DENV proteins can downregulate pDC function. This could explain why some patients respond efficiently to DENV infection and show high levels of IFN-α and sTRAIL, while others do not produce sufficient levels of the factors (severe cases). Furthermore, elevated numbers of activated pDCs could, by releasing high levels of IFN-α, protect target cells and activate other innate immunity actors, like Natural Killer cells that are associated with mild DF [61]. Therefore we suggest a protective role of activated pDCs during acute phase of dengue virus infection.
We asked whether pDCs could acquire IKpDC phenotype and have a protective role against DENV infection in vitro. DENV-2 induced IFN-α production and TRAIL relocalization from the intracellular compartment (in resting pDC) to pDC membrane (activated pDC) soon after viral exposure, supporting the idea of a rapid response to viruses. A high viral load was necessary to activate pDCs that was only achieved after a concentration procedure using ultracentrifugation protocols [47], [48]. Although purification and ultra-centrifugation protocols may decrease infectious-to-particle ratio [62] our concentrated inoculum displayed improved infectious features. However, we cannot rule out that both non-infectious and infectious particles are activating pDCs in synergism, as it was shown for HIV. Indeed, infectious and AT-2-treated HIV (non-infectious) were both able to activate TLR-7 pathway in pDCs [63]. Apparently, it seems that pDCs need large quantities of virus to be activated or high frequency of viral receptor [64]. HTLV-1 also required high viral loads to activate IKpDCs [48]. Indeed, some discrepancies of IFN-α production by DENV-stimulated pDC are reported that may be result from using low viral loads [65] [66]. Flaviviruses may have acquired intrinsic mechanisms to avoid pattern recognition receptors [67] and consequently pDC activation.
To better elucidate pDC activation by DENV, we studied endocytosis pathway in pDC. Lysosomal acidification was crucial for TRAIL expression and IFN-α production by DENV-2 activated pDC. Another report showed that TLR-7 was the endosomal recognition receptor for DENV-2 by using specific inhibitors and acidification blockers [31] and that endocytosis pathway was crucial for co-stimulatory markers upregulation and IFN-α production [30]. Indeed, DENV-2 particles are detectable in pDC in the early stage (2 h) before viral degradation in lysosomes. However, after 18 h we did not detect viral antigens suggesting an absence of viral replication into pDC. Furthermore, lysosomal acidification impairment allowed detection of DENV-2 in pDC, contrasting with non-treated pDCs, suggesting that viruses are not disassembling. We did not observe an increase of non-structural protein 1 in culture supernatants (data not shown) after viral adsorption, supporting the incapacity for virus replication in pDCs. Our data is in accordance with others, as low levels of replicative negative strand RNA were found inside pDCs [30]. Therefore, we suggest that DENV-2 particle sensing occurs in endosomal compartments. Recognition but not infection of DENV-2 is responsible for IKpDC activation, whereas it leads to TRAIL relocalization and IFN-α production.
We next wonder whether IKpDC and IFN-α could inhibit DENV-2 replication in human monocytes, one of main target cells for DENV. Type I interferon have a crucial role during innate immune responses inhibiting viral replication and spreading of many viruses [68]. Binding and activation of IFN receptors triggers transcription of interferon stimulated genes, which induce products that are able to inhibit several steps of virus replication [69]. We found that DENV-2 infection was strongly diminished by treatment with IFN-α in human monocytes. In accordance with these data, other reports demonstrated that pre-treatment of several susceptible cell lines with type I interferon blocked DENV-2 replication through a protein kinase R (PKR)-dependent mechanism [70], [71], [72]. Indeed, recently, several interferon-stimulated genes such as interferon-inducible trans-membrane (IFITM) proteins were able to inhibit dengue infection in cell lines [73], [74]. However, type I interferon pathway is also subject to interference by many viruses that directly target pathways required for type I interferon response. Monocytes and monocyte-derived dendritic cells can produce IFN-α once they are infected by DENV, however at much lower levels compared to other viruses [66], [75], [76]. Moreover, several reports show degradation of downstream [77], [78], [79] and upstream [65], [80] interferon signaling pathways by DENV non-structural proteins. Although DENV blocks type I IFN pathway, the cytokine still remains protective for other uninfected cells reducing viral spreading during infection as described by others [81].
In our study, infected monocytes co-cultured with IKpDCs displayed a dramatic reduction in viral load that could be partially reversed by lysosomal blockage. Viral detection was negatively related to IFN-α detection in cocultures of monocytes and pDCs. IKpDC activation may play an important role for a rapid viral clearance. TRAIL has been reported as a potential antiviral factor for DENV replication [40]. Because TRAIL expression and production by monocytes is induced by IFN-α [37], we tested several concentrations of recombinant TRAIL on monocytes, and we confirmed the antiviral function as published before [40]. However, membrane TRAIL blockage on IKpDC had minimal effect on viral load or apoptosis during cocultures (data not shown). Moreover, IKpDC had no significant effect on DENV-2-infected monocyte apoptosis, suggesting that the anti-viral effect of pDC is mainly due to IFN-α and/or TRAIL on viral replication and not to cell death. Although, both TRAIL and IFN-α were fundamental in reducing viral load in HIV-infected CD4+ T cell/pDC co-cultures [82], [83]. For DENV, type I interferon was sufficient to largely reduce viral infection rates. Therefore, because we could not demonstrate that IKpDCs have a role in killing infected monocytes, this population may modify the outcome of the disease by producing massive quantities of IFN-α that would in turn block dengue replication in monocytes before adaptive immune responses ensues.
Finally, we showed in this work that DF patients harbored higher frequencies of circulating activated pDC and higher IFN-α/TRAIL levels compare to severe cases. DENV is activating pDC response in terms of IFN-α production and membrane TRAIL expression. We demonstrated that DENV mainly activates the endocytosis pathway and not the infection pathway, as we did not detect viral infection in pDC. Furthermore, our in vitro co-cultures data strongly support a crucial antiviral role for activated pDC and IFN-α by dramatically reducing viral spread. Even though, studies on DENV evasion from pDC response are still needed, we believe that pDC activation in patients' blood may contribute in the future to the establishment of good prognostic immune response together with clinical manifestations/warning signs.
|
10.1371/journal.pgen.1006993 | Fibroblast growth factor signaling is required for early somatic gonad development in zebrafish | The vertebrate ovary and testis develop from a sexually indifferent gonad. During early development of the organism, primordial germ cells (the gamete lineage) and somatic gonad cells coalesce and begin to undergo growth and morphogenesis to form this bipotential gonad. Although this aspect of development is requisite for a fertile adult, little is known about the genetic regulation of early gonadogenesis in any vertebrate. Here, we provide evidence that fibroblast growth factor (Fgf) signaling is required for the early growth phase of a vertebrate bipotential gonad. Based on mutational analysis in zebrafish, we show that the Fgf ligand 24 (Fgf24) is required for proliferation, differentiation, and morphogenesis of the early somatic gonad, and as a result, most fgf24 mutants are sterile as adults. Additionally, we describe the ultrastructural elements of the early zebrafish gonad and show that distinct somatic cell populations can be identified soon after the gonad forms. Specifically, we show that fgf24 is expressed in an epithelial population of early somatic gonad cells that surrounds an inner population of mesenchymal somatic gonad cells that are in direct contact with the germ cells, and that fgf24 is required for stratification of the somatic tissue. Furthermore, based on gene expression analysis, we find that differentiation of the inner mesenchymal somatic gonad cells into functional cell types in the larval and early juvenile-stage gonad is dependent on Fgf24 signaling. Finally, we argue that the role of Fgf24 in zebrafish is functionally analogous to the role of tetrapod FGF9 in early gonad development.
| The genes involved in the early stages of vertebrate gonad development remain largely undefined. The gonad begins to form when primordial germ cells and somatic gonad precursor cells coalesce during early development. However, we know little about the signaling events that lead to the subsequent morphogenesis of the early developing gonadal primordium. Using the zebrafish, a model vertebrate, we show that the early somatic gonad organizes into a bi-layered structure, with an outer epithelial layer surrounding an inner mesenchymal core. We demonstrate that the gene encoding the Fibroblast growth factor 24 ligand, fgf24, is expressed by the epithelial population. Utilizing a null mutation in fgf24, we show that Fgf24 signaling is required for the expression of several genes by the inner mesenchymal cells, including those encoding transcription factors, a hormone biosynthesis enzyme, and a TGF-β ligand. Furthermore, the somatic cells in fgf24 mutant gonads have reduced proliferation rates, do not organize into the bi-layered structure seen in wild-type gonads, and most fgf24 mutants are infertile as adults due to the inability of the somatic gonad to support germ cell development. Finally, we argue that the function of Fgf24 during development of the early teleost gonad is analogous to the proposed role of FGF9 during development of the early tetrapod gonad.
| The vertebrate gonad consists of germ cells, the lineage directly responsible for creating the next generation, and somatic gonad cells (SGCs). The somatic gonad serves two important functions. First, it creates an environment that protects germ cells and nurtures their development. For example, with the exception of mammalian females, most animals examined retain the ability to produce gametes throughout their adult life, an activity permitted by the presence of germline stem cells [1]. SGCs form the niche that is required to maintain these germline stem cells. In the adult mouse testis, several SGC types (Sertoli, Leydig, and peritubular myoid cells) secrete growth factors that promote proliferation and suppress differentiation of the germline stem cell population, thereby maintaining fertility through adulthood [2, 3]. Second, a subset of SGCs secretes hormones required for the development of secondary sexual characteristics, so defects in gonad development can result in disorders of sexual development (reviewed in [4]). Although the importance of the somatic gonad to fertility is clear, much remains to be learned about the genes that regulate its early development.
In mammals, the somatic gonad is derived from intermediate and lateral plate mesoderm, and its early development has been divided into three major steps: initiation, growth, and sexual differentiation. During the initiation phase, a portion of coelomic epithelium on the ventral surface of the mesonephros begins to thicken to form the bilateral genital ridges [5–8]. Soon after their formation, primordial germ cells (PGCs) migrate into the ridges [9]. The growth phase is characterized as a period of somatic and germ cell proliferation that results in a larger, multilayered primordium. Finally, sexual differentiation transforms the bipotential gonads into either an ovary or testis. Although the molecular mechanisms that regulate sexual differentiation of the gonad are relatively well understood (reviewed in [10]), less is known about the genetic regulation of the initiation and growth phases. In mice, several transcription factors have been identified that are required for initiation and growth of the primordium: mutations in Gata4, Wilms tumor 1(Wt1), Steroidogenic factor 1 (Sf1/Nr5a1), Lim homeobox protein 9 (Lhx9) or the paired-like homeobox gene Emx2 cause failure to initiate gonad development or undergo early gonad regression [11–17]. However, it is not known if cell signaling is important for these early events.
The Fgf signaling pathway regulates many developmental events in metazoans. The pathway generally consists of secreted ligands that complex with heparan sulfate proteoglycans in the extracellular matrix to bind and activate transmembrane Fgf receptors [18, 19]. One such ligand, Fgf24, is known to play roles in the development of zebrafish posterior mesoderm, forelimb, and pancreas [20–22]. Fgf24 is a member of the FgfD subfamily of Fgf ligands, which, in zebrafish, consists of six members: fgf8a, fgf8b, fgf17, fgf18a, fgf18b, and fgf24 [20, 23, 24]. Although this subfamily is conserved in mammals, the fgf8 and fgf18 duplications are a result of the teleost-specific whole-genome duplication [24]. Furthermore, while Fgf24 is present in basal tetrapods and teleosts (coelacanths and spotted gar, respectively), it was lost early in the tetrapod lineage, so is not present in the mammalian genome [23–27].
In this study, we show that the majority of fgf24 null mutants (ikahx118, hereafter referred to as fgf24hx118; [21]) are sterile as adults, suggesting that it plays a role in either gonad development or maintenance. We show that fgf24 is first expressed in a subset of SGCs by 8 days post fertilization (dpf)—a period that we argue is analogous to the early growth stage of mammalian gonad development—and that this expression is required for early somatic gonad proliferation and morphogenesis into a bi-layered tissue that in wild-type normally occurs by 10 dpf. Coincident with bi-layer formation, we show that cells expressing fgf24 are restricted to the epithelial layer on the surface of the gonad. Furthermore, we show evidence that the cells responding to Fgf24 signaling are a mesenchymal population of SGCs that localizes to the interior of the early gonad, and that loss of Fgf24 function leads to a failure of these cells to differentiate into functional cell types. Finally, we argue that the loss of germ cells in fgf24 mutants is an indirect consequence of defective somatic gonad development.
Gonads are composed of both germ cells and somatic cells that enclose the germ cells and regulate their development. In vertebrates, the somatic gonad is also the main source of sex hormones that regulate secondary sexual characteristics, such as sexually dimorphic appearances and behavior. Zebrafish primordial germ cells (PGCs) are specified during the early cleavage stage by maternal factors, and shortly thereafter migrate to where the somatic gonad will eventually form [28] (Fig 1). Though it is not known with certainty when formation of the somatic gonad initiates, histological analysis has detected SGCs surrounding germ cells as early as 5 dpf [29].
Domesticated zebrafish lack sex chromosomes and it is still unclear how sex determination is regulated or precisely when it occurs [30]. Zebrafish embryos hatch from the chorion around 3 dpf and are free-swimming larvae by 5 dpf. The transition from the larval to juvenile stage occurs around 25 dpf, which coincides closely to when overt sex-specific differentiation becomes apparent (Fig 1). During the early larval stage (i.e. 3–20 dpf), there are no overt differences between animals that will become male or females, and all animals initially produce varying numbers of early stage oocytes (beginning ~13 dpf; [31]). Mutations that reduce or eliminate germ cells, or more specifically, the ability to produce early stage oocytes during the larval period, result in all male development [32–34]. During this period of development, the somatic gonad is bipotential as evident by the simultaneous expression of genes that will eventually be sex-specific. For example, a subset of somatic cells in the larval gonad expresses the female-specific aromatase-encoding gene, cyp19a1a, while neighboring cells express the male-specific amh gene ([35]; this report). These studies have led to the hypothesis that oocytes produce a signal that stabilizes female-specific gene expression in the somatic gonad, thereby promoting female development.
It has previously been shown that zebrafish fgf24 homozygous mutants, called fgf24hx118, are viable but lack pectoral fins [21]. In addition to this defect, we discovered that all fgf24hx118 mutants were males as adults (Fig 2A–2C’). Previous studies have established that zebrafish lacking germ cells, or the ability to produce early-stage oocytes, invariably develop as phenotypic males [32–34]. We therefore examined 27 fgf24 mutants at 3.5 months post fertilization (mpf) and found that 21 animals contained no detectable gonads and were thus sterile (~78%). Each of the remaining 6 animals had one small testis measuring approximately 1/3 the size of a wild-type testis (~22%; S1A and S1B Fig). High-resolution confocal analysis of these latter testes revealed an overall wild-type organization where germ and somatic cells were properly arranged into tubule structures that contained germ cells in all stages of spermatogenesis, including mature sperm (S1C and S1D Fig). Furthermore, in situ hybridization with the Sertoli cell gene markers anti-müllerian hormone (amh) and gonadal soma derived factor (gsdf) showed a similar pattern of expression in both fgf24 mutants and wild-type controls (S1E–S1H Fig). Finally, while mutant fish that have gonads were unable to induce wild-type females to spawn due to their lack of pectoral fins [37], sperm extracted from these gonads and used for in vitro fertilization of wild-type eggs produced viable embryos at frequencies indistinguishable from wild-type sperm (S1 Table).
Although the fgf24hx118 allele is an N-ethyl-N-nitrosourea (ENU)-induced point mutation that introduces a premature stop codon within exon 4 that should truncate the protein within the core Fgf homology domain [21], the incomplete penetrance of its adult sterility phenotype prompted us to investigate whether this allele was hypomorphic (S1I Fig). To test this, we used the CRISPR/Cas9 genome editing technology to induce a new mutation within exon 3 and identified a 5 bp insertion allele, called fgf24uc47, that results in a translational frameshift, and is therefore expected to be a null mutation (S1I Fig). We found that both fgf24uc47 homozygous mutants and fgf24hx118/uc47 transheterozygotes have a phenotype that is indistinguishable from that caused by the fgf24hx118 mutation alone: the resulting animals are homozygous viable, lack pectoral fins, and develop as male with a partially penetrant sterility defect. In subsequent experiments, we use these two alleles interchangeably and conclude that null mutations in fgf24 lead to an incompletely penetrant adult sterility defect.
fgf24 mutants are all male as adults, a phenotype that suggest that mutants may have defects in early gonad development (see Background, above). We therefore first determined if they had defects in early germ cell development by comparing the number of germ cells present in the gonads of wild-type and mutant fish at several stages throughout larval development. For these experiments, we collected confocal images at 5 μm intervals through the whole gonad and quantified distinct germ cells identified by anti-Vasa antibody and DAPI DNA staining. We found that fgf24 mutants had similar numbers of germ cells to wild-type siblings at 8 and 10 dpf. However, at 12 and 14 dpf, mutant animals had significantly fewer germ cells than their wild-type siblings (Fig 2D–2F).
In addition to our finding that fgf24 mutants have fewer germ cells than wild-type, they also appear to have fewer SGCs (arrowheads in Fig 2E and 2F). This feature was most apparent when we compared testes isolated from wild-type and fgf24 mutant animals. A 40 dpf wild-type testis is organized into tubules that contain many premeiotic and spermatogenic germ cells surrounded by SGCs (Fig 2G and 2G’). One such SGC is the Sertoli cell, which expresses the teleost gonad-specific Tgf-β ligand, Gsdf [38]. At 40 dpf, Gsdf-expressing Sertoli cells are abundant in wild-type testes and can be visualized using the Tg(gsdf:mCherry)uc46 transgene as previously reported [38] (n = 7; Fig 2G and 2G’). In stark contrast, the gonads of all fgf24 mutants contain few germ cells that are not organized into tubule-like structures and lack detectable Tg(gsdf:mCherry)uc46 expressing SGCs (n = 7; Fig 2H and 2H’). Notably, gsdf is also expressed in ovarian granulosa cells (S2 Fig). Thus the inability of fgf24 mutant gonads to express Tg(gsdf:mCherry)uc46 argues that fgf24 is required for the development of both male and female SGCs. Finally, the testes and ovaries of wild-type juvenile animals (25-90dpf) contain both premeiotic and postmeiotic germ cells at different stages of gametogenesis (Fig 2G’; S2 Fig) in contrast to juvenile fgf24 mutant gonad, which contain only premeiotic germ cells as evident by their large size and prominent nucleoli (Fig 2H’, inset). In conclusion, loss of fgf24 function affects the development of both testes and ovaries, which, together with the expression of fgf24 during the early bipotential phase, suggests that it is required for development of the early bipotential gonad (Fig 1).
To further test the role of Fgf24 in early bipotential gonad development, we used high resolution fluorescent RNA in situ hybridization to determine which gonadal cells express fgf24. The larval gonad is a rod-like structure that is oriented along the anterior/posterior axis. Germ cells are restricted to the interior core of the gonad and are surrounded by SGCs (Fig 2E). Because the germ cell phenotype of fgf24 mutants is first evident by 12 dpf, we examined the expression of fgf24 mRNA in gonads at various stages between 5 and 20 dpf in whole-mount gonads. In addition, we co-stained gonads for the germ cell-specific Vasa protein to aid in gonad identification during dissection. At 5 dpf, when germ cells and SGCs have just begun to coalesce [29], fgf24 was not detected (n = 4; Fig 3A). However, by 8 dpf, we could detect fgf24 expression in some, but not all, SGCs in 23/27 wild-type animals (Fig 3B). In 10 and 16 dpf animals, fgf24 could be detected in all gonads examined (n = 19 and 12, respectively). As in 8 dpf gonads, fgf24 was detected exclusively in SGCs, but only in a subset of SGCs that appeared to be restricted to the surface of the gonad (Fig 3C and 3D). Finally, in 20 dpf animals, we continued to detect fgf24 only in SGCs, though expression appeared highest in a population of cells that localize to the dorsal edge of the gonad (n = 9; Fig 3E). We conclude that fgf24 is expressed in gonads during the time when the development of fgf24 mutant gonads begins to deviate from those of wild-type animals (Fig 3A’, 3B’, 3C’, 3D’ and 3E’). Furthermore, these results reveal that there are at least two distinct SGC populations in the larval gonad soon after its formation: fgf24(+) and fgf24(-).
The apparent defect in both the somatic and germ cell components of the gonad led us to investigate which cell type(s) responds to Fgf24 signaling. Fgf signaling can activate downstream signaling cascades that can result in gene expression changes. Transcription of the Ets variant (Etv) family of transcription factors is known to be upregulated by Fgf signaling in many developmental contexts (e.g. [39–41]). We therefore asked whether one of these family members, etv4/pea3, is expressed in the wild-type larval gonad. At 8 dpf, when we first detect fgf24 expression in ~85% of wild-type animals (Fig 3B), we can detect expression of etv4 in SGCs in 56% of wild-type gonads (13/23; Fig 4A and 4B). Notably, it is only detected in gonads that also express fgf24 (Fig 4A) and is not detected in 8 dpf fgf24 mutant gonads (n = 14; Fig 4C). By 10 dpf, however, we found that etv4 is strongly expressed in SGCs in all wild-type gonads examined (n = 25). Interestingly, double in situ hybridization of both etv4 and fgf24 reveals that etv4 is expressed in a population of SGCs distinct from that of fgf24 and one that subtends the fgf24-expressing layer (n = 10; Fig 4D and 4D’). In contrast, the gonads of 10 dpf fgf24 mutant siblings express no, or greatly reduced levels of, etv4 (n = 8; Fig 4E). These results suggest that Fgf24 acts as a paracrine signal to regulate the development of an inner population of SGCs.
In various contexts, the Fgf activation of etv4 transcription is mediated by Erk, a terminal kinase of the Map kinase signaling pathway [42, 43]. Erk phosphorylation by the upstream kinase, Mek, allows it to translocate to the nucleus and activate numerous transcription factors (reviewed in [44]). We therefore asked whether Erk is phosphorylated in larval SGCs. Indeed, we detected substantial phosphorylated Erk (pErk) in wild-type SGCs, but not in SGCs of fgf24 mutants (n = 13 and 11, respectively; Fig 4F–4G’). In contrast to etv4 expression, pErk does not appear to be restricted to the inner layer of SGCs, indicating that fgf24-dependent Map kinase activity is present in apparently all SGC populations. These data therefore suggest that Fgf24 may activate etv4 expression via the MAPK pathway, but that other factors must act to limit etv4 expression to only the inner SGC.
Our analysis thus far supports a model where the primary role of Fgf24 is to promote the development of the somatic gonad and that the loss of germ cells in fgf24 mutants is a secondary consequence of defective somatic gonad development. To further this analysis, we analyzed the expression of two main classes of genes: 1) genes reportedly required for early gonad development in the mouse, and 2) genes known to be important for later gonad development and function in both mammals and fish. The transcription factors Gata4, Nuclear receptor subfamily 5 group a1 (Nr5a1, also called Steroidogenic factor 1, Sf1), and Wilms tumor protein 1 (Wt1) regulate early mouse gonad development. Gata4 is required for epithelial proliferation during the initiation phase, while the latter two promote cell survival during the growth phase [11–14]. Using fluorescent RNA in situ hybridization we found that zebrafish gata4 and nr5a1a orthologs are readily detectable in SGCs of wild-type 10 dpf gonads (n = 14 and 15, respectively), but absent or reduced in fgf24 mutant gonads (n = 13 and 12, respectively; Fig 5A–5D). In contrast, we found that wt1a, one of two zebrafish Wt1 orthologs, is expressed in the SGCs of both wild-type and fgf24 mutant gonads at 11 dpf (n = 19 and 12, respectively; Fig 5E and 5F). Notably, wt1a appears to be expressed at lower levels in inner SGCs (arrows, Fig 5E) and most robustly in the outer layer of SGCs on the dorsal edge of the gonad (arrowheads, Fig 5E).
We next assessed the expression of genes associated with differentiated cell types of the larval gonad: cyp19a1a, which encodes an aromatase normally expressed in granulosa and theca cells of the adult ovary, and the Anti Müllarian Hormone-encoding gene amh, which is normally expressed in Sertoli cells of the adult testis [35, 36]. Because the larval gonad is initially bipotential, some genes that are later expressed sex-specifically, including cyp19a1a and amh, can be detected in the gonads of all animals during the early larval stages [35]. Indeed, we found that the expression of both genes can be detected in most wild-type gonads starting at 11 dpf (15/16, 16/16 respectively; Fig 5G and 5I). By contrast, the expression of these genes was not detected, or was detected at very low levels in gonads of 11 dpf fgf24 mutants (2/12 and 1/11 showed low expression, respectively; Fig 5H and 5J). Because we found that two populations of somatic cells can be distinguished in the early larval gonad based on fgf24 and etv4 expression (Fig 4D), we performed high resolution fluorescent RNA in situ hybridization to determine in what cell layer cyp19a1a and amh are expressed. Similar to etv4, we found that both genes are expressed in only interiorly-localized cells. However, we did not detect co-expression of cyp19a1a, amh, or etv4 in the same cells, indicating that the inner SGCs of 12 dpf gonads are comprised of at least three distinct cell populations (Fig 5K–5M). Together, these results further support the model that the primary function of Fgf24 is to promote development of SGCs.
In addition to the defects in somatic gonad gene expression, mutants older than 12 dpf have gonads that are smaller than their wild-type siblings because they have fewer germ cells and apparently fewer SGCs (Fig 2D–2F). Decreased cell numbers could be due to decreased cell proliferation, increased cell apoptosis, or a combination of the two. We first asked if mutant gonads have an increase in cell apoptosis relative to wild-type. We assessed the extent of apoptosis by staining for Cleaved caspase 3 (Cc3) and by performing a TUNEL assay. At 10 and 14 dpf, neither wild-type nor fgf24 mutant gonads displayed appreciable apoptosis in either SGCs or germ cells (Fig 6A–6C’; S3A–S3B’ Fig). In addition, we asked if we could rescue the fgf24 phenotype by blocking Tp53-mediated apoptosis. Using the tp53 M214K allele [45] we produced tp53;fgf24 double mutants, which phenocopied fgf24 single mutants: 100% of double mutants were phenotypically male as adults and 69.2% lacked gonads completely (n = 13). In comparison, only 28.6% of tp53 single mutants were male and 100% had two fully developed gonads (n = 14, S2 Table). Together, these data suggest that the decreased number of cells in fgf24 mutant gonads is not a result of increased apoptosis.
Finally, we asked if decreased cell numbers in fgf24 mutants was the result of reduced proliferation rates. We therefore exposed wild-type and fgf24 mutant animals to the thymidine analog 5-ethynyl-2’-deoxyuridine (EdU) from 8 to 9 dpf to label cells in S-phase of the cell cycle. Although we detected EdU in SGCs of both genotypes, the percentage of SGCs that were EdU-positive was significantly higher in wild-type compared to mutant gonads (70% and 41%, respectively; P < .001; Fig 6D–6F). In contrast to SGCs, we never observed EdU incorporation in germ cells during this or later time frames (Fig 6D–6E’ and S4A–S4D’ Fig). Because there is a vast increase in germ cells during larval development (Fig 2D), we concluded that germ cells may incorporate this thymidine analog less efficiently, and therefore utilized an antibody against the mitosis-specific phospho-Histone H3 (pHH3) to identify germ cells in prophase, when pHH3 is detected throughout the nucleus [46]. While the percentages of pHH3-positive germ cells are similar between genotypes at 8 dpf, by 10 dpf wild-type gonads have a significantly higher proportion of pHH3-positive germ cells than mutant gonads (58% and 34%, respectively; P < .01; Fig 6G–6I). These results argue that decreased SGC and germ cell numbers in fgf24 mutant gonads are due primarily to a decrease in cell proliferation.
The data above indicate that the early zebrafish somatic gonad is composed of two somatic layers. It is therefore possible that the outer fgf24-expressing layer is a developing epithelium. To test this hypothesis, as well as to compare the overall structure of wild-type and fgf24 mutant gonads at high resolution, we analyzed transverse sections by transmission electron microscopy (TEM). At 10 dpf, we found that both wild-type and mutant gonads were arranged with germ cells in the center, surrounded by SGCs (Fig 7A and 7B). However, the thickness of the SGC portion surrounding germ cells appeared to be greater in wild-type compared to mutant gonads. Furthermore, the SGCs in wild-type gonads were divided into two layers, likely corresponding to the fgf24-expressing cells and the etv4-expressing cells (Fig 7A). We noted that these layers were separated by an electron-lucent space, perhaps due to the presence of a basement membrane (n = 6; Fig 7A’). We therefore asked whether Laminin, a central component of the basal lamina, could be detected in this region. While Laminin is not detected in wild-type 8 dpf gonads (n = 8; S5A and S5A’ Fig), there is abundant Laminin deposited between the two layers of SGCs of wild-type 10 dpf gonads, indicating the presence of a basement membrane (n = 14; Fig 7C and 7C’). In contrast to wild-type, fgf24 mutant gonads have only one layer of SGCs (n = 5; Fig 7B’) and lack Laminin staining altogether (n = 10; Fig 7D and 7D’). These results suggest that Fgf24 is required for normal morphogenesis of the early larval gonad.
Frequently, Fgf ligands are secreted by one cell layer (e.g. epithelial) and signal across a basement membrane to Fgf receptor-expressing cells of a second layer (e.g. mesenchymal; [47], reviewed in [48]). Analysis of our TEM data revealed that the outer layer of SGCs in wild-type gonads makes many cell-cell contacts, seen as electron dense patches, characteristic of epithelial cells (Fig 7A”). We therefore asked whether the fgf24- and etv4-expressing cells were adopting epithelial and mesenchymal fates, respectively. To address this question, we determined the cell junction landscape of each cell layer by staining for components of adherens and tight junctions.
Adherens junctions are mediated by transmembrane cadherins, of which there are several types. Cadherin homodimerization helps similar cells associate with each other, and can promote cell sorting within a tissue. Inside the cell, catenins link the intracellular tail of the cadherin to actin, providing mechanical linkage between adjacent cells. In wild-type gonads, we found that β-catenin is expressed in virtually all cells, suggesting that all SGCs and germ cells have some type of adherens junctions (n = 7; Fig 8A and 8A’). Interestingly, we found that Cdh2/N-cadherin is highly localized to the membranes of the outer SGCs and weakly to the membranes of inner SGCs and germ cells (n = 17; arrowheads and arrows, respectively, Fig 8C and 8C’). In contrast, we see very low levels of Cdh1 (E-cadherin) in both germ cells and SGCs (n = 15; S6A–S6C’ Fig).
Tight junctions are a hallmark of epithelial cells, where they function to block the passage of fluids and molecules between cells. Tight junctions are formed by interactions between the transmembrane occludins and claudins and the intracellular Tjp1 (Tight junction protein 1/Zo-1), the latter of which interacts with actin. We found that Tjp1 was expressed in SGCs only and localized most intensely to the outer SGC membranes, similar to Cdh2 (n = 15; arrowheads, Fig 8E and 8E’). These data argue that the outer layer of fgf24-expressing SGCs forms an epithelium.
Our TEM data reveal that both wild-type and mutant SGCs make many electron-dense cell-cell contacts (arrows, Fig 7A” and 7B”). We therefore hypothesized that mutant gonads would also maintain the expression and localization of cell adhesion molecules. Indeed, we found that Cdh2 and Tjp1 remain strongly localized to some SGC membranes (n = 15 and 14, respectively; Fig 8D, 8D’ and 8F, 8F’), while β-catenin and Cdh1 appear slightly reduced (n = 7 and 10, respectively; Fig 8B and 8B’ and S6D and S6D’ Fig).
Finally, we sought to determine the identity of the SGCs that are present in the early fgf24 mutant gonads. In wild-type animals, the outer, fgf24-expressing layer of SGCs showed strong localization of Cdh2 and Tjp1, the two cell adhesion molecules that were maintained in fgf24 mutant somatic gonad cells. We therefore hypothesized that the SGCs that remain in the mutants are most similar to the fgf24-expressing epithelial cells of the wild-type gonad. To test this, we asked whether we could detect fgf24 transcript in fgf24 mutants, as it is known that nonsense mediated decay of transcripts with premature stop codons varies in efficiency [49, 50]. Using fluorescent in situ hybridization, we found that gonads of 11 dpf fgf24 mutants have both SGCs with and without detectable fgf24 transcripts (arrowheads and asterisks in S7A–S7B’ Fig, respectively). Thus, it appears that the fgf24 mutant gonad, like the wild-type gonad, contains two distinct populations of SGCs, but that they fail to form the bi-layered organization observed in wild-type gonads.
Prior to this study, nothing was known about the genetic regulation of early somatic gonad development in zebrafish. Here we have presented evidence that early somatic gonad development in zebrafish is regulated, in part, by Fgf signaling mediated by the Fgf24 ligand. Our results show that fgf24 is expressed in an outer layer of epithelial SGCs, and that these cells surround and signal to a distinct population of internally-located mesenchymal SGCs that are in close contact with the germ cells. In the absence of Fgf24 signaling, the inner SGCs fail to express genes known to be important for gonad development and/or maintenance. Additionally, our results suggest that Fgf24 signaling is required for the formation of this bi-layered early gonad. We find that while these early defects are fully penetrant in larval fish, as adults approximately 22% of animals eventually develop partial testes. Although the phenotypes we describe here are novel for mutants that affect the Fgf signaling pathway in any vertebrate, we argue that the requirement for Fgf signaling during the development of the early bipotential gonad may be widely conserved.
We initiated these studies because of the discovery that all fgf24 mutants are male as adults. This phenotype could suggest that the primary role of Fgf24 is to promote female sex determination or differentiation. However, our results strongly argue that the primary role of Fgf24 is instead to promote the development of the early bipotential gonad, the precursor to both ovaries and testes, and that the effect on sex determination is a secondary consequence of this primary defect. It is well established that germ cells, and in particular oocytes, are required for female sex determination and/or differentiation, as mutations that reduce or eliminate early germ cell development, or specifically early-stage oocytes during the bipotential phase, result in an all-male phenotype [32–34, 51, 52]. We have established here that all fgf24 mutants have significantly reduced germ cells numbers relative to wild-type as early as 12 dpf (early bipotential stage), which can thus explain why all fgf24 mutants develop as males. Importantly, we have presented evidence that gene expression and cell proliferation defects in the somatic gonad can be detected in fgf24 mutants as early as 8 dpf, several days prior to when we can detect significant differences in germ cell numbers between wild-type and mutant larvae. Finally, the expression of both male- and female-specific genes is equally affected by loss of fgf24 function (e.g. amh and cyp19a1a; Fig 9), a result that is inconsistent with Fgf24 having a sex-specific role. Thus, we strongly favor a model where the primary role of Fgf24 is in promoting somatic gonad development during the bipotential phase, and that defects in female development are a secondary consequence to earlier defects in the development of the bipotential gonad.
Our current data suggest that the gonads of fgf24 mutants have fewer SGCs in comparison to their wild-type siblings. This phenotype could result from a failure in the specification and/or migration of somatic gonad precursors cells, a failure of these cells to proliferate, or a combination of these factors. In mammals, SGCs are derived from cells of the coelomic epithelium and, in males, there is also contribution from the mesonephric mesenchyme [54–56]. However, in fish, the origin of SGCs has so far only been investigated in Medaka. Using cell lineage-labeling, SGCs in Medaka have been shown to originate from the lateral mesoderm, which also likely includes precursors of the coelomic epithelium [57]. While similar studies have not been completed in zebrafish, a clear association of SGCs with germ cells can be detected in 5 dpf larvae, at a time when the primitive gonad is in direct contact with the adjacent coelomic epithelium lining the swim bladder [29]. This suggests that the coelomic epithelium is the likely source of SGCs in zebrafish. Although at present we cannot rule out the possibility that fgf24 is involved in early specification of somatic gonad precursor cells, our analysis shows that SGCs are present in fgf24 mutants, and that mutant cells, based on EdU incorporation, have significantly lower rates of proliferation than do wild-type SGCs. These data strongly argue that fgf24 is necessary for the expansion of the SGC population, but not for their initial specification.
An interesting finding from this study is that even though the early larval somatic gonad in zebrafish is composed of relatively few cells, by 8 dpf at least two distinct cell populations can be identified. In wild-type animals, fgf24 is expressed in most gonads by 8 dpf, but at this time point its expression does not appear to be spatially restricted. By 10 dpf, however, all fgf24-expressing cells are localized to an outer layer of SGCs surrounding an inner population of fgf24(-) cells. Coincident with this observation, formation of a basal lamina between these two layers is evident based on TEM and Laminin localization, and cells of the outer layer begin to express cell-junction components that are characteristic of epithelial cells (e.g. Cdh2 and Tjp1). Thus, it appears that by 10 dpf the zebrafish somatic gonad has organized into an inner mesenchymal-like layer of cells that are in direct contact with the germ cells, surrounded by an outer fgf24-expressing epithelial layer. Our data also suggest that development of the inner mesenchymal cells is dependent on Fgf24-mediated cell interactions with the epithelial layer reminiscent of the role of Fgf signaling in other developmental contexts (e.g. limb bud development). It is possible that the single cell layer present in 10 dpf fgf24 mutant gonads represents only one of these two cell populations, but this has been difficult to determine as the expression of nearly all SGC marker genes so far examined is either greatly reduced or absent in mutant gonads. An exception to this is Wilms tumor 1a (wt1a), which appears to be expressed in both outer and inner SGCs in wild-type and in mutant gonads. This indicates that the expression of wt1a is independent of fgf24 and therefore places Wt1a either upstream of, or in parallel with, Fgf24. In addition to wt1a, fgf24 expression can be detected in a subset of the mutant SGCs suggesting that at least some of the remaining cells have an outer SGC-like characteristic, and consistently, analysis of cell-junction components suggests that a population of epithelial-like SGCs is still present in mutants. Thus, it appears that fgf24 mutant gonads may contain at least two populations of somatic gonad precursor cells, but that these cells fail to mature into functional cell types in the absence of Fgf24 signaling.
In wild-type gonads, etv4 appears to be expressed in most if not all of the inner somatic cells present in a wild-type gonad at 10 dpf. As development proceeds, the number of cells expressing etv4 appears to decline, while at the same time, cells expressing differentiation marks such as cyp19a1a and amh increase in number. Interestingly, in 12 dpf gonads, we find little to no overlap between cells that express etv4 and those that also express cyp19a1a or amh. Furthermore, cyp19a1a-expressing cells are also distinct from amh-expressing cells. These results together indicate that by 12 dpf, the inner population of SGCs is composed of at least three distinct cell populations: etv4(+), cyp19a1a(+), and amh(+). Given the dynamics of etv4 expression relative to cyp19a1a and amh, we speculate that the etv4-expressing cells are a somatic gonad progenitor cell population that in turn gives rise to the differentiated functional cells of the gonad. If this is the case, the role of Fgf24 may therefore be to promote the development and proliferation of this progenitor population. Cell lineage analysis will be necessary to test this hypothesis.
Thus far, no Fgf ligand has been shown to be necessary for mammalian bipotential gonad development analogous to the role we have described here for Fgf24 in zebrafish. Although it is possible that early gonad development in zebrafish is fundamentally different from that of tetrapods, there is reason to believe that this is not the case. Even though our understanding of gonad development in any teleost lags behind what is known in mammals, there are likely to be more similarities than differences in the genetic mechanisms that regulate gonad development and function in these two vertebrate lineages; many genes that are known to play essential roles during gonad development and sex determination in mammals are expressed at comparable time points during the development of the teleost gonad, and in some instances mutational analysis has confirmed their conserved roles. Examples include, but are not limited to: Wt1/wt1a ([58]; this report), Nr5a1/nr5a1a (Steroidogenic factor 1/Sf1; [59]; this report), gata4 (this report), Sox9/sox9a [35, 60], and dmrt1 [61, 62]. In mice, Gata4, Nr5a1, and Wt1 are all required for early gonad development [11–14]. Molecular epistasis analysis has shown that Gata4 is required for the expression of Nr5a1, but not Wt1 [11]. Interestingly, we have shown that while all three genes are expressed in SGCs in wild-type zebrafish, gata4 and nr5a1a transcripts are not detected in fgf24 mutants, while wt1a expression appears to be normal. Thus, as in mice, the regulation of wt1a expression appears to be independent of gata4 and nr5a1a.
An apparent exception to this conservation appears to be the role that Fgf signaling plays in gonad development, although in mammals, FGF9 and FGFR2 are known to play an important role during sex determination and differentiation [63–65]. Specifically, in mice, Fgf9 is initially expressed in the sexually indifferent gonads of both sexes (starting as early as E9.5) after which its expression is stabilized only in the male gonad in response to expression of the male sex determinant Sry [64]. FGF9 both antagonizes the expression of WNT4, a female-promoting signal, and promotes the stable expression of the male-promoting SOX9 transcription factor ([66, 67]; reviewed in [68]). Disrupting the FGF9 signaling pathway leads to partial male-to-female sex reversal ([63, 64]; reviewed in [68]). However, similar to the fgf24 mutant phenotype we present here, XY mice mutant for Fgf9 experience a defect in SGC proliferation prior to the expression of Sry, signifying an earlier role for FGF9 in gonad development [62].
In addition to mice, the role of FGF9 has also been investigated in chick. In both sexes of chick, FGF9 is expressed first in the mesonephroi immediately adjacent to the early bipotential gonads and later in the epithelium that surrounds these bipotential gonads [69], a pattern that is strikingly similar to what we have reported here for fgf24. Furthermore, and regardless of sex, ectopic expression of FGF9 in the early chick gonad is sufficient to expand the apparent number of SGCs, while inhibition of FGF signaling using the Fgf receptor inhibitor SU5402 leads to an apparent reduction of SGCs. Collectively, these data argue that in tetrapods, FGF9 may function first during the formation of the early bipotential gonad in both sexes (similar to the role of Fgf24 reported here) and then again later to promote testis differentiation in males.
Finally, although there are striking similarities between the expression patterns and functions of FGF9/Fgf9 in chick and mouse early gonads and that of fgf24 in the early zebrafish gonad, Fgf24 and FGF9 belong to different FGF superfamilies: FGF9 is a member of the FGF9 superfamily, which includes FGF16 and FGF20, whereas Fgf24 is a member of the FGF8 superfamily, which includes FGF17 and FGF18 ([20, 23]; reviewed in [70]). It should be noted however that the FGF8 and FGF9 subfamilies are thought to have similar, though not identical, receptor binding specificity as measured in cell culture assays (reviewed in [70]). Interestingly, while genes encoding FGF9 and Fgf24 are both present in the genomes of representative basal ray-finned and lobed-finned fish (i.e. Spotted Gar and Coelacanth, respectively; [27, 71, 72]), FGF9 orthologs have not been found in any teleost genome to date, including zebrafish [73], and Fgf24 appears to have been lost in the lobed-fin lineage prior to the evolution of land dwelling tetrapods [23, 24]. An attractive hypothesis is that in animals with both genes, Fgf9 and Fgf24 function redundantly during early gonad development. If so, this could provide a means by which they could be lost after the divergence of the two main vertebrate lineages.
Although a limited role for Fgf signaling in early mammalian gonadogenesis has been established, it is noteworthy that upregulation of the FGF-FGFR signaling pathway has been implicated as a causal factor for promoting certain types of aggressive ovarian cancers. For example, increased expression of each of the four mammalian Fgf receptors have been found in various epithelial ovarian cancers (EOC), and drugs that block or attenuate Fgf signaling have been shown to sensitize some EOCs to certain chemotherapeutic drugs [74–79]. In addition, the Fgf-responsive gene Etv4 and its close family member Etv5 are overexpressed in certain ovarian cancers [80–82]. It is therefore possible that, like in many cancers, ovarian cancers are caused, in part, by the unregulated activity of genetic pathways that are required for normal ovarian development during embryogenesis.
We have so far focused our attention on the role of Fgf24 during the development of the larval gonad. While all fgf24 mutants have severe defects in larval gonad development, as adults 22% of fgf24 mutants have partial testes that can produce functional sperm. Although it is not known how the mutant gonads resume apparently normal development, it is clear that this development can occur in the absence of Fgf24 function. One explanation for this phenomenon is that Fgf signaling is not involved in gonad formation at later stages; however it is also plausible that a second Fgf ligand expressed during juvenile development can rescue somatic gonad development in the absence of Fgf24. If this is the case, then latent testis development could occur in mutants that retain germ cells until expression of this ligand initiates. Future experiments will be necessary to test these models.
Despite the important role the vertebrate somatic gonad plays in protecting germ cells and in regulating their maintenance and differentiation into gametes, little is known about the genetic regulation of somatic gonad development, and, in particular, the cell-cell interactions that are necessary for its development. Here, we have identified the Fibroblast growth factor ligand Fgf24 as a key player in this process. These results help to establish the zebrafish as a model for understanding the genetic regulation of early somatic gonad development in vertebrates.
The University of California Davis IACUC approved all animals used in this study (protocol #18483), and all animals used were euthanized using the American Veterinary Medical Association-approved method of hypothermal shock.
The wild-type strain *AB was used for the generation of fgf24uc47. Zebrafish husbandry was performed as previously described [83], with the following modifications to the larval fish (5-30dpf) feeding schedule: 5-12dpf: 40 fish/250mL in static fish water (4parts/thousand (ppt) ocean salts) were fed rotifers (Brachionus plicatilis, L-type) twice daily ad libitum. 12–15 dpf: 40 fish/ 1 liter gently flowing fish water (<1ppt ocean salts) were fed both rotifers and freshly hatched Artemia nauplii ad libitum twice daily. 15-30dpf: 40 fish/1 liter gently flowing fish water (<1ppt ocean salts) were fed freshly hatched Artemia nauplii ad libitum twice daily. The following alleles were used in this study: fgf24hx118, fgf24uc47, tp53zdf1. The following transgenic lines were used: Tg(ziwi:EGFP)uc02, Tg(gsdf:mCherry)uc46.
The sgRNA and cas9 mRNA components were produced as previously described [84]. Briefly, the sgRNA was designed to target exon 3 of fgf24 (zifit.partners.org/ZiFiT/). Two oligonucleotides (5’-TAGGCAAGAAGATTAACGCCAA-3’ and 5’-AAACTTGGCGTTAATCTTCTTG-3’) were annealed and cloned into plasmid pDR274 (Addgene Plasmid #42250). The plasmid was linearized with DraI, and in vitro transcription was performed with the T7 polymerase (Roche, Cat. No. 10881775001). The cas9-expressing pMLM3613 plasmid was also obtained from Addgene (Plasmid #42251) and mRNA synthesis was performed as described [84]. The sgRNA and cas9 mRNA were coinjected into one-cell embryos with phenol red (5% in 2M KCl) at a concentration of 12.5 ng/μL and 300 ng/μL, respectively. CRISPR efficiency was determined by comparing the DNA isolated from eight injected embryos with eight uninjected control embryos (24 hpf) using High Resolution Melt Analysis (HRMA) as described [85] (the primers used are listed in S1 Methods). At three months post-injection, germline mutations were identified from extracted sperm of injected males by PCR analysis and gel electrophoresis. PCR products with evident indels were cloned into the pGEM-T Easy vector (Promega, Cat. No. A137A) and sequenced. The individual containing the fgf24uc47 allele was outcrossed to *AB to obtain a heterozygous line.
Fish were genotyped by extracting gDNA from caudal fin tissue. fgf24uc47, Tg(ziwi:EGFP)uc02, and Tg(gsdf:mCherry)uc46 were assayed using standard PCR conditions and gel electrophoresis. fgf24hx118 and tp53zdf1 were analyzed using HRMA. See S1 Methods for primer sequences.
Male fish were euthanized in an ice water bath. Testes were removed and macerated with scissors in Hank’s solution [83]. Eggs were squeezed from wild-type females according to standard protocols [81] and were fertilized with 30 μL sperm from either wild-type or fgf24-/- males. After three hours, the numbers of fertilized and unfertilized eggs were recorded.
RNA probes that detect the following genes were used: cyp19a1a and amh [33]; gata4 [86]; wt1a [58]; etv4 [87]. For all other plasmids for probe synthesis, mRNA was isolated from 24 hpf embryos, adult testis, or ovary using TRI reagent (Sigma-Aldrich, Cat. No. T9424) and synthesized into cDNA using the RETROScript Reverse Transcription Kit (ThermoFisher, Cat. No. AM1710). Targets were PCR amplified with Takara Ex Taq (Clontech, Cat. No. RR001A) and primers found in S1 Methods, Table 2. PCR products were cloned into an appropriate vector, and plasmids were linearized using endonucleases from New England Biolabs (see S1 Methods for details). in vitro transcription yielded antisense probes (Roche T7 or SP6 RNA polymerases (Cat. Nos. 10881775001 or 10810274001, respectively)). Probes were ethanol precipitated and G-50 sephadex column purified to remove excess nucleotides (GE Healthcare, Cat. No. 45-001-398). Probes were used at a concentration of 0.5-2ng/μL in hybridization solution.
Fish were fixed in 4% paraformaldehyde (PFA) overnight at 4°C or 4 hours at room temperature. Samples were transferred to 100% methanol and stored at -20°C for at least 16 hours. Samples were bleached for 10–15 minutes, as needed, prior to proteinase K digestion in 3% H2O2, 0.5% KOH. Color in situ hybridizations were performed similar to Thisse and Thisse, 2008 [88], with the exception that 5% dextran sulfate was included in the hybridization solution. Fluorescent in situ hybridizations were performed as in Lauter et al., 2011 [89]. Tyramide reactions were performed with commercially available tyramides (Life Technologies, Cat. Nos. T20948, T20950, and T20951). Gonads were dissected, and DNA was stained with DAPI. Samples were dehydrated by an increasing glycerol gradient. Gonads were mounted whole and imaged with an Olympus FV1000 laser scanning confocal microscope.
Tissue was prepared and treated as for ISH. After initial washing, nonspecific antibody was blocked with 2% BSA and 2% goat serum in PBS-DT (1% PBS + 0.1% Triton-X + 1% DMSO) for one hour at room temperature. Antibodies were diluted in blocking solution according to S1 Methods and applied to tissue overnight at 4°C. Following washing, blocking was repeated. Alexa Fluor secondary antibodies (Thermo Fisher Scientific, Cat. nos. A-11008, A-11012, A-11005, A-11001, A-11039, A-11042) were diluted at 1:500 in blocking solution and incubated with tissue overnight at 4°C to detect primary antibodies. DNA was stained with DAPI, and samples were dehydrated by an increasing glycerol gradient. Gonads were dissected, mounted whole, and imaged with an Olympus FV1000 laser scanning confocal microscope.
For the experiment shown in Fig 6G–6H’, an anti-Mouse-HRP conjugated secondary antibody (ThermoFisher, Cat. no. G-21040) was used to detect anti-pHH3, and tyramide reactions were performed as described above. For the experiment shown in Fig 7C and 7D, both Laminin and Vasa antibodies were raised in rabbit. Staining was therefore performed as above, but sequentially; briefly, samples were incubated overnight at 4°C with Rabbit anti-Laminin, washed, and incubated with a Goat anti-Rabbit IgG, Alexa Fluor 488 overnight at 4°C. After extensive washing, samples were incubated overnight at 4°C with Rabbit anti-Vasa, washed, and incubated with a Goat anti-Rabbit IgG, Alexa Fluor 594 overnight at 4°C. Samples were then treated as above.
Apoptosis was detected with the ApopTag Apoptosis Detection Kit (Millipore, Cat. No. S7110). Samples were treated according to the manufacturer’s manual, with additional post-fixation steps after proteinase K digestion: Samples were treated with 4% PFA for 20 minutes at room temperature, washed 5 X 5 minutes in PBSTw (PBS + 0.1% Tween-20), incubated in pre-chilled 2:1 EtOH:acetic acid for 10 minutes at -20°C, and washed 3 X 5 minutes in PBSTw.
Larvae were allowed to swim freely in 200 μM EdU + 0.1% DMSO. To diminish any systemic affects of treatment, fish were kept at normal densities and on normal feeding schedules. Fish were euthanized and fixed in 4% PFA immediately following exposure. After extensive washing in PBSTw, EdU was detected by “click” chemistry (10 μM Alexa Fluor 594 Azide (ThermoFisher, Cat. no. A10270), 1 mM CuSO4, 100 mM Tris pH8.5, 100 mM Ascorbic acid; incubate for 30 minutes at room temperature) and visualized on an Olympus FV1000 laser scanning confocal microscope.
Sagittal optical sections were collected at 5 μm intervals throughout the entirety of whole-mount gonads with an Olympus FV1000 laser scanning confocal microscope. Intervals of 5 μm were used to allow for identification of virtually every cell in a gonad. Individual cells were manually documented with the Cell Counter plugin for FIJI. Germ cells were identified by Vasa expression. Because we lack a pan-SGC marker and because gonads of the stages described here do not readily dissect from the body wall, we were not confident in our ability to quantify the total number of SGCs in any given gonad. However, for the EdU experiment, we were able to roughly identify SGCs based on their proximity to germ cells and overall shape of the tissue. Once somatic cells of a gonad were identified and recorded, overlap of EdU signal was scored.
Tg(ziwi:gfp)uc02 fish were euthanized in an ice water bath. Tails were removed immediately posterior to the gonads as visualized by the germ cell-specific GFP. They were then fixed 24 hours in Karnovsky's fixative (2.5% glutaraldehyde + 2% paraformaldehyde in 0.1 M sodium cacodylate), following 2 X 15 min rinses in 0.1 M sodium cacodylate buffer. Samples were then treated with 2% osmium tetroxide for 1 hour, followed by 2 X 15 minutes rinses in 0.1 M sodium cacodylate buffer. Tissue was dehydrated in an ethanol gradient (30 minutes each: 50%, 75%, 95% EtOH; 2 X 20 minutes 100% EtOH) and treated with propylene oxide 2 X 10 minutes. Tissue was pre-infiltrated with 1:1 propylene oxide:Poly/Bed 812 resin overnight and infiltrated with 100% Poly/Bed 812 resin for three hours (Polysciences, Inc). Finally, samples were embedded in fresh resin, polymerized in a 60°C oven for 24 hours, and sectioned to 100 nm. Sections were imaged using a Philips BioTwin CM120 TEM.
Images were analyzed using FIJI, and only linear manipulations of brightness and contrast were applied. Statistical analysis and graphing were completed in R using standard packages, ggplot2, and ggbeeswarm.
|
10.1371/journal.pgen.1005977 | Palmitoylation of the Cysteine Residue in the DHHC Motif of a Palmitoyl Transferase Mediates Ca2+ Homeostasis in Aspergillus | Finely tuned changes in cytosolic free calcium ([Ca2+]c) mediate numerous intracellular functions resulting in the activation or inactivation of a series of target proteins. Palmitoylation is a reversible post-translational modification involved in membrane protein trafficking between membranes and in their functional modulation. However, studies on the relationship between palmitoylation and calcium signaling have been limited. Here, we demonstrate that the yeast palmitoyl transferase ScAkr1p homolog, AkrA in Aspergillus nidulans, regulates [Ca2+]c homeostasis. Deletion of akrA showed marked defects in hyphal growth and conidiation under low calcium conditions which were similar to the effects of deleting components of the high-affinity calcium uptake system (HACS). The [Ca2+]c dynamics in living cells expressing the calcium reporter aequorin in different akrA mutant backgrounds were defective in their [Ca2+]c responses to high extracellular Ca2+ stress or drugs that cause ER or plasma membrane stress. All of these effects on the [Ca2+]c responses mediated by AkrA were closely associated with the cysteine residue of the AkrA DHHC motif, which is required for palmitoylation by AkrA. Using the acyl-biotin exchange chemistry assay combined with proteomic mass spectrometry, we identified protein substrates palmitoylated by AkrA including two new putative P-type ATPases (Pmc1 and Spf1 homologs), a putative proton V-type proton ATPase (Vma5 homolog) and three putative proteins in A. nidulans, the transcripts of which have previously been shown to be induced by extracellular calcium stress in a CrzA-dependent manner. Thus, our findings provide strong evidence that the AkrA protein regulates [Ca2+]c homeostasis by palmitoylating these protein candidates and give new insights the role of palmitoylation in the regulation of calcium-mediated responses to extracellular, ER or plasma membrane stress.
| Palmitoylation is a reversible post-translational modification catalyzed by palmitoyl acyltransferases (PATs) and proteins that undergo this modification are involved in numerous intracellular functions. Yeast Akr1p was the first characterized PAT whilst HIP14, an Akr1p homolog in human, is one of the most highly conserved of 23 human PATs that catalyze the addition of palmitate to the Huntington protein which is of major importance in Huntington’s disease. Calcium serves numerous signaling and structural functions in all eukaryotes. However, studies on the relationship between calcium signaling and palmitoylation are lacking. In this study, we demonstrate that the palmitoyl transferase Akr1 homolog in the filamentous fungus Aspergillus nidulans, similar to the high-affinity calcium uptake system (HACS), is required for normal growth and sporulation in the presence of low extracellular calcium. We find that AkrA dysfunction decreases the transient increase in cytosolic free calcium induced by a high extracellular calcium stress, tunicamycin (which induces endoplasmic reticulum stress) or the antifungal agent itraconazole (which induces plasma membrane stress). The influence of AkrA on all of these processes involves its DHHC motif, which is required for palmitoylation of various proteins associated with many processes including calcium signaling and membrane trafficking. Our findings provide evidence for a crucial link between calcium signaling and palmitoylation, suggesting a possible role in the mechanistic basis of human PAT-related diseases. These results also indicate that regulators of posttranslational modification may provide promising antifungal targets for new therapies.
| In all eukaryotic cells, the cytosolic free calcium ([Ca2+]c) concentration is strictly and precisely controlled by complex interactions between various calcium-channels, calcium-pumps and calcium-antiporters and by calcium buffering in the cytoplasm. Finely tuned changes in [Ca2+]c mediate a variety of intracellular functions, and disruption of [Ca2+]c homeostasis can lead to various pathological conditions [1]. In fungi, numerous studies have shown that calcium signaling is involved in regulating a wide range of processes including cell morphogenesis, cell cycle progression, stress responses and virulence [2]. Two different calcium uptake systems in the plasma membrane have been identified in most fungal species: the high-affinity Ca2+ influx system (HACS) and the low-affinity calcium influx system (LACS) [3–5]. The main components of the HACS are primarily composed of an α-subunit of the mammalian voltage-gated Ca2+-channel homolog Cch1 and a stretch-activated β-subunit called Mid1. Loss of the HACS results in an inability to grow under low-calcium conditions. In addition, fungi possess a range of other calcium P-type ATPases and calcium transporters that play important roles in calcium signaling and homeostasis [6]. Upon stimulation, calcium is rapidly taken up from the extracellular environment or released from these intracellular calcium stores and either interacts with the primary intracellular calcium sensor/receptor calmodulin or directly regulates that activity of other proteins. When the calcium signal binds to calmodulin this results in a conformational change in the protein allowing it to interact with and regulate the activity of various target proteins involved in converting the original stimuli into cellular responses. The [Ca2+]c increase is transient because various calcium-pumps and calcium-antiporters, as well as the cytoplasmic calcium buffering, subsequently return the [Ca2+]c to its normally low resting level within the cytosol [7,8].
The phosphatase calcineurin is an important [Ca2+]c transient effector and is conserved from yeast to humans. Its most well known target in fungi is the transcription factor Crz1 (calcineurin responsive zinc finger 1) [9,10]. In vegetatively growing S. cerevisiae cells, [Ca2+]c concentrations are normally maintained at low non-signaling levels. During this stage, Crz1 is fully phosphorylated, localized to the cytoplasm, and transcriptionally inactive [11,12]. When fungal cells are exposed to chemicals that induce plasma membrane stress (e.g. by azole antifungals) or endoplasmic reticulum (ER) stress (e.g. by tunicamycin), or are under low calcium conditions, the HACS is activated. These stimuli result in calcium uptake and a transient increase in [Ca2+]c which leads to calcineurin activation and subsequent Crz1 de-phosphorylation. Crz1 is then recruited to nuclei where it transcriptionally regulates downstream signaling pathways to alleviate cellular stress and promote cell survival [13,14]. Interestingly, there are no known mammalian Crz1 orthologs, but mammals express another calcineurin sensitive transcription factor target, known as NFAT (nuclear factor of activated T-cells). Crz1 does not belong to the NFAT family, but the Zn-finger domains in Crz1 and NFAT bind specific DNA sequences within the promoter regions of calcineurin-dependent response elements (CDREs) to activate transcription [15,16]. In the filamentous fungus Aspergillus nidulans, there is a calcineurin-dependent Crz1 homolog, known as CrzA. Interestingly, calcineurin deletion causes more severe growth defects than CrzA deletion in this species, suggesting that calcineurin has additional target proteins other than CrzA [17,18].
Palmitoylation is a reversible posttranslational modification that catalyzes the attachment of palmitate to cytoplasmic cysteine residues of soluble and transmembrane proteins. Palmitoyl transferases (PATs) are known to be responsible for palmitoylation. The defining feature of PATs is the presence of a cysteine-rich domain (CRD) with an Asp-His-His-Cys (DHHC) motif, which is required for PAT activity. Many proteins that require palmitoylation are involved in cellular signaling, membrane trafficking and synaptic transmission [19–21]. There are more than 20 encoded DHHC proteins in mammalian genomes, and there is now a major effort to verify DHHC-substrate partners and determine how their interaction specificity is encoded [22]. Several lines of recent evidence have shown that protein palmitoylation influences various cell functions, physiology and pathophysiology [23–25].
In this study, we have demonstrated that AnAkrA in A. nidulans and AfAkrA in A. fumigatus, which are homologs of the yeast palmitoyl transferase ScAkr1p, have similar function to the HACS in the presence of low extracellular calcium. The akrA deletion resulted in marked defects in hyphal extension and conidiation, especially under low calcium conditions. Moreover, using codon-optimized aequorin as a calcium reporter in living cells, we found that AkrA dysfunction significantly decreased the amplitude of the [Ca2+]c transient induced by an extracellular calcium stimulus, ER stress caused by tunicamycin or plasma membrane stress resulting from itraconazole, respectively. Our data suggest that these [Ca2+]c responses are mediated by the palmitoylation of the cysteine residue of the DHHC motif in AkrA. Moreover, we have identified that two new putative P-type ATPases (Pmc1 and Spf1 homologs), a putative proton V-type proton ATPase (Vma5 homolog) and three putative CrzA-dependent proteins, are palmitoylated substrates of the AkrA protein. To our knowledge, this is the first report that a palmitoylation protein is involved in regulating eukaryotic calcium signaling.
Based on a NCBI BLASTp search (http://www.ncbi.nlm.nih.gov/BLAST/), we identified a putative ortholog of NFAT in A. nidulans, AkrA (AN5824.4, Accession: XP_663428.1), which encodes a putative palmitoyltransferase. However, it showed low identity (less than 20%) or similarity (less than 30%) to mammalian NFAT based on full-length sequences. Interestingly, a bioinformatic analysis revealed that the promoter region contains a putative calcineurin-dependent-response-element (CDRE-like) motif. As shown in Fig 1A, we identified a CDRE-like sequence at 398 bp (akrA, AN5824.4), upstream of this gene’s start codon [26,27]. These data suggest that AkrA may be a component of the calcium signaling machinery. To further explore the function of the akrA gene and its relationship to calcineurin, the full-length deletion strain was constructed by homologous gene replacement employing a self-excising recyclable cassette that contains an AfpyrG gene as a selectable marker. Diagnostic PCR analysis of the resulting strain ΔakrA confirmed the homologous replacement (S1A Fig). We also generated ΔakrAΔcnaA double mutants through genetic crosses (the cnaA gene encodes the catalytic subunit of calcineurin). The ΔakrA mutant produced smaller colonies compared to that of the parental wild-type strain, when grown on minimal medium. In comparison, the ΔcnaA mutant exhibited severe growth defects on minimal medium. Moreover, the double mutant had a smaller colony size and underwent less conidiation than the single mutants (Fig 1B). These results suggest that akrA and cnaA may have different functions in A. nidulans. Therefore, the double deletion mutant exacerbates the growth defects on minimal medium.
We next tested whether low external calcium conditions could affect the colony phenotype in the akrA deletion mutant. When conidia were spot inoculated onto the solid minimal medium containing the calcium chelator EGTA and were allowed to grow at 37°C for 2.5 days, the ΔakrA mutant exhibited increased EGTA sensitivity compared to the parental wild-type strain. As shown in Fig 1C, the akrA deletion exhibited markedly reduced conidial formation and colony growth under low-calcium conditions. Since, mutants of the HACS components have been previously shown to exhibit similar defects under low calcium conditions [28–30], we next examined whether AkrA was a potential novel HACS component. To determine whether the defects in the ΔakrA mutant could be rescued by high extracellular calcium, we inoculated ΔakrA mutant conidia on minimal medium supplemented with 20 mM Ca2+. We found that the colony diameter of the ΔakrA mutant was restored almost to the same diameter of the parental wild-type strain by the addition of extracellular calcium (Fig 1C), indicating that exogenous calcium could completely rescue the colony growth defect caused by AkrA loss. We further examined conidiation in the ΔakrA mutant in a calcium-limited environment (i.e. in the presence of EGTA) with a stereomicroscope (Fig 1D left panels). The results showed that the vegetative mycelia from the parental wild-type strain were capable of producing numerous conidia under low-calcium conditions. In contrast, conidiation was almost completely abolished in the ΔakrA mutant on minimal media supplemented with EGTA (1 mM) (Fig 1D left panels). In submerged liquid culture, the wild-type strain displayed robust polarized hyphal growth around the margins of mycelial balls, whereas the ΔakrA mutant showed smooth margins around small mycelial balls (Fig 1D right panels). Consistently, the ΔakrA mutant had a significantly reduced biomass, germination rate, and colony size compared to the parental strain on minimal media (S3 Fig). Moreover, ectopically expressed akrA was able to completely rescue these defects in the akrA deletion strain (Fig 1D), establishing that these phenotypes were specific to the loss of akrA. In addition, we deleted the akrA homolog gene in A. fumigatus. Similar to the ΔakrA phenotypes in A. nidulans, the ΔAfakrA mutant displayed hypersensitivity to the low calcium conditions, and its phenotypic defects could be rescued by high extracellular calcium (S2 Fig). Thus, these data are consistent with AkrA being involved in calcium uptake especially in a calcium-limited environment.
To further confirm and assess the localization and the molecular mass of AkrA, we generated a conditional expression allele, alcA(p)::GFP-akrA, referred to here as ZYA09 (S1B Fig). In this conditional allele, akrA expression was assumed to be regulated by the carbon source, as it was not induced by glucose, induced by glycerol, and overexpressed to high levels by L-threonine [31]. To determine whether this conditional allele behaved as predicted, we inoculated the ZYA09 strain in liquid media for 18 h, which promoted induction, non-induction or overexpression. As expected, the akrA mRNA level was approximately 20-fold higher when grown in overexpressing medium compared to that grown in non-inducing medium, which was 12-fold higher than that in inducing medium (S4B Fig). Moreover, the conditional strain ZYA09 displayed an identical phenotype to the parental wild-type strain when grown on the inducing or the overexpressing media, indicating that the fusion GFP-AkrA protein was functional and that the assumed akrA over-expression had no detectable effects in A. nidulans. In comparison, when grown on the non-inducing medium, the conditional allele alcA(p)::GFP-akrA exhibited an identical phenotype to the ΔakrA mutant, confirming a consistent phenotype for the loss of AkrA and for the knock-down of AkrA (Figs 2A and 1C). Western blotting showed a band at approximately 110 kDa in the GFP-AkrA strain grown under inducing or overexpressing conditions using an anti-GFP antibody but no such a band appeared in the parental wild-type strain or the conditional allele (ZYA09) under the non-inducing condition (Fig 2B). These results indicate that the molecular mass of AkrA is approximately 80 kDa because GFP is a 27 kDa protein.
Microscopic examination showed that the AkrA-GFP localization pattern resembled that of the Golgi previously reported in A. nidulans [32]. To confirm this we generated the strain ZYA13 by genetically crossing the alcA(p)::GFP-akrA strain ZYA09 with the MAD2013 strain in which the late Golgi marker (gpdAmini::mRFP-PHOSBP), consisting of the pleckstrin homology domain of the human oxysterol binding protein (PHOSBP) fused to mRFP was included [33,34]. Spores of the ZYA13 strain were incubated in non-inducing medium at 37°C for 10 h and were then shifted to the overexpression medium for 6 h. Microscopic examination of the young germlings produced under these conditions showed the majority of GFP-AkrA proteins colocalized with mRFP-PHOSBP late Golgi marker (Fig 2C).
Because the bioinformatic analysis showed that AkrA contains a conserved DHHC motif required for its palmitoylation activity [19–21], we next investigated whether the DHHC motif was required for the normal function of AkrA under low calcium conditions. We first constructed a C-terminal AkrA truncation lacking the region from the DHHC motif through to the stop codon by homologous gene replacement (Fig 3A). The colony phenotype of the truncation mutant was similar to that resulting from the complete deletion of the akrA gene when grown in minimal medium plus EGTA, indicating that the DHHC motif is required for AkrA function (Fig 3B). To rule out the possibility that a loss of function in the truncated mutant might result from a conformational change that prevented a true reflection of the function of the DHHC motif, we performed site-directed mutagenesis. Since Cys487 in the DHHC motif has previously been shown to be crucial for palmitoyl transferase activity, we therefore mutated Cys487 to Ser487 in the DHHC motif (Fig 3A) [35,36]. Consequently, we found that the C487S site-mutated DHHS fragment could not rescue the defect of the akrA deletion mutant under either the control of a native promoter (native(p)::akrAC487S) or a GPD promoter (GPD(p)::akrAC487S) (Fig 3B). In comparison, the wild-type akrA gene completely rescued the growth defects in the akrA deletion recipient strain. To confirm that these fusion cassettes were transcribed in the transformant, we performed quantitative real-time PCR to verify the akrA mRNA levels. The results showed that both the GPD and native promoters induced normal akrA mRNA expression, even though the mRNA expression level under the control of the GPD promoter was higher than that with the native promoter (S4D and S4E Fig), indicating that the AkrA-DHHS cassettes were fully transcribed. Next, we generated Flag-tagged AkrA and the site mutated AkrAC487S strains to further confirm the expression of the AkrA protein. As shown in Fig 3C, the predicted bands on a Western blot were observed clearly, suggesting that both Flag-AkrA and Flag-AkrAC487S proteins were fully expressed in vivo. In addition, the Flag-tagged AkrAC487S strain displayed an identical phenotype to that of the Flag-untagged (native(p)::akrAC487S) mutant, suggesting that the Flag tag could not phenotypically change the function of the targeted protein AkrA (Fig 3B and 3D). These data suggest that the growth defect caused by akrA deletion was closely associated with the Cys487 site in the DHHC motif.
Because the loss of akrA caused a similar defect phenotype to that of deletion mutants of the HACS components cchA and midA under the low calcium conditions, we hypothesized that AkrA forms a complex with CchA or MidA to perform its function. To assess whether AkrA physically interacts with CchA or MidA, we performed yeast two-hybrid assays. We cloned the cDNA fragments corresponding to the cytosolic C-terminus of cchA and the full-length cDNA of midA, respectively. They were then amplified and cloned into the pGADT7 vector, which contains the GAL4 DNA-AD and the LEU2 marker. In addition, a full-length cDNA of akrA was cloned into the pGBKT7 vector, which contains the GAL4 DNA-BD and TRP1 marker. As a result, some small colonies of pGBKT7-akrA with pGADT7-cchA were obtained, and there was no detectable growth of colonies of pGBKT7-akrA with pGADT7-midA under the high stringency screening conditions compared to the positive colonies of pGADT7-T and pGBKT7-53, which showed robust growth (S4A Fig). These data suggest that AkrA and MidA do not directly interact, and that AkrA and CchA might weakly and transiently interacted.
We next investigated the functional interaction(s) between AkrA and CchA and between AkrA and MidA by a genetic phenotypic analysis. The ΔakrAΔmidA, ΔakrAΔcchA double mutants were generated by genetic crossing. As shown in Figs 4A and S6, phenotypic defects in colony size and conidiation were exacerbated in the double mutants compared to the parental single mutants, especially in the presence of EGTA. Notably, the growth retardation of the ΔakrAΔmidA and ΔakrAΔcchA double mutants under low calcium conditions was reversed by the addition of 20 mM calcium to the minimal medium. These results suggest that AkrA, CchA, and MidA are all required under the calcium-limited condition, but may have some non-overlapping roles in growth. To determine whether overexpression of cchA could rescue the ΔakrA defects under the low calcium condition, we crossed ΔakrA (ZYA02) and alcA(p)::GFP-cchA (ZYA11) to generate the ZYA12 strain. Real-time PCR verified that the mRNA level of cchA in ZYA12 was approximately 15-fold higher in the overexpressing medium than in the inducing medium when cultured for 18 h (S4C Fig). However, overexpression of cchA did not rescue the ΔakrA defects under low calcium conditions (Fig 4B).
Previous studies have demonstrated that pmr1, which encodes a Ca2+/Mn2+ P-type ATPase and is involved in Ca2+ homeostasis, localizes to the Golgi in yeast [37]. In A. nidulans, ΔpmrA had no discernible effect on fungal physiology, but the cells were hypersensitive to low extracellular calcium [38]. To investigate the link between AkrA and PmrA, we crossed the ΔakrA and ΔpmrA mutants. Surprisingly, the double mutant had no detectable defect when grown in minimal medium compared to the ΔakrA strain, which had a reduced-colony size (Fig 4A). These data suggest that the pmrA deletion suppressed the ΔakrA growth defect. However, when cultured on minimal medium with 1 mM EGTA, the double mutant showed an exacerbated growth retardation phenotype compared to the parental single mutants. In addition, the phenotypic defects of ΔakrAΔpmrA were completely suppressed by the addition of 20 mM calcium. These results suggest that AkrA and PmrA may operate together in regulating cellular calcium homeostasis in a reverse way.
Previous studies with yeast reported that Cch1 and Mid1 mutations reduced calcium uptake and affected [Ca2+]c accumulation under both stimulating and non-stimulating conditions [5,39–41]. We monitored the extracellular calcium-induced [Ca2+]c changes in living cells of A. nidulans wild type and mutant strains in which we expressed codon-optimized aequorin [42–44]. When treated with 0.1 M CaCl2, the [Ca2+]c concentration in wild type cells transiently increased from a resting level of approximately 0.1 μM to a peak concentration of 1.2 μM (Fig 5). In comparison, cchA or midA mutants showed a reduction of 17 ± 11% or 25 ± 12% in the [Ca2+]c amplitudes, respectively, under the same stimulating conditions. Surprisingly, the decrease in the [Ca2+]c amplitude in akrA mutants was much larger than that observed in the HACS mutants. The [Ca2+]c amplitudes were decreased as follows: 53 ± 13% in the akrA deletion strain ZYA02, 54 ± 9% in the DHHC truncated mutant ZYA15, and 55 ± 8% in the site-mutated native(p)::akrAC487S mutant ZYA16. These data suggest the significant reduction in calcium influx due to the loss of AkrA is mediated by the DHHC motif and, in particular, the cysteine residue within the DHHC motif.
The [Ca2+]c amplitude in the ΔpmrA mutant exposed to the 0.1 M CaCl2 stimulus was similar to that of the parental wild-type strain, which is different from that previously reported for yeast [45–47], suggesting that other Ca2+-ATPases may compensate for the loss of PmrA function in response to the extracellular calcium stimulus. However, loss of pmrA in the akrA deletion background was able to recover the decreased [Ca2+]c amplitude in the akrA mutant to a similar level as that in the parental wild-type strain in response this extracellular calcium stimulus, indicating that the perturbation of calcium homeostasis induced by AkrA could be rescued by loss of pmrA.
The protein palmitoylation inhibitor 2-bromopalmitate (2-BP) is a palmitate analog that blocks palmitate incorporation into proteins [48,49]. To determine whether inhibition of palmitoyl transferase activity influences calcium influx into the cytoplasm, we measured the [Ca2+]c amplitude of the wild type pre-incubated in 2-BP (20 μM) for 2 h. Following this drug treatment, the amplitude of the [Ca2+]c increase following stimulation with 0.1 M CaCl2 was significantly reduced by approximately 40% of the untreated cells in response to stimulation with 0.1 M CaCl2 (Fig 5). These data suggest that the inhibition of palmitoyl transferase activity can significantly block calcium influx.
Activation of Ca2+ channels, calmodulin, calcineurin and other factors is necessary for the long-term survival of cells undergoing ER stress, and during this process the HACS components, CchA and MidA, are required for Ca2+ influx from the extracellular environment [41,50,51]. To verify whether AkrA is involved in the calcium influx response during ER stress, we measured the influence of the ER-stress agents, tunicamycin (TM) and dithiothreitol (DTT) on [Ca2+]c. When the parental wild-type strain was treated with 5 μg/mL tunicamycin, we observed an immediate transient increase in [Ca2+]c with an amplitude of 0.60 ± 0.03 μM (Fig 6B). In comparison, the [Ca2+]c amplitude in the ΔcchA mutant (but not the ΔmidA mutant) in response to tunicamycin was decreased by approximately 32 ± 6%, suggesting that the loss of CchA but not MidA mediates the ER stress-induced calcium influx in A. nidulans. Furthermore, in response to tunicamycin treatment the [Ca2+]c amplitude decreased by 40 ± 5%, 34 ± 8% and 34 ± 6% in the ΔakrA, akrAΔC, native(p)::akrAC487S mutants, respectively. We next examined the [Ca2+]c response after addition of DTT, another agent causing ER-stress. 10 mM DTT induced a rapid increase in [Ca2+]c which peaked at approximately 0.40 μM in the wild-type and ΔmidA strains, but the [Ca2+]c amplitudes decreased by approximately 40% in the ΔakrA (36 ± 10%), akrAΔC (37 ± 7%), and native(p)::akrAC487S (36 ± 8%) mutants, and by 15 ± 9% in the ΔcchA mutant (S7 Fig). These data suggest that CchA, but not MidA, influences the ER stress-induced calcium influx in A. nidulans, which is different from that previously reported in yeast [41,51]. Furthermore, loss of AkrA, or mutations in its DHHC significantly decreased the ER stress-induced calcium influx. We further tested whether the amplitude of the [Ca2+]c increase in response to tunicamycin was dependent on the extracellular calcium concentration. We found that there was no significant change when mycelia were cultured in media with or without 5 mM calcium (S8A Fig). In contrast, exposure of cells to 1 mM EGTA prior to tunicamycin treatment completely abolished the increase in [Ca2+]c in the ΔakrA, akrAΔC and native(p)::akrAC487S mutants, but not in the parental wild-type, ΔcchA or ΔmidA strains (Fig 6A). Similar data was obtained when we used the more selective, calcium chelator BAPTA (S9 Fig). These data suggest that intracellular calcium stores contribute to the transient increase in [Ca2+]c induced by agents causing ER stress.
Because azole antifungal drugs induce plasma membrane stress [13,14,52], we next compared the differences in the [Ca2+]c transient between wild-type and relevant mutant strains after treatment with the azole antifungal agent itraconazole (ITZ), which is currently used as a primary antifungal drug in the clinic. In all the tested mutants and the wild-type strain, the [Ca2+]c resting levels were similar at approximately 0.05 μM. After addition of 1 μg/mL ITZ to the medium, all strains responded with a transient increase in [Ca2+]c (Fig 7B). However, all the akrA defective mutants exhibited significantly lower increases in [Ca2+]c compared to their parental wild-type strain: the amplitudes of the [Ca2+]c transients were reduced by 36 ± 11% in the ΔakrA, 29 ± 10% in the akrAΔC, 24 ± 8% in the native(p)::akrAC487S and 27 ± 8% in the ΔcchA mutants, respectively, compared to that of the parental wild-type strain. In marked contrast to these mutants, the ΔmidA mutant exhibited a similar [Ca2+]c amplitude in response to ITZ as observed in the wild-type strain. In addition, the amplitude of the ITZ-induced [Ca2+]c elevation increased when mycelia were cultured in media containing 5 mM CaCl2 (S8B Fig). We next examined whether the [Ca2+]c transient induced in response to ITZ was dependent on external calcium or internal calcium stores. We exposed hyphal cells to media supplemented with EGTA (1 mM) prior to ITZ treatment, and found that [Ca2+]c transients were dramatically abolished in all the ΔakrA mutants, whereas the [Ca2+]c transients in the wild type, and the ΔcchA and ΔmidA mutants, were still observed (Fig 7A). Similar data were obtained when we used the calcium chelator BAPTA (S9 Fig). These data indicate that the loss of AkrA or disruption of its DHHC motif in the absence of extracellular calcium completely block calcium influx after treatment with chemicals that induce ER or plasma membrane stress from both extracellular and intracellular sources. Furthermore, both extracellular calcium and intracellular calcium stores play roles in generating these [Ca2+]c transients induced by these stress treatments.
Our evidence above indicates that the cysteine residue in the DHHC motif of AkrA is involved in regulating the calcium response to high extracellular calcium-, ER- and plasma membrane-stress. To test whether the cysteine residue of DHHC is required for AkrA palmitoylation, we set up an acyl-biotin exchange (ABE) chemistry assay to detect palmitoylation in potential target proteins based on selective thioester hydrolysis by hydroxylamine (HA) (Fig 8A). Compared to the control, the treatment of hydroxylamine combined with N-ethylmaleimide (NEM) (which blocks free sulhydryls), efficiently enriches palmitoylated proteins. Subsequent treatment with HA cleaves the thioester bond between palmitate and cysteine residues, exposing bound thiols, which are then covalently linked to HPDP-biotin. The controls were protein samples not treated with HA. Lastly, the biotinylated proteins were bound to streptavidin agarose, washed with buffer, and eluted by cleavage of the cysteine-biotin disulfide linkage following by SDS-PAGE. Several previous reports have suggested that the process of palmitoylation involves in a two-step mechanism in which palmitoyl transferase is auto-acylated by itself to create an intermediate followed by the transfer of the palmitoyl moiety to its substrate [53,54]. Therefore, to investigate whether the cysteine residue in the DHHC motif is responsible for AkrA auto-acylation, we used the ABE assay to detect whether AkrA palmitoylates itself [20]. As shown in Fig 8B, when HA was present, Flag-AkrA can be clearly detected with an anti-Flag antibody. However, a site-directed mutation of the cysteine residue in the DHHC motif and the parental wild-type strain pre-cultured with 2-bromopalmitate (2-BP) completely abolished palmitoylation of AkrA, which resulted in no signal being detected in the enriched pamitoylated proteins. These results indicate that AkrA is able to be auto-acylated and the cysteine residue in the DHHC motif is required for this process. In addition, we found that treatment with 2-BP (24 h, 50 and 100 μM) virtually abolished the Golgi localization of GFP-labelled AkrA (Fig 8D) and resulted in a similar defective growth defect phenotype to the ΔakrA mutant on minimal medium (S10 Fig).
We constructed another alcA(p)::GFP-akrAC487S mutant and confirmed by Western blotting (Fig 8C) to further check whether site directed mutagenesis of the Cys487 in the DHHC motif disrupted the normal localization of AkrA in the Golgi. The GFP-AkrAC487S was less distinctly localized in the punctate Golgi structures characteristic of wild-type GFP-AkrA and some appeared to be localized in the cytoplasm (Fig 8D). These data collectively suggest that the cysteine residue in the DHHC motif of AkrA and the palmitoylation activity are closely associated with AkrA auto-acylation, which is required for normal AkrA localization and palmitoylation.
To further explore palmitoylated protein substrates specifically mediated by AkrA, total proteins of the wild-type and ΔakrA strains were treated and analyzed using the ABE chemistry assay combined with liquid chromatograpy-mass spectrometry (LC-MS) for comparative proteomics (Fig 8E). Using this approach, 334 proteins were identified as potential AkrA substrates in the parental wild-type strain because they were completely absent in the ΔakrA strain. As shown in Table 1, AkrA belonged to one of the AkrA-mediated pamitoylated substrates suggesting it is able to auto-acylate itself. Among the palmitoylated protein candidates identified, Yck2, Lcb1, Ras2, Cdc48 and Pab1 have been previously identified as palmitoylated proteins in S. cerevisiae but only Yck2 has been characterized as an Akr1 substrate [20,55–57]. These data indicated that the ABE chemistry assay combined with LC-MS was a valid approach to identify proteins palmitoylated by AkrA and it also indicated that A. nidulans may palmitoylate some of the substrates previously reported in S. cerevisiae. In our study we notably identified the following protein substrates palmitoylated by AkrA: a vacuolar Ca2+-ATPase Pmc1 homolog (AN5088.4); a P-type ATPase Spf1 homolog (AN3146.4) involved in calcium homeostasis [58]; a putative V-type H+-ATPase Vma5 homolog (AN1195.4) that has been linked to Ca2+-ATPase function [59], and three uncharacterized proteins (AN8774.4, AN3420.4 and AN2427.4), the transcripts of which have previously been shown to be induced by extracellular calcium stress in a CrzA-dependent manner [53]. These results provide strong evidence that the AkrA protein regulates [Ca2+]c homeostasis in A. nidulans by palmitoylating these protein candidates. Other candidate substrates of AkrA that we identified included the P450 enzymes, Cyp51A (Erg11A), Cyp51B (Erg11B) and Erg5 homologs, which are all involved in ergosterol biosynthesis and azole resistance. Thus AkrA may influence the azole resistance by these biosynthetic enzymes.
Palmitoylation is a reversible post-translational modification that is involved in regulating the trafficking and the functional modulation of membrane proteins. Many proteins that rely on palmitoylation are key players in cellular signaling, membrane trafficking and synaptic transmission [19–21]. Yeast Akr1p was the first characterized palmitoyl transferase (PAT) [36,60]. AkrA, a human AkrA homolog HIP14, is involved in palmitoylation and plays an important role in the trafficking of multiple neuronal proteins associated with Huntington’s disease [61]. Calcium serves a multitude of signaling and structural functions in all eukaryotes. Recent studies in mammalian systems have shown that the skeletal muscle ryanodine receptor/Ca2+-release channel RyR1 is subject to S-palmitoylation modification in ''hot spot'' regions containing sites of mutations implicated in malignant hyperthermia and central core disease [62]. However, studies on the relationship between calcium signaling components and palmitoylation are very scarce. In this study, we identified that homologs of the yeast palmitoyl transferase in A. nidulans (AkrA) and A. fumigatus (AfAkrA) are required for hyphal growth and sporulation under low external calcium conditions. High extracellular calcium-, ER- and plasma membrane-stress conditions all elicited transient increases in [Ca2+]c. These [Ca2+]c responses were all mediated by AkrA and involved the cysteine residue in its DHHC motif, which was shown to be required for AkrA palmitoylation. Candidate protein substrates that the AkrA protein is involved in palmitoylating were found to include many key components involved in membrane trafficking and cellular signaling processes including known palmitoylated Ras-like proteins (Table 1). Among them were: a vacuolar Ca2+ ATPase Pmc1 homolog [63]; a putative P-type ATPase Spf1 homolog, which is involved in ER function and calcium homeostasis in budding yeast and Candida albicans [58,64,65]; a Vma5 homolog that has been linked with Pmr1 Ca2+-ATPase function [59], and three calcium signaling- related proteins (encoded by AN8774.4, AN3420.4 and AN2724.4), the transcripts of which have been previously shown to be induced in response to high extracellular calcium stress which is dependent on the transcription factor CrzA [53]. Key P450 enzymes in the ergosterol biosynthesis pathway were also identified as AkrA palmitoylated proteins. Thus, our findings suggest that mutation of the DHHC motif in AkrA results in the disruption of [Ca2+]c homeostasis that is mainly due to the absence of the post-translational, palmitoylated-modification of key proteins involved in calcium signaling/homeostasis. PmcA and SpfA are homologs of two Ca2+ ATPases which response for sequestrating calcium into intercellular compartments in S. cerevisiae [63,64].
Deletion of the akrA gene exhibited marked growth and conidiation defects under low calcium conditions, which is similar to the defects caused by mutations in the CchA/MidA HACS [28–30]. In addition, the akrA deletion conferred increased sensitivity to Li+, Na+, K+, Mg2+, but slightly increased resistance to the cell wall disrupting agents compared to the parental wild-type strain (S5 Fig). Moreover, the ΔakrAΔcchA and ΔakrAΔmidA double mutants exacerbated the ΔakrA defects under calcium-limited conditions, suggesting that AkrA may have independent functions to those of the CchA-MidA complex. AkrA localized to trans Golgi structures (Fig 2C), while the CchA-MidA complex probably localizes to the plasma membrane as reported for yeast [40,66,67]. In addition, results from the Y2H assays (S4A Fig) suggested that there were no direct, or only very weak, interactions between AkrA and CchA and between AkrA and MidA. Nevertheless, the [Ca2+]c transient in the ΔakrA mutant had a much lower amplitude (approximately 53 ± 13% lower) than the wild-type control following treatment with a high extracellular calcium stress stimulus, suggesting that the loss of AkrA reduced calcium influx into the cytoplasm. In contrast, loss of CchA and MidA caused a 25% decrease in the [Ca2+]c amplitude in response to this treatment with high external calcium, consistent with the results from previous studies on yeast cells lacking either Cch1 or Mid1, which exhibited a low calcium uptake [5,39–41]. The akrA deletion also had a bigger impact on inhibiting calcium influx in response to ER stress than observed in the ΔcchA and ΔmidA mutants. Overall our data suggests that AkrA regulates calcium uptake from the external medium as well and its release from intracellular Ca2+ stores through a pathway that is independent of the previously identified CchA/MidA HACS as shown in Fig 9. PmrA is an A. nidulans homolog of yeast Pmr1, which is a P-type Golgi Ca2+/Mn2+ ATPase responsible for Ca2+ transport into the Golgi and widely accepted as responsible for Ca2+ efflux from the cytoplasm into the Golgi to regulate calcium signaling and homeostasis and prevent calcium toxicity. Loss of Pmr1 function in budding yeast is believed to inhibit the return of [Ca2+]c to its resting level following stimulus-induced [Ca2+]c increases [37,45–47]. In contrast, our data showed that the pmrA deletion in A. nidulans exhibited no significant change in the calcium signature following a high extracellular calcium stress stimulus compared with the wild-type strain, suggesting that other paralogs of pmrA (e.g. other Ca2+-ATPases) may compensate or play more important roles in returning the elevated [Ca2+]c back to its resting level. Surprisingly, loss of pmrA alleviated the decreased response of the ΔakrA mutants to the external calcium stimulus, resulting in the amplitude of the [Ca2+]c increase of the double mutant ΔpmrAΔakrA being almost back to the normal level of the wild type. Thus deletion of PmrA reverses the effects of the AkrA deletion in regulating calcium influx following extracellular calcium stress. The lower amplitude of the [Ca2+]c increase of the ΔakrA mutant in response to the high extracellular calcium stimulus indicate that AkrA and its pamitoylated targets play a role in mediating the calcium influx into the cytoplasm and then PmrA may store cytoplasmic calcium into Golgi. When both PmrA and AkrA were absent, the increase in [Ca2+]c following extracellular calcium stimulation was back to almost the normal level in the wild-type (Fig 5). This suggests that the [Ca2+]c increase in the ΔpmrAΔakrA double mutant following treatment with high extracellular calcium is compensated by some other unknown component(s) of the calcium signaling/homeostatic machinery. Furthermore, our data (Fig 4A) showed that loss of pmrA suppressed the colony growth defect of ΔakrA mutants, providing further evidence to support interactive regulatory roles of PmrA and AkrA in A.nidulans.
Previous studies have verified that exposure of fungi to ER or plasma membrane stress stimulates store-operated calcium influx through the HACS to promote fungal cell survival [13,14,41,50–52]. Consistent with previous studies, in A. nidulans we observed a transient increase in [Ca2+]c after treatment with the ER-stress agents tunicamycin (TM) or dithiothreitol (DTT). The ΔcchA mutant exhibited reduced [Ca2+]c amplitudes by 32 ± 6% and 15 ± 9% upon treatment with TM or DTT, respectively (Figs 6 and S7). In contrast, we did not detect a change in the [Ca2+]c response to the ER stress agents in the ΔmidA mutant compared to its parental wild-type strain. This suggests that as a complex of CchA and MidA, CchA may have a more predominant role than MidA during the ER stress response. Moreover, the ΔakrA mutant displayed a decreased response to ER and plasma membrane stress inducing drugs, as the [Ca2+]c amplitude of ΔakrA mutants decreased by approximately 36–40% of the wild-type strain following treatment with these drugs (Figs 6 and S7). These data suggest that, in addition to HACS components, AkrA is also involved in ER and plasma membrane stress-induced calcium influx. Moreover, these responses were completely abolished in the ΔakrA mutant but not in the wild-type strain in the presence of EGTA or BAPTA that chelate external calcium. These results indicate that both extracellular calcium and calcium stores contribute to the transient [Ca2+]c changes following ER or plasma membrane stress. Because calcium release from intracellular stores in response to these types of stress was abolished in the akrA mutants (Figs 6, 7 and S9), our results are consistent with AkrA regulating calcium influx across the plasma membrane, which in turn activates the release of calcium from intracellular pools.
Altogether, our results provide the first report that AkrA is a putative palmitoyl transferase in A. nidulans, and it mediates calcium influx in a DHHC-dependent mechanism to perform an essential function in calcium homeostasis/signaling for survival under high extracellular calcium-, ER- or azole antifungal-stress conditions.
Calcium signaling regulators have been previously identified as antifungal target candidates, including FK506, which targets calcineurin [8]. However, most of the fungal homologs of known calcium signaling components in mammalian cells are of proteins also required for mammalian cell growth and metabolism [68]. Thus, potential antifungals against these components may cause side effects in mammalian hosts. The use of drugs that target regulators of posttranslational modification of calcium signaling that show significant differences to their mammalian homologs (e.g. AkrA only exhibits 24.8% identity to the human AkrA homolog HIP14), may circumvent this problem. The potential for developing novel antifungal drugs of this type has been greatly facilitated by our study that has shown a critical link between palmitoylation and calcium signaling.
Previous studies have shown that all AkrA homologs across different species require the DHHC motif to be active and function normally as palmitoyl transferases [69–71]. Three approaches were initially employed to determine AkrA function: deletion of the DHHC motif; site-directed mutagenesis of the cysteine residue in the DHHC motif; and use of a specific palmitoyl transferase analogue inhibitor (2-bromopalmitate), to determine AkrA function [48,49]. Our data from these experiments suggested that the DHHC motif and its cysteine residue are required for the function of AkrA, especially when extracellular calcium is limited. To further test whether the cysteine residue in the DHHC motif, is correspondingly required for AkrA palmitoylation, we used the acyl-biotin exchange (ABE) chemistry assay to detect palmitoylation based on selective thioester hydrolysis by hydroxylamine. Compared to the treatment without hydroxylamine, the newly exposed cysteine residues are disulfide-bonded to a biotin analogue, affinity purified and digested into peptides, leaving the labeled peptides on the affinity beads so that palmitoylated proteins have been enriched. As the ABE chemistry detects palmitoylation through identification of all the thioester linkages. A subsequent Western experiment was used to further confirm palmitoylated proteins by specific antibodies. Consequently, among these enriched palmitoylated proteins, Flag-AkrA was clearly detected with an anti-Flag antibody. Site-directed mutation of the cysteine residue in the DHHC or treatment of the parental wild-type strain with the palmitoyl transferase analogue inhibitor 2-BP completely abolished palmitoylation of AkrA (Fig 8B). Previous studies have demonstrated that although the exact mechanism of S-acylation is not known, palmitoylation of the purified DHHC-CRD palmitoylated proteins zDHHC2, zDHHC3 and yeast Erf2, involves a two-step mechanism, in which the zDHHCs form an acyl-enzyme intermediate (auto-acylation), with the acyl group later transferred to the target protein [53,54]. Our results indicated that AkrA auto-acylated itself before palmitoylating its target proteins. In mammalian cells, any protein that contains a surface-exposed and freely accessible cysteine that has transient access to Golgi membranes is susceptible to palmitoylation. Our data suggests AkrA both auto-acylated itself and palmitoylates target proteins in association with Golgi membranes. Moreover, we found that site directed mutagenesis of the Cys487 in the DHHC motif significantly affect normal localization of AkrA in the Golgi. When we treated cells with a specific palmitoyl transferase analogue inhibitor 2-BP, AkrA localization within the Golgi localization was completely lost (Fig 8D), suggesting that the 2-BP treatment not only prevented AkrA auto-acyltation but also prevented the normal subcellular localization of AkrA. The reason for the different localization pattern, if any, caused by the site directed mutagenesis and the treatment of 2-BP as shown in Fig 8D is likely to be due to a side effect of the 2-BP reagent.
In conclusion, our results provide the first report that AkrA is a palmitoyl transferase in A. nidulans, and that it mediates calcium influx in a DHHC-dependent mechanism to perform an essential role in calcium homeostasis to survive high extracellular calcium-, ER- and plasma membrane-stress conditions. A working model of AkrA function in regulating [Ca2+]c homeostasis in A. nidulans is presented in Fig 9. Our findings provide new insights into the link between palmitoylation and calcium signaling that may be of relevance for understanding the mechanistic basis of human PAT-related diseases. Regulators of posttranslational modification in fungi may provide promising targets for new therapies against life threatening fungal diseases.
All fungal strains used in this study are listed in S1 Table. Minimal media (MM), and MMPDR (minimal media + glucose + pyrodoxine + riboflavin), MMPDR+UU (minimal media + glucose + pyrodoxine + riboflavin+ uridine + uracil), MMPGR (minimal media + glycerol + pyrodoxine + riboflavin) have been described previously [29,72]. MMPGRT was MMPGR with 100 mM threonine. Fungal strains were grown on minimal media at 37°C, harvested using sterile H2O and stored for the long-term in 50% glycerol at −80°C. Expression of tagged genes under the control of the alcA promoter was regulated by different carbon sources: non-induced by glucose, induced by glycerol and overexpressed by glycerol with threonine. Growth conditions, crosses and induction conditions for alcA(p)-driven expression were as previously described [73].
In order to generate constructs for akrA null mutant (ΔakrA), the fusion PCR method was used as previously described [74]. Primers used to design constructs are listed in S2 Table. The A. fumigatus pyrG gene in plasmid pXDRFP4 was used as a selectable nutritional marker for fungal transformation. The transformation was performed as previously described [75].
For creating an ΔakrA construct, a 5′ flank and a 3′ flank DNA fragments were amplified using the primers akrA-P1 and akrA-P3, akrA-P4 and akrA-P6, respectively, using genomic DNA (gDNA) of the A. nidulans wild-type strain TN02A7 as the template for PCR. As a selectable marker, a 2.8 kb DNA fragment of A. fumigatus pyrG was amplified from the plasmid pXDRFP4 using the primers pyrG-5’ and pyrG-3’. The three PCR products were combined and used as a template to generate a 4.8 kb DNA fragment using the primers akrA-P2 and akrA-P5. The final PCR product was transformed into a wild-type strain. A similar strategy was used to construct akrA-truncated mutants.
To design the revertant strain construct, a 3.7 kb DNA fragment, which included a 1.2 kb promoter region, a 2.4 kb coding sequence, and a 3′ flank was amplified using the primers primer A and primer D from A. nidulans gDNA. As a selectable marker, a 1.7 kb pyroA fragment was amplified from the plasmid pQa-pyroA using the primers pyro-5’ and pyro-3’. The two PCR products were co-transformed into the ΔakrA strain to produce the revertant strain.
To generate the alcA(p)::GFP-akrA vector, a 1 kb akrA fragment was amplified from the gDNA in the wild-type strain TN02A7 with primers akrA-5’ and akrA-3’ (S2 Table) and then ligated into the plasmid vector pLB01 yielding plasmid pLB-alcA(p)::GFP-akrA which contains GFP-N under the control of the alcA promoter with the N. crassa pyr4 as a marker.
For site-directed mutation, a 3.7 kb akrA DNA fragment with a site directed mutation in which cysteine487 was replaced by serine and a selective marker pyroA were co-transformed into the ΔakrA strain to obtain native(p)::akrAC487S strain. The fragment containing the site mutation was amplified with two steps. First, fragment AB and fragment CD were amplified from A. nidulans gDNA with primers A and B, primers C and D, respectively, and complementary regions contained the desired mutation (cysteine487 to serine487). Second, using fragment AB and fragment CD as a template, the final 3.7 kb fragment was generated through fusion PCR using primer A and primer D.
The GPD(p)::akrAC487S and alcA(p)::GFP-akrAC487S strains were constructed using a similar strategy. In brief, the GPD promoter was amplified with the GPD-5’ and GPD-3’, and 2.4 kb akrA DNA fragment including a 2.4 kb coding sequence, and a 0.5 kb 3’ flanking was amplified with akrA-GPD-5’ and primer D. These two fragments were combined using GPD-5’ and primer D, Lastly, the aboved fusion PCR products and the selective marker pyroA were co-transformed into the ΔakrA strain to obtain the GPD(p)::akrAC487S strian. For the alcA(p)::GFP-akrAC487S construction, a 5′ flank and a 3′ flank DNA fragments were amplified from genomic DNA of alc-akrA mutant using the primers alc-up and primer B, primer C and new primer D, respectively. Then the two PCR products were combined and used as a template to generate a 3.9 kb DNA fragment using the primers alc-up and new primer D, and then this fragment was ligated into a plasmid vector yielding the pEA-C487S. The pyroA fragment was amplified from the pQa-pyroA using the primers pyro-cre-5’ and pyro-cre-3’, then recombined into the plasmid pEA-C487S. Finally the plasmid was transformed into the ΔakrA strain to obtain the alcA(p)::akrAC487S strian.
All N-terminal Flag constructs were designed and fabricated using restriction-free cloning protocols outlined at http://www.rf-cloning.com using PrimerSTAR MAX DNA polymerase (TAKARA, R045A) [76]. Then, N-Flag tagged cassettes and selective marker pyroA were co-transformed into the ΔakrA strain.
For the mutants expressing the codon-optimized aequorin, the plasmid pAEQS1-15 containing codon-optimized aequorin and selective markers pyroA or riboB genes were co-transformed into the indicated mutants. Transformants were screened for aequorin expression using methods described previously [77] and high aequorin expressing strains were selected after homokaryon purification involving repeated plating of single conidia.
For each experiment, at least three replicate plates were used to test phenotypes for each strain. To assess the influence by the extracellular calcium to the colony phenotype, minimal medium was supplemented with 20 mM CaCl2 or 1 mM EGTA, respectively. The influence of osmotic stress or ionic stress was tested by adding 600 mM NaCl, 600 mM KCl, 10 mM MnCl2, 400 mM MgCl2, 400 mM CaCl2 or 300 mM LiCl into minimal medium, respectively. For the cell wall integrity test, the reagent of 60 μg/mL Calcofluor White or 100 μg/mL Congo Red was added to the minimal medium, respectively. 2 μL of conidia from the stock (1×106 conidia/mL) for indicated strains were spotted onto relevant media and cultured for 2.5 days, at 37°C, and then the colonies were observed and imaged.
For microscopic observations, conidia were inoculated onto pre-cleaned glass coverslips overlaid with liquid media. To observe co-localization of GFP-AkrA and mRFP-PHOSBP, strain ZYA13 (S1 Table) was cultured at 37°C for 10 h in non-inducing medium (non-inducing conditions for the alcA(p) driving expression of AkrA) and shifted for 6 h to the inducing medium (in which the alcA promoter was induced) before microscopic observation [34]. Differential interference contrast (DIC) and fluorescence images of the cells were captured with a Zeiss Axio imager A1 microscope (Zeiss, Jena, Germany) equipped with a Sensicam QE cooled digital camera system (Cooke Corporation, Germany). The images were processed with MetaMorph/MetaFluor software (Universal Imaging, West Chester, PA) and assembled in Adobe Photoshop (Adobe, San Jose, CA).
Germination was assessed in liquid non-inducing medium at 37°C with a total number of 106 conidia/mL for each strain in their stationary phase [78]. The percentage rate of germination was measured at 4, 5, 6, 7 and 8 h by microscopic examination. Spores were considered as germinated ones when length of the germ tube was almost equal to the conidium in diameter. For each strain, three replicates of 100 spores were quantified at each time point to determine the germination rate.
Saccharomyces cerevisiae strain AH109 (Clontech, Palo Alto, CA) was used as the host for the two-hybrid interaction experiments. The analysis was performed using the Matchmaker Library Construction & Screening system (BD Clontech). For strain generation, a cDNA fragment corresponding to the cytosol C-terminus of cchA and the full-length cDNA of midA were amplified and cloned into the pGADT7 vector, which contains the GAL4 DNA-AD and the LEU2 marker (BD Clontech). Full-length cDNA of akrA were used for the pGBKT7 vector (Clotech, Palo Alto, CA).
The strains expressing the codon-optimized aequorin gene were grown on minimal media for 2.5 days to achieve maximal conidiation. 106 spores with liquid media were distributed to each well of a 96-well microtiter plate (Thermo Fischer, United Kingdom). Six wells were used in parallel for each treatment. The plates were incubated at 37°C for 18 h. The medium was then removed and the cells in each well were washed twice with PGM (20 mM PIPES pH 6.7, 50 mM glucose, 1 mM MgCl2). Aequorin was reconstituted by incubating mycelia in 100 μL PGM containing 2.5 μM coelenterazine f (Sigma-Aldrich) for 4 h, at 4°C in the dark. After aequorin consititution, mycelia were washed twice with 1 mL PGM and allowed to recover to room temperature for 1 h [79,80]. To chelate extracellular Ca2+, 1 mM EGTA or 8 mM BAPTA was added to each well 10 min prior to stimulus injection.
At the end of each experiment, the active aequorin was completely discharged by permeabilizing the cells with 20% (vol/vol) ethanol in the presence of an excess of calcium (3 M CaCl2) to determine the total aequorin luminescence of each culture. Luminescence was measured with an LB 96P Microlumat Luminometer (Berthold Technologies, Germany), which was controlled by a dedicated computer running the Microsoft Windows-based Berthold WinGlow software. Conversion of luminescence (relative light units [RLU]) into [Ca2+]c was done using Excel 2007 software (Microsoft). The relative light units (RLU) values were converted into [Ca2+]c concentrations by using the following empirically derived calibration formula: pCa = 0.332588 (-log k) + 5.5593, where k is luminescence (in RLU) s-1/total luminescence (in RLU) [77]. Error bars represent the standard error of the mean of six independent experiments, and percentages in the figures represent peak of [Ca2+]c compared to that of the wild-type (100%).
ABE was performed as described previously with some modifications [81]. Briefly, the strain mycelium was ground to a fine powder in liquid nitrogen and resuspended in 5 mL lysis buffer. Samples were incubated for 1 h at 4°C followed by centrifugation at 4°C, 13,000 g to remove insoluble material. 5 mg of protein was incubated overnight with 50 mM N-ethylmaleimide (NEM) at 4°C to reduce proteolysis while allowing free sulhydryls to be blocked. Proteins were precipitated at room temperature using methanol/chloroform. The pellet was resuspended in 200 μL resuspension buffer and the solution divided into two equal aliquots. One aliquot was combined with 800 μL of 1 M fresh hydroxylamine (HA), 1 mM EDTA, protease inhibitors and 100 μL 4 mM biotin-HPDP (Thermo Scientific). As a control the remaining aliquot was treated identically but hydroxylamine (HA) was replaced with 50 mM Tris pH 7.4. Proteins were precipitated and resuspended in 100 μL of resuspension buffer. 900 μL PBS containing 0.2% Triton X-100 was added to each sample, aliquots were removed as a loading control, and the remaining reactions were incubated with 30 μL of streptavidin-agarose beads (Thermo scientific). The streptavidin beads were washed four times with 1 mL PBS containing 0.5 M NaCl and 0.1% SDS. Proteins were eluted by heating at 95°C in 40 μL 2× SDS sample buffer containing 1% 2-mercaptoethanol v/v. Samples were analyzed by silver staining or Western blotting as described below. In some cases, cells were treated with 50 or 100 μM of the palmitoylation inhibitor 2-bromopalmitate (2-BP) before the ABE assay.
For mass spectrometry (MS), total protein (100 μg) extracted from each sample was chemically reduced for 1 h at 60°C by adding DTT to 10 mM and carboxyamidomethylated in 55 mM iodoacetamide for 45 min at room temperature in the dark. Then trypsin gold (Promega, Madison, WI, USA) was added to give a final substrate/enzyme ratio of 30:1 (w/w). The trypsin digest was incubated at 37°C for 16 h. After digestion, the peptide mixture was acidified by 10 μL of formic acid for further MS analysis. After protein digestion, each peptide sample was desalted using a Strata X column (Phenomenex), vacuum-dried and then resuspended in a 200 μL volume of buffer A (2% ACN, 0.1% FA). After centrifugation at 20000 g for 10 min, the supernatant was recovered to obtain a peptide solution with a final concentration of approximately 0.5 μg/μL. 10 μL supernatant was loaded on a LC-20AD nano-HPLC (Shimadzu, Kyoto, Japan) by the autosampler onto a 2 cm C18 trap column. The peptides were then eluted onto a 10 cm analytical C18 column (inner diameter 75 μm) packed in-house. The samples were loaded at 8 μL/min for 4 min, then the 35 min gradient was run at 300 nL/min starting from 2 to 35% buffer B (95% ACN, 0.1% FA), followed by a 5 min linear gradient to 60%, then followed by a 2 min linear gradient to 80%, and maintenance at 80% buffer B for 4 min, and finally returned to 5% in 1 min.
Data acquisition was performed with a TripleTOF 5600 System (AB SCIEX, Concord, ON) fitted with a Nanospray III source (AB SCIEX, Concord, ON) and a pulled quartz tip as the emitter (New Objectives, Woburn, MA). Data was acquired using an ion spray voltage of 2.5 kV, curtain gas of 30 psi, nebulizer gas of 15 psi, and an interface heater temperature of 150. The MS was operated with a RP of greater than or equal to 30,000 FWHM for TOF MS scans. Raw data files acquired from the Orbitrap were converted into MGF files using Proteome Discoverer 1.2 (PD 1.2, Thermo), [5,600 msconverter] and the MGF file were searched. Protein identification was performed by using Mascot search engine (Matrix Science, London, UK; version 2.3.02) against a database containing 13,597 sequences.
To extract proteins from A. nidulans mycelia, conidia from alcA(p)::GFP-akrA and the wild-type strains were inoculated in the liquid inducing medium, then shaken at 220 rpm on a rotary shaker at 37°C for 24 h. The mycelium was ground in liquid nitrogen with a mortar and pestle and suspended in ice-cold extraction buffer (50 mM HEPES pH 7.4, 137 mM KCl, 10% glycerol containing, 1 mM EDTA, 1 μg/mL pepstatin A, 1 μg/mL leupeptin, 1 mM PMSF). Equal amounts of protein (40 μg) per lane were subjected to 10% SDS–PAGE, transferred to PVDF membrane (Immobilon-P, Millipore) in 384 mM glycine, 50 mM Tris (pH 8.4), 20% methanol at 250 mA for 1.5 h, and the membrane was then blocked with PBS, 5% milk, 0.1% Tween 20. Next, the membrane was then probed sequentially with 1:3000 dilutions of the primary antibodies anti-GFP or anti-FLAG or anti-actin and goat anti-rabbit IgG-horseradish peroxidase diluted in PBS, 5% milk, 0.1% Tween 20. Blots were developed using the Clarity ECL Western blotting detection reagents (Bio-Rad), and images were acquired with the Tanon 4200 Chemiluminescent Imaging System (Tanon).
The mycelia were cultured for 18 h in liquid media and were then ground to a fine powder in liquid nitrogen. Total RNA was isolated using Trizol (Invitrogen, 15596–025) following the manufacturer’s instructions. 100 mg of mycelia per sample was used as the starting material for the determination of total RNA. The reverse transcription polymerase chain reaction (RT-PCR) was carried out using HiScript Q RT SuperMix (Vazyme, R123-01), and then cDNA was used for the real-time analysis. For real-time reverse transcription quantitative PCR (RT-qPCR), independent assays were performed using SYBR Premix Ex Taq (TaKaRa, DRR041A) with three biological replicates, and expression levels normalized to the mRNA level of actin. The 2-ΔCT method was used to determine the change in expression.
|
10.1371/journal.ppat.1001054 | A Subset of Replication Proteins Enhances Origin Recognition and Lytic Replication by the Epstein-Barr Virus ZEBRA Protein | ZEBRA is a site-specific DNA binding protein that functions as a transcriptional activator and as an origin binding protein. Both activities require that ZEBRA recognizes DNA motifs that are scattered along the viral genome. The mechanism by which ZEBRA discriminates between the origin of lytic replication and promoters of EBV early genes is not well understood. We explored the hypothesis that activation of replication requires stronger association between ZEBRA and DNA than does transcription. A ZEBRA mutant, Z(S173A), at a phosphorylation site and three point mutants in the DNA recognition domain of ZEBRA, namely Z(Y180E), Z(R187K) and Z(K188A), were similarly deficient at activating lytic DNA replication and expression of late gene expression but were competent to activate transcription of viral early lytic genes. These mutants all exhibited reduced capacity to interact with DNA as assessed by EMSA, ChIP and an in vivo biotinylated DNA pull-down assay. Over-expression of three virally encoded replication proteins, namely the primase (BSLF1), the single-stranded DNA-binding protein (BALF2) and the DNA polymerase processivity factor (BMRF1), partially rescued the replication defect in these mutants and enhanced ZEBRA's interaction with oriLyt. The findings demonstrate a functional role of replication proteins in stabilizing the association of ZEBRA with viral DNA. Enhanced binding of ZEBRA to oriLyt is crucial for lytic viral DNA replication.
| Epstein-Barr virus encodes a protein, ZEBRA, which plays an essential role in the switch between viral latency and the viral lytic cycle. ZEBRA activates transcription of early viral genes and also promotes lytic viral DNA replication. It is not understood how these two functions are discriminated. We studied five ZEBRA mutants that are impaired in activation of replication but are wild-type in the capacity to induce transcription of early viral genes. We demonstrate that these five mutants are impaired in binding to viral DNA regulatory sites. Therefore, replication required stronger interactions between ZEBRA and viral DNA than did transcription. Three components of the EBV-encoded replication machinery, including the single-stranded DNA binding protein, the polymerase processivity factor and the primase markedly enhanced the interaction of ZEBRA with viral DNA. These three components partially rescued the defect in ZEBRA mutants that were impaired in replication. The results suggest that through protein-protein interaction, replication proteins play a role in enhancing ZEBRA's association with the origin of DNA replication and other regulatory sites.
| There are many gaps in our understanding of the process by which the Epstein-Barr virus (EBV) lytic replication machinery assemble on DNA sites present in the viral genome.
EBV encodes an essential bZIP protein known as ZEBRA (aka Zta, Z and BZLF1) that functions as a transcription activator of viral and cellular genes and as an origin binding protein during lytic DNA replication. An EB viral genome that lacks the open reading frame encoding ZEBRA, bzlf1, loses its ability to activate lytic gene expression and DNA replication [1]. ZEBRA interacts both with promoters and with origins of lytic replication through DNA sequences known as ZEBRA response elements (ZREs) that are common to both types of DNA regulatory regions [2], [3], [4]. It is unknown how ZEBRA distinguishes between a replication site and a transcription activation site. The mechanism by which ZEBRA activates transcription relies on its capacity to bind DNA and to form physical contact with a number of cellular proteins. ZEBRA binds to a wide variety of ZREs located in target promoters. Some of these response elements contain methylated CpG motifs to which ZEBRA binds with high preference [5]. The protein also forms stable transcriptional initiation complexes with basic components of the transcription machinery such as TBP, TFIID, and the transcription co-activator CBP [6], [7], [8]. Since ZEBRA augments the histone acetyl transferase (HAT) activity of CBP, interaction of ZEBRA with CBP increases promoter accessibility [9].
Activation of viral DNA synthesis during the lytic phase of the EBV life cycle is dependent on the capacity of ZEBRA to efficiently recognize a large (∼1 kb) complex intergenic region that serves as the origin of replication. This region, known as oriLyt, consists of essential and auxiliary segments [10]. The two essential components of oriLyt, the upstream and downstream elements, together constitute the minimal origin of DNA replication [2], [11], [12]. The auxiliary component serves as an enhancer element that augments DNA replication [13], [14]. ZEBRA recognizes the origin of lytic DNA replication (oriLyt) by interacting with seven ZEBRA-binding sites [12], [15]. Mutation of all seven binding motifs in the background of a recombinant virus drastically reduces production of infectious virus particles [16]. These ZEBRA binding elements are located in two non-contiguous regions of oriLyt. Four elements are present in the upstream core region of oriLyt and overlap with the promoter of the BHLF1 open reading frame [3]. Knocking out any of these four elements was deleterious for amplification of an oriLyt-containing plasmid in a transient replication assay [17]. Three additional ZEBRA binding elements located mainly in the enhancer region are dispensable for viral replication [17].
The current model for the role of ZEBRA in lytic DNA replication suggests that the protein serves as a physical link between oriLyt and core components of the replication machinery [18], [19]. The six core replication factors encoded by EBV are the DNA polymerase (BALF5); the polymerase processivity factor (BMRF1); the helicase (BBLF4); the primase (BSLF1); the primase associated factor (BBLF2/3), and the single-stranded DNA binding protein (BALF2) [4]. Corroboration for the proposed role of ZEBRA in replication is inferred from data showing that ZEBRA interacts with almost all components of the viral replication machinery, with the exception of the single-stranded DNA binding protein (BALF2) [18], [20], [21], [22]. The function of tethering replication proteins to oriLyt is not limited to ZEBRA; the transactivation domains of Sp1 and ZBP89 interact with BMRF1 and BALF5 and target them to the downstream region of oriLyt [18], [23]. Similarly, ZBRK1, a cellular DNA binding zinc finger protein, serves as a contact point for BBLF2/3 on oriLyt [19]. Deletion of the ZBRK1 binding site present in the downstream region of oriLyt reduced oriLyt-dependent replication of a transiently transfected plasmid. Binding of these cellular transcription factors is not essential but contributes to replication efficiency.
ZEBRA mutants that activate transcription but not replication are valuable in furthering our understanding of the process of EBV lytic DNA replication. ZEBRA is phosphorylated in vivo at multiple sites [24]. Phosphorylation of ZEBRA at S173 regulates lytic viral replication [25]. Serine 173 is located in a region N-terminal to the DNA binding domain of ZEBRA. This region, known as the regulatory domain, regulates the DNA binding activity of the protein [25], [26], [27]. Alanine substitution of the phosphoacceptor site S173 reduced the capacity of ZEBRA to bind to DNA in vitro and in vivo [25]. Attenuation in DNA binding correlated with a defect in the capacity of ZEBRA to stimulate lytic viral replication. However, it had no effect on the ability of ZEBRA to activate transcription of downstream viral target genes. Thus phosphorylation of S173 segregates the two main functions of ZEBRA, namely activation of transcription and activation of viral replication. In addition, the S173A mutant demonstrates that activation of transcription is not sufficient to stimulate viral replication. Additional proof for the role of phosphorylation of S173 in replication was attained when a phosphomimetic substitution mutant Z(S173D) activated both transcription and replication and was competent to bind DNA to the same extent as wild-type (wt) ZEBRA. Therefore, phosphorylation of ZEBRA at S173 functionally mimics ATP binding in other origin binding proteins by enhancing the DNA binding activity of ZEBRA to all ZREs in general and not to a specific site [25].
In a comprehensive mutagenesis study of the DNA binding domain of ZEBRA we identified ZEBRA mutants that arrested the EBV lytic cycle at different stages [28]. Two of these mutants, Z(Y180E) and Z(K188A), caused lytic cycle arrest prior to viral replication. They reproducibly activated expression of viral early genes but were defective in inducing amplification of EBV DNA and late gene expression [28]. These mutants did not affect the phosphorylation site in the regulatory domain, S173, but changed specific residues within the DNA recognition domain. The availability of replication defective (RD) ZEBRA mutants prompted us to investigate the effect of alterations in the DNA binding activity of ZEBRA on viral replication. If replication is indeed less tolerant than transcription for weak interaction between ZEBRA and DNA, then stronger association with oriLyt is necessary and might play a critical role in origin activation. Augmentation of ZEBRA binding to oriLyt is likely to be mediated by factors specific for viral replication. For example in budding yeast, interaction of the ORC with Cdc6 enhances its interaction with the origin of replication [29]. Here we describe a new role for three components of the EBV replication complex, namely, the primase, the single-stranded DNA binding protein and the DNA processivity factor. We show that over-expression of these three replication proteins is sufficient to increase the association of ZEBRA with viral DNA. This augmentation in DNA binding suppressed the phenotype of ZEBRA replication defective mutants and partially restored viral genome amplification and late gene expression. Our findings represent the first indication that three replication proteins play a role in enhancing the interaction between ZEBRA and viral DNA thereby promoting origin recognition, a process that is exquisitely sensitive to the DNA binding activity of ZEBRA.
Previously we described three ZEBRA mutants which activated expression of early genes but failed to activate viral replication and late gene expression. The ZEBRA mutants that reproducibly exhibited replication defective phenotype were: Z(S173A) in the regulatory domain and Z(Y180E) and Z(K188A) in the DNA recognition domain [25], [28]. In further exploration of this phenomenon we identified a fourth ZEBRA RD mutant with a conservative arginine to lysine substitution at position 187. Fig. 1A, C and D compare the phenotype of Z(R187K) to wt ZEBRA, Z(K188A) and Z(F193E). Z(K188A) served as a typical ZEBRA RD mutant; the mutant Z(F193E) was partially defective in induction of late genes and DNA replication. Expression of Z(R187K) in BZKO cells induced a pattern of lytic gene expression that mimicked Z(K188A); it fully activated expression of two early proteins, Rta and EA-D (aka BMRF1), encoded by brlf1 and bmrf1, but failed to activate synthesis of two late proteins BFRF3 (FR3) (a component of the viral capsid) and BLRF2 (LR2) (a tegument protein) (Fig. 1A). Rta and EA-D are direct targets of ZEBRA; their expression is governed by the ability of ZEBRA to bind to their corresponding promoters, Rp and BMRF1p, respectively [30], [31], [32]. Activation of expression of the two late proteins, FR3 and LR2, is associated with the capacity of ZEBRA to induce lytic viral replication [33].
To demonstrate that the introduced point mutations were the sole cause for the observed defect in late gene expression, the mutant Z(Y180E) was reverted to its original amino acid composition, i.e. tyrosine. As expected, Z(Y180E) was impaired in activating late gene expression while the revertant mutant Z(Y180E→Y) was competent to activate late gene expression to the same level as wt ZEBRA (Fig. 1B).
To examine whether the defect in late gene expression was due to a failure in stimulating viral replication, we tested the capacity of Z(R187K) to induce viral genome amplification by probing for two different regions of viral DNA. First, we probed for a region upstream of the viral terminal repeats (TRs). During lytic viral replication linear viral genomes are synthesized. These linear forms differ in their number of terminal repeats and are detected on a Southern blot as a ladder [34]. In Fig. 1C, wt ZEBRA induced the formation of a replication ladder. Z(F193E) was slightly impaired and resulted in a less intense ladder than wt ZEBRA. The two late mutants Z(R187K) and Z(K188A) failed to induce the replication ladder. Comparable results were observed when a Southern blot of a parallel experiment was probed for the reiterated BamH1 W sub-fragment of EBV DNA (Fig. 1D). All the RD mutants were defective at amplifying viral DNA when assessed by qPCR (Fig. S3). Based on these results and our previous studies, we conclude that RD mutants Z(S173A), Z(Y180E), Z(R187K), and Z(K188A) are competent to activate expression of early viral proteins but incompetent to activate lytic viral DNA replication and late gene expression (see also Fig. S3).
Although the data in Fig. 1 showed that the four replication defective mutants activated expression of two early proteins, Rta and EA-D to the same level as wt ZEBRA, this result did not directly assess the capacity of the mutants to activate transcription from early promoters. The level of brlf1 transcripts is particularly important since activation of the brlf1 promoter by direct binding of ZEBRA is a crucial initial event in activation of the EBV lytic cycle [5], [30], [31], [32], [35]. Therefore, using quantitative RT-PCR we measured the level of endogenous brlf1 mRNA, encoding Rta, in BZKO cells expressing each of the four ZEBRA RD mutants. We found that wt ZEBRA induced expression of the brlf1 message by 776-fold relative to the background level of brlf1 mRNA detected in cells transfected with empty vector (Fig. 2A). The level of brlf1 expression was corrected for the corresponding level of the gapdh transcript measured in each sample (Fig. 2B). The ZEBRA RD mutants, Z(S173A), Z(Y180E), Z(R187K) and Z(K188A), reproducibly activated expression of the brlf1 message to levels similar or higher than that of wt ZEBRA (Fig. 2, S4A and S5). Therefore, despite a clear defect in the capacity of these ZEBRA mutants to activate viral replication, the mutants were fully competent to activate transcription of the early brlf1 gene.
In addition to its role in replication, expression of ZEBRA leads to activation of transcription of early lytic cycle genes, six of which constitute core components of the viral lytic replication machinery [4], [36]. The defect observed with the ZEBRA RD mutants could be attributed to failure to activate transcription of one or more genes encoding essential replication proteins. To investigate this possibility, we examined the capacity of the ZEBRA mutants to activate transcription of the different components of the viral replication machinery. Expression of balf2, the gene encoding the single-stranded DNA binding protein was examined by expressing five ZEBRA mutants in BZKO cells. Three of these mutants, Z(Y180E), Z(K188A) and Z(R187K), are markedly defective in activating late gene expression and viral replication, Fig. 1, Fig. S3 and [28]. The other two mutants, Z(F193E) and Z(K194A), are slightly to moderately impaired in activating viral replication and late gene expression (Fig. 1, [28] and unpublished data). 48 h after transfection of BZKO cells, we compared the level of balf2 expression among the mutants using Northern blot analysis. All five mutants activated the balf2 message to a level equivalent to wt ZEBRA (Fig. 3A). As a positive control for migration of the balf2 transcript we used RNA from HH514-16 cells induced into the lytic cycle with sodium butyrate.
Using quantitative RT-PCR we assessed the level of transcripts encoding the heterotrimeric helicase-primase complex in cells expressing five RD mutants: the regulatory mutants, Z(S173A) and Z(S167A/S173A) and the three basic domain mutants, Z(Y180E), Z(R187K) and Z(K188A). We employed two different methods to prepare cDNA from purified RNA samples. In the experiment illustrated in Fig. 3B and 3C, we synthesized cDNA using gene specific primers that were complementary to viral helicase (BBLF4) or viral primase (BSLF1). In Fig. 3D, 3E and 3F, we used a mixture of random hexamers and poly-dT to synthesize cDNA. It is important to note that each of the DNA fragments amplified by RT PCR acquired the same melting point and electrophoretic mobility on agarose gels as DNA fragments amplified by PCR from an expression vector containing a cloned version of the corresponding gene (data not shown). To confirm that the purified RNA samples were not contaminated with genomic DNA we omitted the reverse transcriptase enzyme from the reaction mixture. As a result no DNA amplification was detected (Fig. 3B).
Regardless of the method used for cDNA preparation, we found that the levels of mRNAs for viral helicase, primase and primase-associated factor (BBLF2/3) in cells expressing wt ZEBRA were several fold higher than in cells transfected with empty vector. All four RD mutants were competent to activate expression of the viral helicase and primase to levels comparable or higher than those activated by wt ZEBRA. The mutants, particularly the basic domain mutants, activated twice as much helicase and primase transcripts as the wild type protein. For example, Z(K188A) activated between 2.3 to 2.6-fold more bblf4 mRNA than wt ZEBRA (Fig. 3B and 3D).
The primase-associated-factor (BBLF2/3) was the only gene that exhibited lower transcript levels in cells expressing RD mutants compared to those expressing wt ZEBRA (Fig. 3F). However, the level of bblf2/3 mRNA was still 5–9-fold above background. To determine whether ectopic expression of BBLF2/3 could rescue the defect in these mutants, we co-expressed BBLF2/3 with two ZEBRA RD mutants, Z(S173A) and Z(Y180E), in BZKO cells. Forty-eight hours after transfection, cells were harvested and analyzed for late gene expression and viral replication. We found that over-expression of BBLF2/3 had no effect on the level of the late protein, FR3, induced by wt ZEBRA, Z(S173A) or Z(Y180E) (Fig. 4A). Similarly, using quantitative PCR to determine the extent of viral genome amplification, we found the same levels of viral genome in cells expressing Z(S173A) or Z(Y180E) in the absence or presence of BBLF2/3. The level of viral DNA present in cells transfected with the mutants was approximately equal to that in control cells transfected with empty vector (Fig. 4B). These experiments showed that impairment of ZEBRA RD mutants to induce late gene expression and viral replication was not the result of the slightly reduced levels of the bblf2/3 transcript detected following expression of this class of ZEBRA mutants. Moreover, over-expression of BBLF2/3 protein could not rescue the late mutants.
Previously, we showed that reduction in the DNA binding activity of ZEBRA, due to alanine substitution of the phosphorylation site S173, correlated with a defect in the capacity of ZEBRA to induce viral replication. The same impairment of binding was detected between Z(S173A) and the Rta promoter, but Z(S173A) was competent to activate expression of Rta to the same extent as wt ZEBRA [25]. This finding provoked the hypothesis that the DNA binding affinity of ZEBRA was of relatively greater importance for activation of viral replication than for activation of transcription. To further investigate this correlation we used an electrophoretic mobility shift assay (EMSA) to assess the DNA binding activity of ZEBRA RD mutants located in the basic domain of the protein. Fig. 5 compares the DNA binding activity of Z(Y180E) and K(188A) with that of wt ZEBRA and with Z(K188R), a mutant with a conservative change that manifests a wild phenotype. An EMSA assay was performed using cell extracts obtained from EBV negative HKB5/B5 cells transfected with the indicated expression vectors. Four ZEBRA response elements, ZIIIB and ZREs 1 to 3, were used as probes. ZIIIB represents the highest affinity binding site for ZEBRA; it mediates auto-stimulation of the ZEBRA promoter [37], [38]. ZREs 1–3 represent a cluster of sites present in the upstream essential region of oriLyt. Both Z(Y180E) and Z(K188A) were markedly impaired in binding to each of the four probes relative to wt ZEBRA. The efficiency of binding was calculated as the percentage of probe shifted by each mutant protein. Z(Y180E) shifted between 0.1% and 0.7% depending on the probe used in the shift assay; Z(K188A), 1% to 9.2%, and wt ZEBRA, 23.4% to 46% (Fig. 5A). The ZEBRA mutant, Z(K188R), which is fully competent to activate the lytic cycle [28], shifted the same set of ZEBRA specific DNA probes to percentages that were markedly higher than those observed with the ZEBRA RD mutants, namely 12.3% and 38.8% of the total probe (Fig. 5A). These in vitro DNA binding studies clearly indicated that Z(Y180E) and Z(K188A) are both significantly impaired in their capacity to bind to ZEBRA response elements present in regulatory sites for transcription or replication. The differences in DNA binding between wt ZEBRA and the mutants were not due to variable protein levels. Western blot analysis with an antibody against ZEBRA demonstrated that all EMSA extracts contained similar levels of ZEBRA protein (Fig. 5B).
To analyze the ability of the ZEBRA RD mutants to associate with the viral origin of lytic replication (oriLyt) in vivo, we employed chromatin immunoprecipitation (ChIP). To study the associations of the three basic domain ZEBRA RD mutants with oriLyt we transfected BZKO cells with expression vectors encoding each of the ZEBRA RD mutants, wt ZEBRA and a non-DNA binding form of ZEBRA, Z(R183E), which does not activate transcription or replication. In this experiment, the wild type protein was the only form of ZEBRA that was capable of inducing viral replication. To compare the amount of oriLyt immunoprecipitated by each ZEBRA protein we maintained equivalent levels of viral DNA by blocking viral replication with phosphonoacetic acid (PAA). We found that all ZEBRA RD mutants were more efficient than the non-DNA binding mutant Z(R183E), but less competent than wt ZEBRA in precipitating the upstream region of oriLyt. 3.7-fold less oriLyt was immunoprecipitated from cells expressing Z(Y180E) compared to those expressing wt ZEBRA (Fig. 6A). Similarly, Z(R187K) and Z(K188A) pulled down 2.9 and 8.3-fold less DNA than wt ZEBRA. The extent of association of each mutant with oriLyt was corrected for the total amount of oriLyt detected in the corresponding input sample. Fig. 6B shows that the level of input oriLyt was approximately the same in cells transfected with wild type and all three mutants. These results suggest that amino acid changes introduced in the three ZEBRA RD mutants did not completely abolish interaction of ZEBRA with oriLyt as was observed with the non DNA binding mutation R183E. Nonetheless, the ability of the RD mutants to bind to oriLyt in cells was 3- to 8-fold impaired compared with wt ZEBRA.
The Rta promoter (Rp) is a direct target for activation by ZEBRA. In Fig. 1, 2, S4 and S5, we showed that all ZEBRA RD mutants were fully competent to induce wild type levels of brlf1 (Rta) mRNA and protein. However, EMSA experiments showed that the ZEBRA RD mutants were similarly defective in binding to ZEBRA response elements regardless of their presence in transcription or replication regulatory regions (Fig. 5 and [25]). To investigate whether the ZEBRA RD mutants are impaired in their capacity to associate with Rp, in a separate experiment we carried out ChIP experiments to compare directly the capacity of two RD mutants to precipitate oriLyt and Rp DNA relative to wt ZEBRA. We found that Z(Y180E) and Z(K188A) were two-to three-fold defective in interacting both with Rp and with oriLyt when compared to the wild type protein (Fig. 6C and D). However, these RD mutants displayed higher efficiency to interact with oriLyt and Rp than the non-DNA binding ZEBRA mutant Z(R183E). The Z(R183E) mutant pulled down amounts of oriLyt and Rp that were equivalent to those detected in ChIP experiments performed with cells transfected with empty vector or precipitated with pre-immune serum (Fig. 6). The finding that RD mutants are equally impaired in binding to Rp and oriLyt suggests that activation of brlf1 transcription is more tolerant of weaker interaction between ZEBRA and its response elements than is stimulation of replication.
The ChIP assay measures the amount of DNA associated with ZEBRA, but does not measure how much ZEBRA interacts with DNA. Therefore, we employed a different approach to assay for the capacity of ZEBRA to bind DNA in cells (Fig. 7). The assay relied on co-transfecting vectors encoding wild type ZEBRA or ZEBRA mutants together with biotin-conjugated probes. The BUR probe is 167 bp long and encompasses the four ZREs in the upstream region of oriLyt that are crucial for lytic replication. BRpS and BRpL represent short (156 bp) and long (277 bp) segments of Rp. BRpS contains the ZIIIA site, while the BRpL has all three identified ZREs present in Rp. After 48 h, BZKO cells were harvested and biotinylated probes were captured using avidin coated beads. The level of ZEBRA protein bound to each probe was determined by western blot. The relative binding of ZEBRA to each probe was corrected for the total amount of ZEBRA protein present in each sample. In cells transfected with ZEBRA RD mutants, all three biotinylated probes pulled down less ZEBRA protein compared to cells transfected with wt ZEBRA. The defect in binding relative to wt ZEBRA averaged between 75% to 89% for the oriLyt probe (Fig. 7A); 57% to 93% for the short Rp probe (Fig. 7B), and 66% to 95% for long Rp probe (Fig. 7C). Our results with the transfected biotinylated probe assay confirm the EMSA and ChIP experiments. These three different assays show that replication defective mutants of ZEBRA are markedly impaired in binding to DNA. This defect in DNA binding can be seen with probes for oriLyt and Rp.
Our findings suggest that weak association of ZEBRA with oriLyt has significant ramifications for subsequent events that lead to lytic viral DNA replication. These events might involve a specific protein-protein interaction between ZEBRA and one or more of the replication proteins. In an attempt to restore this interaction we over-expressed the six components of the EBV replication machinery together with each of the ZEBRA RD mutants in BZKO cells. Over-expression of replication proteins partially rescued late gene expression by all four ZEBRA RD mutants. The extent of rescue ranged between 3- to 4-fold regardless of the level of late gene expression induced by each mutant in the absence of replication proteins (Fig. 8A, C and D). For example, in case of Z(S173A), expression of replication proteins reproducibly increased FR3 expression by 3.2-fold reaching 55% that of wt ZEBRA alone. This effect on late gene expression was not an anomalous feature of these mutants; a similar increase was detected with the wild type ZEBRA protein and ranged between 1.6- and 2.5-fold (Fig. 8C and D). While expression of the late FR3 protein can be used as an indirect marker for viral replication, we also examined the effect of over-expressing replication proteins on the capacity of wt ZEBRA and ZEBRA RD mutants to induce viral genome amplification. Expression of high levels of replication proteins reproducibly augmented the capacity of wt ZEBRA and Z(S173A) to stimulate EBV lytic replication by 1.9- and 3.4-fold respectively. In this experiment no similar effect of the complete mixture of replication proteins on DNA amplification was observed with the other ZEBRA RD mutants (Fig. 8B). However, subsequent experiments defined a subset of replication proteins that was capable of rescuing replication by all the RD mutants (Fig. S3).
In experiments illustrated in Figs. 5 to 7 and previously published [25] we found a direct correlation between strong association of ZEBRA with oriLyt and viral replication. To explore the possibility that replication proteins enhance interaction of ZEBRA with oriLyt, thereby partially restoring EBV lytic replication, we carried out ChIP experiments combined with quantitative PCR. In Fig. 9A, BZKO cells were transfected with empty vector (CMV), Z(S173A) or wt ZEBRA in the presence and absence of the six EBV replication proteins. In ChIP experiments, we found that BZKO cells transfected with Z(S173A) or wt ZEBRA yielded more oriLyt when replication proteins were co-expressed, 1.58-fold and 1.72-fold, respectively (Fig. 9A). This increase was independent of the level of ZEBRA expressed or immunoprecipitated. Western blot analysis showed that similar levels of ZEBRA protein were present in each immune-precipitate (Fig. 9B). Expression of the six replication proteins had no effect on the amount of oriLyt immunoprecipitated from cells transfected with empty vector.
A biological replicate experiment was performed and included two additional ZEBRA RD mutants, Z(Y180E) and Z(S167A/S173A) [25]. Wild type and mutant ZEBRA were expressed in BZKO cells minus and plus replication proteins. Co-expression of replication proteins enhanced the ability of wild type and mutant forms of ZEBRA protein to associate with oriLyt in vivo. A 1.8-fold increase in association with oriLyt was detected with wt ZEBRA; 2.2-fold with Z(S173A); 3-fold with Z(Y180E), and 4.24-fold with Z(S167A/S173A) (Fig. 9C).
A compilation of several chromatin immunoprecipitation experiments showed that replication defective ZEBRA mutants weakly associated with oriLyt (Fig. S2A). Z(S173A) was the least defective while Z(K188A) was the most impaired. For wild type ZEBRA and three of the mutants, Z(S173A), Z(Y180E) and Z(S167A/S173A), we demonstrated an increase in their association with oriLyt as a result of overexpressing the EBV replication proteins. The effect of replication proteins on association of ZEBRA with oriLyt was greatest with the mutant Z(S167A/S173A), 6.87-fold. Z(Y180E) precipitated 3 times more oriLyt in the presence replication proteins; Z(S173A), 1.8-fold, and wild type ZEBRA 1.6-fold (Fig. S2A). The two ZEBRA RD mutants which were most defective in binding to oriLyt, namely Z(R187K) and Z(K188A) were the least affected by replication proteins.
To investigate further the effect of replication proteins on interaction of ZEBRA with oriLyt we transfected BZKO cells with Biotin-conjugated oriLyt Full length (BOF) and expression vectors encoding wild type and the RD ZEBRA mutants with and without replication proteins. Cells were harvested 48 h after transfection; ZEBRA bound to oriLyt was purified using avidin coated beads. Both input and BoF-captured ZEBRA proteins were analyzed by Western blot. The effect of replication proteins on binding of ZEBRA to oriLyt was calculated after correcting for the amount of ZEBRA present in the corresponding input samples. We found that co-expression of the six core components of the replication machinery enhanced binding of wt ZEBRA, Z(Y180E), Z(R187K) and Z(S173A) to oriLyt by 2.1-, 6.0-, 16.4- and 5.0-fold, respectively. In summary the ChIP and iBDAA experiments demonstrate that the core components of the EBV replication machinery augment the interaction between ZEBRA and oriLyt.
To determine whether the effects of replication proteins were specific for ZEBRA's association with oriLyt, we examined the effect of replication proteins on association of ZEBRA with other lytic viral regulatory sites by studying interaction of ZEBRA and RD mutants with Rp and two other ZEBRA responsive promoters, BZLF1p (Zp) and BMRF1p (EAp). Zp is auto-stimulated by ZEBRA while BMRF1p is activated by synergy between ZEBRA and Rta. Over-expression of the six EBV replication proteins increased the relative amount of Rp, Zp and BMRF1p DNA precipitated by wt ZEBRA, Z(S167A/S173A) and Z(S173A) (Fig. S2B and Supplementary Table S1). The effect of replication proteins on the amount of viral DNA pulled down by Z(Y180E) was more pronounced on Rp (2.2-fold). The amount of Z(Y180E) bound to Zp and BMRF1p was minimally enhanced by replication factors, 1.3-fold and 1.25-fold, respectively. No difference was detected by ChIP for the effect of replication proteins on the relative binding capacity of Z(R187K) and Z(K188A) to Rp. This could be attributed to the marked defect in the DNA binding capacity of these two mutants or limitations in the ChIP technique to detect small changes in association with a particular site. Our results show that replication proteins enhance the interaction of ZEBRA and the phosphorylation site mutants with oriLyt, and with at least three transcription regulatory sites, Rp, Zp and EAp.
To delineate the contribution of each replication protein in restoring lytic viral DNA synthesis, Z(S173A) was co-expressed with different mixtures of replication proteins. In each mixture one of the six components was omitted. After 48 h, DNA was purified from BZKO cells and analyzed for its viral DNA content using quantitative PCR (Fig. S1). Elimination of individual components of the mixture of replication proteins led to several distinct outcomes. Exclusion of BBLF2/3 had no significant effect. Omission of BALF2 and BBLF4 reduced the efficacy of the replication proteins complex to rescue replication by Z(S173A). Eliminating BSLF1 or BMRF1 from the mixture of replication proteins abolished its activity. In contrast, omitting the expression vector of BALF5 augmented the capacity of the other five replication proteins to restore viral replication by Z(S173A). These results suggest that over-expression of different mixtures of replication proteins can stimulate, inhibit or have no effect on viral replication.
To select the minimum subset of replication proteins sufficient to suppress the phenotype of these RD ZEBRA mutants, we examined the effect of expressing the primase individually or together with various combinations of replication proteins excluding the polymerase (BALF5) that had been shown to be inhibitory (Fig. S1). After 48 h, transfected BZKO cells were analyzed by Western blot for the level of the FR3 protein as a marker for late gene expression. While co-expression of all six replication proteins with Z(S173A) induced late gene expression to 33.4 and 35.4% that of wt ZEBRA (Fig. 10A compare lane 3 to 4 and 13 to 14), addition of the primase alone had no significant effect on the level of the FR3 protein as compared to cells transfected with the S173A mutant in absence of RP. However, combining the primase with either the viral single-stranded DNA binding protein (BALF2) or the viral DNA polymerase processivity factor (BMRF1) enhanced late gene expression to 21.8 and 27.2% of wild type, respectively. A mixture containing all three proteins, the primase, the ssDNA-binding protein and the DNA polymerase processivity factor, restored late gene expression to 49.1%, a level higher than that induced by all six replication proteins (Fig. 10A lane 17). Addition of the viral helicase and/or the primase associated factor was either inhibitory or had no effect on the level of FR3.
To assess the effect of the different combinations of replication proteins on viral replication, we purified DNA from the same group of cells and analyzed it using quantitative PCR. The findings obtained by qPCR were similar to those seen by analyzing late gene expression. A mixture of the primase, the single-stranded DNA binding protein and the DNA polymerase processivity factor suppressed the defect in Z(S173A) and restored replication to approximately 44% that of the level activated by the wild type protein (Fig. 10A).
To determine if the same tripartite mixture of replication proteins could complement the defect in viral genome amplification observed in ZEBRA mutants in the DNA recognition domain, we repeated the same experiment using Z(R187K). Addition of all replication proteins induced viral replication 2.2-fold above that induced by Z(R187K) alone. Transfection of the primase and the DNA polymerase processivity factor together with Z(R187K) had no effect on late gene expression or viral DNA synthesis (Fig. 10B). However, addition of the single-stranded DNA binding protein to this mixture resulted in the highest impact on viral genome amplification, a 4.2-fold increase compared to replication induced by Z(R187K) alone (Fig. 10B). Similar results were observed for the effect of these three replication proteins on late gene expression (Fig. 10B).
The capacity of BALF2, BMRF1 and BSLF1 to rescue viral genome amplification by all five identified ZEBRA RD mutants was examined. BZKO cells were transfected with expression vectors encoding Z(S167A/S173A), Z(S173A), Z(Y180E), Z(R187K), Z(K188A) and wild type ZEBRA in the absence and presence of plasmids encoding the tripartite mixture of replication proteins. Cells were harvested at 48 h and 72 h and DNA was purified. The amount of EBV lytic replication induced by each condition was assessed by qPCR. At both time points, over-expression of the three components of the replication machinery enhanced activation of EBV lytic DNA replication by all five replication defective mutants as well as wt ZEBRA (Fig. S3).
Our findings stress the importance of the primase, the DNA polymerase processivity factor and the single-stranded DNA binding protein on suppressing the effect of ZEBRA mutations that render the protein incompetent to activate lytic DNA replication.
The upstream region of oriLyt encompasses four ZEBRA binding sites that are essential for oriLyt replication. Therefore it was important to assess directly the effect of the three replication proteins that rescued the function of the RD mutants on the capacity of ZEBRA to interact with the upstream region of oriLyt. In an iBDA assay, we transfected BZKO cells with a biotinylated upstream region of oriLyt (BUR) together with expression vectors for wt ZEBRA or Z(S173A) in the absence and presence of the tripartite replication mixture. BUR-bound ZEBRA was captured on avidin coated beads and the amount of ZEBRA bound was analyzed by western blot. We found that over-expression of the primase, the ssDNA-binding protein and the polymerase associated factor resulted in a 2.5- to 3.7-fold increase in the amount of ZEBRA that interacted with BUR (Fig. 10C). This finding supports a role for the tripartite mixture of replication proteins in lytic origin recognition by ZEBRA.
The results presented in supplemental Fig. S2B show that over-expression of replication proteins enhanced the capacity of wt ZEBRA, the phosphorylation site mutants and Z(Y180E) to interact with Rp, the BRLF1 promoter. The functional significance of expressing this subset of replication proteins on transcriptional activation of brlf1 by wt ZEBRA or mutant ZEBRA was studied in BZKO cells. To maintain an equal number of viral genome templates in each group, viral replication was blocked by phosphonoacetic acid (PAA) and the cells were harvested after 24 hours. Total RNA was purified and the level of brlf1 transcript was assessed using quantitative RT-PCR. Fig. S4A represents the average of two biological replicate experiments in which each value is the mean of three distinct RT-PCR reactions. As previously demonstrated in Fig. 2, expression of ZEBRA replication defective mutants induced the synthesis of up to 2.4-fold more brlf1 mRNA than did wt ZEBRA. Over-expression of the tripartite mixture of replication proteins co-stimulated synthesis of the brlf1 transcript to various levels depending on the form of the ZEBRA protein being expressed. Replication proteins had a modest effect on the capacity of wt ZEBRA, Z(S173A) and Z(S167A/S173A) to activate transcription of brlf1 (1.3 to 1.5- fold). A significant 2.4-fold to 3.6-fold increase in the level of the brlf1 transcript was detected when BALF2, BMRF1 and BSLF1 were co-expressed with each of the three DNA binding domain ZEBRA mutants, Z(Y180E), Z(R187K) and Z(K188A). These results show that despite the defect in activating DNA lytic replication, all ZEBRA RD mutants were capable of activating transcription to levels equal to or higher than that of wt ZEBRA. In addition, replication proteins enhanced the capacity of wt ZEBRA and ZEBRA RD mutants to activate transcription of brlf1. This effect was more prominent when the BALF2-BMRF1-BSLF1 mixture was co-expressed with any of the three mutant forms of ZEBRA proteins containing single point mutations in the DNA recognition domain. The tripartite mixture of replication factors also enhanced the level of Rta protein activated by wild-type and mutant ZEBRA proteins. The enhancement was most marked for RD mutants Z(Y180E) and Z(R187K) (Fig. S4B).
In this study we provide evidence that a subset of virally encoded replication proteins enhance origin recognition by ZEBRA during lytic viral replication by promoting the capacity of ZEBRA to bind to viral DNA. ZEBRA binds specifically to a set of DNA sequences that are scattered throughout viral and cellular genomes. The EBV origin of lytic replication, oriLyt, is recognized by ZEBRA which also serves as a strong transcription activator by binding to lytic gene promoters. The ability of ZEBRA to perform two distinct functions in the same cell poses a biologically important question, namely, how would a protein like ZEBRA distinguish between a site that promotes transcription and another one that triggers replication? We addressed this question by characterizing a set of ZEBRA mutations that specifically disrupted the protein's ability to activate lytic viral replication (Fig. 1). These mutations are not significantly impaired at activating transcription and on most targets are enhanced as transcription activators (Fig. 2, 3, S4A and S5).
A common defect observed among all five mutants was reduced DNA binding activity. Impairment of the ZEBRA mutants to interact with DNA was not specific to a particular ZEBRA response element and was detected whether we studied binding of ZEBRA to oriLyt or to promoters that regulate expression Rta, ZEBRA and BMRF1 (Fig. 5, 7, S2B and Table S1). However, the defect in DNA binding seemed to specifically disrupt activation of viral replication without affecting transcription (Fig. S5). This feature of the ZEBRA RD mutants allowed us to investigate the effect of EBV replication proteins on the interaction between ZEBRA and oriLyt and to correlate this effect with activation of viral replication. Increasing the concentration of all EBV replication proteins rescued the defect in viral replication and enhanced the formation of the ZEBRA-oriLyt complex (Fig. 8 and 9). A similar effect for replication proteins was observed with wt ZEBRA suggesting that this effect is not an artifact caused by the mutations installed in ZEBRA or due to failure of the RD mutants to activate expression of replication proteins. Subtraction experiments indicated that removal of the DNA polymerase (BALF5) from the mixture of replication proteins enhanced DNA replication while removal of expression vector encoding the viral primase (BSLF1) or the polymerase processivity factor (BMRF1) was detrimental (Fig. 10 and Fig. S1). In a reconstruction experiment, three EBV replication proteins were found to be sufficient to suppress the defect in replication and DNA binding associated with the ZEBRA RD mutants; these are the primase, the single-stranded DNA binding protein and the DNA polymerase processivity factor (Fig. 10 and Fig. S3). Expression of this tripartite replication mixture increased the level of Rta (brlf1) transcript and protein (Fig. S4). Thus, replication proteins seem to co-stimulate the capacity of ZEBRA to activate expression of Rta and consequently expression of replication factors prior to viral genome amplification. In summary, our findings support a model (Fig. 11) in which replication proteins promote lytic viral DNA synthesis in at least three different ways: i) replication proteins co-stimulate the capacity of ZEBRA to express Rta and other early lytic cycle products; ii) replication proteins augment the ability of ZEBRA to interact tightly with oriLyt, and iii) replication proteins comprise the EBV lytic replication machinery.
ZEBRA RD mutants with compromised DNA binding activity can be divided into two subclasses: the phosphorylation site mutants, Z(S173A) and Z(S167A/S173A), and the basic domain mutants, Z(Y180E), Z(R187K) and Z(K188A) [25], [28]. The defect in DNA binding was demonstrated using three different DNA binding assays: EMSA, ChIP and in vivo biotin-conjugated DNA affinity assay (iBDAA). Each of these assays addressed a different aspect of the DNA binding activity of ZEBRA. EMSA compared the capacity of the ZEBRA RD mutants to bind to individual ZREs in vitro. Four ZREs were tested, three present in oriLyt (ZRE1–3) and a fourth (ZIIIB) in Zp. The defect in binding to these sites by the ZEBRA RD mutants was severe relative to the wild type protein. However, examining the ability of the mutants to associate with oriLyt in vivo using ChIP revealed a milder defect (2- to 8-fold) (Fig. 6). This difference could be attributed to several factors; for example, ZEBRA binds to ZREs present in oriLyt in a cooperative manner [39], other viral proteins affect ZEBRA association to oriLyt (Fig. 9), and formation of pre-replication foci increases the local concentration of ZEBRA [40]. To directly assess the level of ZEBRA protein bound to oriLyt or Rp, we examined interaction of ZEBRA with biotinylated probes in BZKO cells. Using this in vivo biotinylated DNA affinity assay (iBDAA), we showed that all the mutants were impaired in their capacity to associate with both the oriLyt and Rp probes (Fig. 7).
Our studies on the DNA binding activity of ZEBRA revealed important correlations between strong interaction of ZEBRA with oriLyt and its capacity to activate viral replication. These correlations are: 1) all five ZEBRA mutants defective in activating viral replication exhibited a 2- to 8-fold defect in interacting with oriLyt (Fig. S2A). 2) The level of reduction in the DNA binding of each mutant correlated with its defect in stimulating viral replication (Fig. S5). 3) Replication proteins that enhanced interaction of ZEBRA with oriLyt restored viral replication (Fig. 9 and S2A). 4) At position S173, a mutation that disrupts DNA binding, e.g. Z(S173A), also abolishes viral replication, while another substitution that maintains DNA binding, e.g. Z(S173D), has no effect on viral replication [25]. All together these correlations point to the importance of strong interaction between ZEBRA and oriLyt to stimulate viral replication.
Specific DNA binding by a protein that regulates different processes is not sufficient to confer specificity; additional levels of regulation likely play an important role beyond the initial step of DNA recognition. Consistent with the notion that initial recognition of the origin by the origin binding protein per se is not sufficient to induce replication, in Saccharomyces cervisiae, interaction with Cdc6p increased sequence-specific binding of ORC to the origin by altering its structure [29]. Also, the herpes simplex virus polymerase processivity factor (UL42) facilitated loading of the origin binding protein (UL9) to single-stranded or partially duplex DNA [41]. This study was done in vitro and the effect of replication proteins on the process of replication was not directly assessed in infected cells.
Here, we showed that expression of replication proteins enhanced interaction of ZEBRA with both oriLyt and Rp. This enhancement in binding is likely to have more impact on replication than on transcription of brlf1 for the following reasons: 1) the ZEBRA RD mutants were fully competent to activate transcription of Rta and other lytic products. 2) Replication, and not transcription, was dependent on the capacity of ZEBRA to strongly bind to the corresponding viral DNA regulatory sites. 3) The replication proteins not only enhanced oriLyt recognition by the ZEBRA RD mutants but also restored their capacity to activate viral replication and late gene expression.
Over-expression of the tripartite mixture of replication factors did not rescue viral replication by the ZEBRA RD mutants to wild type level. This could be due to the presence of additional defects, other than DNA binding, in the ZEBRA RD mutants. Alternatively, over-expression of other viral or cellular proteins might be necessary to completely suppress the phenotype of these mutants in replication. However, a complete rescue of the mutants might be technically challenging since it is unlikely that all the cells will obtain and express the transfected plasmids.
Two findings suggest that replication proteins exert their effects early, during the assembly of the pre-replication complex or in the initial stages of replication rather than in extension. First, omission of the viral DNA polymerase (BALF5) expression vector markedly enhanced viral replication (Fig. S1). Second, addition of phosphonoacetic acid (PAA), an inhibitor of the viral DNA polymerase, had no effect on the ability of replication proteins to enhance ZEBRA association with oriLyt (Fig. 9). This effect of replication proteins is specific to three of the six replication proteins and is unlikely to be due to over-expression. In other EBV-infected cell lines, such as the Burkitt lymphoma derived cell line HH514-16, replication proteins are expressed at much higher levels than in transfected BZKO cells (Fig. 3A and [25]).
Origin recognition is a complex process that is regulated at multiple levels. In addition to the role of replication proteins in enhancing association of ZEBRA with oriLyt, other mechanisms must also be involved. For example, interaction of BBLF2/3 with ZBRK1 serves as a tethering point on oriLyt for other replication proteins [19]. The involvement of multiple mechanisms in regulation of origin recognition reflects the complexity of such an initial but essential step for activation of EBV replication.
At the initial stage of the EBV lytic cycle, stimulation of Rta expression by ZEBRA is independent of the presence of any replication proteins. As the lytic cycle proceeds into the early phase, ZEBRA and Rta, solely or synergistically, activate transcription of genes encoding the different components of the replication machinery. Our data shows that expression of three replication proteins, BALF2, BMRF1 and BSLF1 positively modulate the capacity of ZEBRA to stimulate expression of Rta (Fig. 2, S2B and S4). The co-stimulatory effect of this subset of replication proteins on Rta expression is likely to be a secondary event that occurs later during the pre-replicative phase of the lytic cycle. Our findings suggest that replication proteins trigger a positive feedback mechanism prior to viral replication that increases the level of Rta and replication proteins (Fig. 11). Upsurge in expression of replication proteins is likely to play a significant role in origin recognition, assembly of the replication complex and the process of viral DNA synthesis.
Evidence supporting the possible role of replication proteins in a positive feedback loop comes from a recent report suggesting that the DNA polymerase processivity factor, BMRF1, enhances the capacity of ZEBRA to activate the BALF2 promoter [42]. BMRF1 has also been shown to modulate the ability of Sp1 and ZBP-89 to activate the early viral BHLF1 promoter and the cellular gastrin promoter [43], [44]. The mechanism responsible for the transcriptional co-activation function of BMRF1 is still unknown. It is possible that the effect of replication proteins in augmenting the capacity of ZEBRA to activate transcription is mediated by BMRF1 only.
The following models might account for the role of replication proteins in origin recognition. First, ZEBRA interacts with sub-complex(es) containing the three replication proteins, the primase, the ssDNA-binding protein and the DNA polymerase processivity factor, off DNA. This interaction results in the formation of a high affinity quaternary origin recognition complex. Second, ZEBRA binds independently to oriLyt and interacts with replication proteins that are already tethered to oriLyt through other cellular transcription factors, e.g. Sp1 and ZBRK1 [18], [19], [23]. The formation of this network of protein-protein interactions with multiple contacts among replication proteins, ZEBRA and oriLyt is likely to have a synergistic effect on the stability of this protein-DNA complex and to facilitate recruitment of other replication proteins [45]. Third, formation of the ZEBRA-oriLyt complex results in a specific DNA-protein architecture that functions as a landing pad for the three replication proteins which in turn augment and stabilize the interaction between ZEBRA and oriLyt. One possible function for the three replication proteins is to enhance the capacity of ZEBRA to occupy all the ZREs present in oriLyt (Fig. 9D and Fig. 10C). ZEBRA-oriLyt complexes that are not recognized by these three proteins are likely to become unstable and will fail to assemble a functional replication complex. These models do not yet account for the precise role of individual proteins. For example, it is possible that only one of these proteins, such as the ssDNA-binding protein, alters the origin binding capacity of ZEBRA while the two other proteins are important in subsequent events.
Based on our results, we propose that a tripartite mixture of replication proteins plays a role in EBV lytic origin recognition. This is a novel role for replication proteins. Additional experiments will be necessary to investigate the mechanism(s) by which each of these three replication proteins modulate the binding activity of ZEBRA to oriLyt and other ZEBRA response elements and enhance viral replication and transcription.
The plasmids pHD1013/Z, pHD1013/Z(S173A), pHD1013/Z(R187K), pHD1013/Z(Y180E), pHD1013/Z(K188A), pHD1013/Z(K188R), pHD1013/Z(F193E) and pHD1013/Z(K194A) were prepared as described previously [28], [46]. Expression vectors for the viral open reading frames encoding BALF5, BBLF4, BBLF2/3 and BSLF1 were a kind gift from Dr. Diane Hayward [4]. The full length coding sequences for BALF2 and BMRF1 were amplified from EBV genomes purified from HH514-16 cells by PCR. The amplified fragments were cloned in pFLAG-CMV2 using EcoR1 and Xba1 restriction sites.
BZKO cells were previously described [1]. HKB5/B5 cells represent an EBV negative subclone that was initially generated by hybridizing HEK293 cells with the EBV positive cell line HH514-16 [47]. All transient transfection experiments were performed in 25 cm2 flasks using 3 µg of ZEBRA expression vector and 2 µg of each construct encoding a replication protein. The DMRIE-C reagent (Invitrogen) was used for transfection according to the manufacturer's protocol.
Immunoblotting was performed as previously described [28]. The following antibodies were used: anti-ZEBRA, anti-FR3 and anti-LR2 are polyclonal rabbit sera raised to TrpE-fusion proteins expressed in E. coli. The anti-Rta antibody was generated by expressing the C-terminal 320 a.a of Rta using the pET-expression system. The fragment was purified on a nickel column and used for rabbit immunization. The monoclonal antibody against BMRF1 (EA-D) (R3.1) was a kind gift from G. Pearson. Anti-FLAG is a mouse monoclonal antibody (Sigma). ChIP experiments were performed as previously detailed [25]. Sequences for the primers used are available upon request.
RNA was purified from 8×106 BZKO or HH514-16 cells using RNeasy kits (Qiagen) according to the manufacturer's protocol. All RNA samples were treated with 30 U DNase1 (Qiagen). Twenty micrograms of RNA was separated on 1% agarose gel and transferred by capillarity to a Hybond N+ filter (Amersham). The membrane was hybridized to two 32P-labeled probes detecting the H1 component of RNase P (a loading control) and BALF2. The probes were generated from a 370-bp NcoI-Pst1 fragment of RNase P and full length BALF2 DNA using random primers.
DNA was isolated from 107 BZKO cells as detailed previously [28]. Ten micrograms of DNA was digested with 40 units BamH1 (New England Biolabs) for 3 h at 37°C. DNA fragments were separated by electrophoresis in a 0.8% agarose gel and transferred to a Zeta probe GT genomic membrane (Bio-Rad). Formation of a replication ladder was detected using a probe complementary to a 336-bp sequence in the unique Xho 1.9-kb sequence upstream of the viral terminal repeats [34]. The template for the 336-bp probe was generated by PCR using the following primers 5′-CTCACGAGCAGGTGG-3′ and 5′-CGCAGTCTTAGGTATCTGG-3′. An excised EBV BamH1 W fragment was used as a template to generate a corresponding probe [48]. Radioactive probes were synthesized using 10 units of the Klenow fragment of DNA polymerase (New England Biolabs), [α-32P]dCTP and 1 ng random primers. The probes were purified using a Sephadex-G50 column.
RNA samples were prepared 48 h after transfection of BZKO cells. Phosphonoacetic acid (PAA) was added to inhibit viral replication. RT-PCR was performed on 100 ng of total RNA using reagents and instructions described in the manual for the SuperScript One-Step RT-PCR with platinum Taq kit (Invitrogen). In reactions where the reverse transcriptase was omitted, 2 units of platinum Taq was added. Random hexamers or gene specific primers were used to generate cDNA. A 131-bp fragment was amplified by the BBLF4 primers; a 121-bp fragment by the BSLF1 primers, and 211-bp by the BBLF2/3 primers. The sequences for the primers used are available upon request. Incorporation of Sybr green into DNA was detected using Cepheid Smart Cycler II or a Bio-Rad MyiQ real time PCR machines. Standard curves were generated using 10-fold serial dilutions of expression vectors encoding each of the three open reading frames. Quantitative PCR for detection of viral genome amplification was previously described [25].
Preparation of supernatants of HKB5/B5 cell extracts expressing wt ZEBRA or mutants as well as the DNA binding reactions were previously described [28]. Super-shifts were performed using BZ1, a monoclonal antibody against the dimerization domain of ZEBRA. The percent probe shifted is calculated as previously described [28], [49].
Full length oriLyt was cloned into pBSKII+ from HH514-16 cells using primers containing EcoR1 and BamH1 sites 5′-GCGCGAATTCTGGGGTCTCTGTGTAATACTTTAAG-3′ and 5′-GCGCGGATCCGTTA ATAAGGAGCC GTCCTTATTC-3′. Biotin-labeled full length oriLyt (BoF) was prepared by PCR using primers that were conjugated to biotin at their 5′ends. BZKO cells were co-transfected with 150 ng BoF and the indicated expression vectors. Cells were harvested after 60 h and re-suspended in lysis buffer containing 15 mM Tris-HCl pH 8.1, 167 mM NaCl, 1.2 mM EDTA, 3 mM MgCl2, 0.01% SDS and 1.1% Triton X-100. Cell extracts were briefly sonicated and supernatants were collected. The amounts of total protein were assessed using the Bradford reagent (Bio-Rad) and equalized. ZEBRA bound to BoF was captured using Avidin coated beads. The beads were washed and heated in SDS-PAGE sample buffer. The amount of captured ZEBRA protein was determined using Western blot analysis.
|
10.1371/journal.pcbi.1000954 | A Covering Method for Detecting Genetic Associations between Rare Variants and Common Phenotypes | Genome wide association (GWA) studies, which test for association between common genetic markers and a disease phenotype, have shown varying degrees of success. While many factors could potentially confound GWA studies, we focus on the possibility that multiple, rare variants (RVs) may act in concert to influence disease etiology. Here, we describe an algorithm for RV analysis, RareCover. The algorithm combines a disparate collection of RVs with low effect and modest penetrance. Further, it does not require the rare variants be adjacent in location. Extensive simulations over a range of assumed penetrance and population attributable risk (PAR) values illustrate the power of our approach over other published methods, including the collapsing and weighted-collapsing strategies. To showcase the method, we apply RareCover to re-sequencing data from a cohort of 289 individuals at the extremes of Body Mass Index distribution (NCT00263042). Individual samples were re-sequenced at two genes, FAAH and MGLL, known to be involved in endocannabinoid metabolism (187Kbp for 148 obese and 150 controls). The RareCover analysis identifies exactly one significantly associated region in each gene, each about 5 Kbp in the upstream regulatory regions. The data suggests that the RVs help disrupt the expression of the two genes, leading to lowered metabolism of the corresponding cannabinoids. Overall, our results point to the power of including RVs in measuring genetic associations.
| We focus on the problem of detecting multiple rare variants (RVs) that act together to influence disease phenotypes. In considering this problem, we argue that the detection of causal rare variants must necessarily be different from typical single-marker analysis used for common variants and propose a novel algorithm, RareCover, to accomplish this analysis. RareCover combines a disparate collection of RVs, each with very low effect and modest penetrance. Extensive simulations over a range of values for penetrance and population attributable risk (PAR) illustrate the power of our approach over other published methods, including the collapsing and weighted-sum strategies. To showcase the method, we applied RareCover to data from 289 individuals at the extremes of Body Mass Index distribution (NCT00263042), sequenced around the FAAH and MGLL genes. RareCover analysis identified exactly one significantly associated region in each gene, each about 5Kbp in the upstream regulatory regions. The data suggests that the RVs help disrupt the expression of the two genes leading to lowered metabolism of the corresponding endocannabinoids previously linked with obesity. Overall, our results point to the power of including RVs in measuring genetic associations, and suggest that whole genome, DNA sequencing-based association studies investigating RV effects are feasible.
| The Common Disease, Common Variant (CDCV) hypothesis [1]–[3] postulates that the etiology of common diseases is mediated by commonly occurring genomic variants in a population. This has served as the basis for genome wide association (GWA) studies that test for association between individual genomic markers and the disease phenotype. Using genome-wide panels of common SNPs, GWA studies have been successful in identifying hundreds of statistically significant associations for many common diseases as well as several quantitative traits [4]–[7]. Nevertheless, the success of GWA studies has been mixed. Significant genetic loci have not been detected for several common diseases that are known to have a strong genetic component [4]. Additionally, for many common diseases, associations discovered in GWA studies can account for only a small fraction of the heritability of the disease. While many factors could potentially confound GWA studies, we focus on the possibility that multiple, rare variants may act in concert to influence disease etiology.
The alternative to the CDCV hypothesis, the ‘Common Disease, Rare Variant (CDRV)’ hypothesis has been the topic of much recent debate [8], and has shown promise in explaining disease etiology in multiple studies. For example, rare variants (RVs) have been implicated in reduced sterol absorption and, consequently, lower plasma levels of LDL [9], [10] and colorectal cancer [11]. While some studies have shown RVs to increase risk, a recent study indicates that RVs also act ‘protectively’, with multiple RVs in renal salt handling genes showing association with reduced renal salt resorption and reduced risk of hypertension [12]. Additionally, rare mutations in IFIH1 have been shown to act protectively against type 1 diabetes [13].
The aforementioned studies and others focused on re-sequencing of the coding regions of candidate genes using Sanger sequencing (see Table 1 in Schork et al. [8] for a summary). Recent technological advances in DNA sequencing have made it possible to re-sequence large stretches of a genome in a cost-effective manner. This is enabling large-scale studies of the impact of RVs on complex diseases. However, several properties of rare variants make their genetic effects difficult to detect with current approaches. Bodmer and Bonilla provide an excellent review of the properties of RVs, and the differences between rare, and common variant analysis [14]. As an example, if a causal variant is rare ( MAF ), and the disease is common, then the allele's Population-Attributable-Risk (PAR), and consequently the odds-ratio (OR), will be low. Additionally, even highly penetrant RVs are unlikely to be in Linkage Disequilibrium (LD) with more common genetic variations that might be genotyped for an association study of a common disease. Therefore, single-marker tests of association, which exploit LD-based associations, are likely to have low power. If the CDRV hypothesis holds, a combination of multiple RVs must contribute to population risk. In this case, there is a challenge of detecting multi-allelic association between a locus and the disease.
Methods to detect such associations are only just being developed. A natural approach is a collapsing strategy, where multiple RVs at a locus are collapsed into a single variant. Such strategies have low power when ‘causal’ and neutral RVs are combined (See for example, Li and Leal [15]). Madsen and Browning have recently proposed a weighted-sum statistic to detect loci in which disease individuals are enriched for rare variants [16]. In their approach, variants are weighted according to their frequency in the unaffected sample, with low frequency variants being weighted more heavily. Each individual is scored as a sum of the weights of the mutations carried. The test then determines if the diseased individuals are weighted more heavily than expected in a null-model. Madsen and Browning show that with of variants in a group being causal and a combined odds ratio , the weighted-sum statistic detects associations with high power. While effective, this approach depends upon the inclusion of high proportion of causal rare variants in the formation of the test statistics and strong penetrance to detect significant association. In their simulations, the PAR of the locus is partitioned equally among all variants, an assumption that may not always hold.
The Combined Multivariate and Collapsing Method (CMC), proposed by Li and Leal, combines variants into groups based upon predefined criteria (e.g. allele frequency, function) [15]. An individual has a ‘1’ for a group if any variant in the group is carried and a ‘0’ otherwise. The CMC approach then considers each of the groups in a multivariate analysis to explain disease risk. This combination of the collapsing approach and multivariate analysis results in an increase of power over single-marker and multiple marker approaches. However, as Li and Leal point out, the method relies on correct grouping of variants. The power is reduced as functional variants are excluded and non-functional variants are included in a group. Assignment of SNPs to incorrect groups may, in fact, decrease power below that attainable through single marker analysis. Indeed, a recent analysis by Manolio and colleagues suggests that new methods might be needed when the causal variants have both low PAR and low penetrance values [17].
Here, we focus on a model-free method, RareCover, that collapses only a subset of the variants at a locus. Informally, consider a locus encoding a set of rare variants. RareCover associates with a phenotype by measuring the strongest possible association formed by collapsing any subset of variants at . At first glance, such an approach has many problems. First, selecting an optimal subset of SNPs is computationally intensive, scaling as . We show that a greedy approach to selecting the optimal subset scales linearly, making it feasible to conduct associations on a large set of candidate loci.
A second confounding factor is that the large number of different tests at a locus increase the likelihood of false association. The adjustment required to control the type I error could decrease the power of the method. However, extensive simulations show otherwise. Our results suggest that moderately penetrant alleles with small PAR , and moderately sized cohorts ( cases and controls) are sufficient for RareCover to detect significant association. This compares well with the current power of single-marker GWA studies on common variants, and outperforms other methods for RV detection.
We also applied RareCover to the analysis of two genes, FAAH, and MGLL, in the endocannabinoid pathway in a large sequencing study of obese and non-obese individuals. The endocannabinoid pathway is an important mediator of a variety of neurological functions [18], [19]. Endocannabinoids, acting upon CB1 receptors in the brain, the gastrointestinal tract, and a variety of other tissues, have been shown to influence food intake and weight gain in animal models of obesity. Using a selective endocannabinoid receptor (CB1) antagonist, SR141716 (Rimonanabt; Sanofi-Synthelabo) leads to reduced food intake in mice. Correspondingly, elevation of leptin levels have been shown to decrease concentrations of endogenous CB1 agonists, Anandamide, and 2-AG in mice, thereby reducing food-intake [20]. The FAAH and MGLL enzymes serve as regulators of endocannabinoid signaling in the brain [21], by catalyzing the hydrolysis of endocannabinoid including anandamide (AEA), and 2-AG. Gene expression studies in lean and obese women show significantly decreased levels of AEA and 2-AG, as well as over-expression of CB1 and FAAH in lean, as opposed to obese women [22]. While evidence points to a genetic association of these loci with obesity, multiple recent studies using common SNPs in the FAAH region have failed to confirm an association [23]–[26]. A Pro129Thr polymorphism was tentatively associated with obesity in a cohort of Europe and and Asian ancestry, but has not been replicated in other data [27].
We tested the hypothesis that multiple, rare alleles at these loci are associated with obesity. We have used unpublished (submitted) data from Frazer and colleagues, where the FAAH (Kbp) and MGLL (Kbp) regions were re-sequenced using next generation technologies in 148 obese and 150 non-obese individuals taken as extremes of the body mass index distribution from subjects in a large clinical trial (the CRESCENDO cohort, NCT00263042). The resequencing identified a number of common, and rare variants in the region. We applied RareCover to determine if multiple RVs, i.e., allelic heterogeneity, mediated the genetic effects of FAAH and MGLL on obesity. RareCover identified a single region at each locus with permutation adjusted p-values of and . In each case, the significant locus was immediately upstream of the gene, consistent with a regulatory function for the rare variants.
We define a locus as a genomic region of fixed size (nucleotides). Let denote the set of RVs in the locus. We abuse notation slightly by using to also denote the locus itself. A case-control study at includes a set of individual genotypes. For genotype , and RV , let denote the number of minor alleles that genotype carries for variant . Extending the notation to subsets, of RVs, define . For a subset , denote a union-variant as follows: individual has the allele if and only if . Otherwise, . The union-variant is a virtual construct that helps combine the effect of multiple RVs. Let (respectively, ) represent the case (respectively, control) status of an individual.
For an individual chosen at random, and , let Xcorr denote an association test statistic between the union-variant and the disease status . Here, we will use Pearson's as the test-statistic, but the method remains unchanged for other measures. Using this notation, the collapsing strategy described by Li and Leal [15] uses the test-statistic Xcorr to associate a locus with the disease. Instead, we define the association statistic for locus by(1)
Our method, RareCover, accepts a locus containing a set of RVs in a window of fixed size (nucleotides). It returns the test-statistic, Xcorr , a -value on the statistic, and the subset of RVs that contribute to the union variant. The window size is a parameter. When the input locus is larger than the window size, RareCover looks at overlapping windows of size , where each window is shifted one RV away from the previous window. For each window, the Xcorr statistic is output, along with a non-adjusted -value.
The computation for the Xcorr statistic on a single window is described in the Algorithm below. Given a set of RVs over individuals, the naive computation for computing Xcorr needs computations. A reduction from the MAX-COVER problem can be used to show that the problem is NP-hard, indicating that no provably efficient algorithm is likely [28]. Similar reductions can also be used to show the hardness result for a variety of other proposed association statistics. Therefore, we employ a greedy heuristic that is fast ( computations), and does well in practice. In each step, (see Algorithm), we select the variant that adds the most to the statistic, until no further improvement is possible. On a standard Linux workstation, the computation is fast, about windows per second.
Consider a locus with a set of rare-variants. Let a subset of RVs be causal, in the sense that a mutation at any increases the likelihood of disease. For an individual, , we use and as short-forms of the events , and , respectively. Similarly, events reflect case-control status for the individual. We work with the following parameters for power calculations:
Note that the PAR for a variant is often described by the following (Ex:Bodmer and Bonilla, 1999 [14])(3)where is the number of individuals with the phenotype, and is the number of individuals that show the phenotype, but do not have the variant allele. In our terminology
The choice of these parameters is intuitive as we expect an RV to have moderate penetrance, but very low PAR . However, the multiple RVs in have roughly additive effect, leading to moderate locus-PARs. These parameters are tightly related to other, more common measures of locus association, such as the Odds-Ratio (OR), as shown below:To compute , we start with a Bayesian relation for computing the likelihood of a genotype containing a causal RV as(4)Then,(5)and,(6)
We simulate multiple case-control studies over a range . A simulation of individuals begins with the division of the individuals into cases and controls. Once this is done two additional steps take place.
We start by generating causal RVs. As RVs do not show high LD, we can model the population by generating each RV independently. We adapt Pritchard's argument that the frequency distribution of rare, deleterious, RVs must follow Wright's model under purifying selection [29]. Therefore, the allele frequencies are sampled according to:(7)where,
We choose [29]. Note that we do not control the number of causal RVs, , directly, in our simulation. Recall thatFurther,Therefore, setting a value for R limits the size of the causal RV.(8)Further, the sampling procedure occasionally generates SNPs with a high individual PAR. These variants would show up as being significant even with a single marker analysis. Therefore, these are discarded. The procedure SimulateRV describes the method to generate causal RVs. To generate neutral RVs, we use Fu's model of allele distributions [30] on a coalescent, which suggests that the number of mutations that affect individuals in a population with mutation rate is given by . For the purposes of our simulation we use .
For both cases and controls, each RV is sampled independently. For non-causal variants , the probability of picking a minor allele is , for both case and control individuals. To sample alleles from causal SNPs, recall that under the union model, for all . Therefore, the minor allele frequencies are given byWe assume HW equilibrium to sample genotypes for case and control individuals.
Recently, Kryukov and colleagues [31] described a demographic model that explicitly models European ancestry. The population is assumed to be relatively stable for a long period, but is followed by a bottleneck, and rapid expansion after the bottleneck (about 7500–9000 years ago, with 20–25 years per generation). They validate their model by comparing observed versus predicted allelic frequencies. To this model, they add ‘causal’ (mostly deleterious) mutations using a distribution of selection coefficients from a gamma distribution. The causal alleles are associated with a change in a quantitative trait (QT). The QT values are normally distributed. Individuals carrying any causal mutation have QT values drawn from a Normal distribution with a shifted mean. For Rare variant analysis, individuals are chosen from the lower (Control) and upper (Case) tails of the QT distribution.
For our study, the authors provided us with individual genotypes simulated according to their demographic model, with causal mutations contributing to the following shifts: (Low), (Medium), (High). The highest and lowest , and % of the QT distributions were used for the Case and Control populations. For the % population (500 controls, 500 cases), the locus PAR varied as –. For the % populations, the number of individuals is larger (1000 controls, 1000 cases), but the PAR values decrease to (Low), (Medium), and (High).
For the purposes of comparison, we reimplemented the collapsing statistic proposed by Li and Leal [15] as well as the weighted-sum statistic used by Madsen and Browning [16]. Both publications discuss the separation of variants into groups based upon function (i.e. non-synonymous coding SNPs) or other property. However, because we are performing our studies on model free, unannotated data, we do not perform any such grouping.
As a result, the CMC approach proposed by Li and Leal [15] is equivalent to collapsing all variants in the locus and calculating the association. Li and Leal show that the assignment of variants to functional groups, separately collapsing these groups, and finally performing a multivariate analysis improves power to detect causal loci. However, separation of variants into groups is inexact and the authors show that errors in group assignment can confound tests for significance. Additionally, performing this separation on a genome wide scale may be intractable.
The weighted-sum statistic proposed by Madsen and Browning [16] is used to detect association between a pre-defined group and a disease state. To compare fairly we defined the group of mutations as all mutations at a locus. We reimplemented the weighting approach based upon allele frequency as well as the sum and ranking approach to determine a score. Finally, we implemented a single-marker test as a bi-allelic -statistic with 1df. The tests were used to score windows over a wide range of simulation parameters to better understand how RareCover performed in comparison to the collapsing, and weighted strategies. For each strategy, a -value of significance was established by doing randomized trials using permuted case and control data. All three methods were run on the same sets of permuted data, and the -values were used to compare. Code for all methods is available upon request from the authors.
Recently, Yeager and colleagues [32] resequenced a Kbp region including the micro-seminoprotein- (MSMB) gene (chr10:–) for prostate cancer cases, controls, plus another CEPH individuals. While the number of individuals is too small to derive rare variants, we used the prediced genotypes supplied by the authors for RV analysis. For this analysis, we used individuals together as controls, and all 284 variants with MAF % were used as input to RareCover.
In a recently submitted study LR-PCR amplicons (Harismendy et al., unpublished) were used to re-sequence Kbp from the FAAH locus (NCBI36 chr1:46621328–46653043) and Kbp from the MGLL locus (NCBI36 chr3:128880456–129037011). A total of individuals were selected for sequencing from two tails of the BMI distribution of the CRESCENDO cohort (http://clinicaltrials.gov/ct/show/NCT00263042). individuals had BMI lower than kg/ and individuals a BMI greater than kg/. DNA sequencing libraries were prepared, and sequenced as previously described in Harismendy, 2009 [33] with the following modifications: sequencing libraries were indexed by nt barcode located downstream of the adapter [34] and between one and six libraries were loaded per lane of the Illumina GAII instrument. The reads obtained from several lanes were merged, aligned and the variant called using MAQ mapmerge, map and cnsview+SNPfilter options respectively [35]. All samples had an average coverage greater than . Allowing for a minimum coverage of reads and a minimum base quality (Phred ), a raw set of single nucleotides variants (SNVs) were identified in the population. The SNVs were filtered for Hardy Weinberg Equilibrium in the controls () and genotyping rate % of the samples to obtain a final set of SNVs (220 FAAH, 1173 MGLL). Of these, SNVs had MAF , and were selected for RV analysis. The list and location of the RVs identified by RareCover as supporting the association is available in Supplemental Table S1.
We simulated cases and controls for a collection of sample sizes, ranging from to over individuals with equal numbers of cases and controls. The MAF for rare variants ranged from to . Throughout, we assume the disease prevalence in the population to be . The PAR for the locus was set to . The penetrance, was varied in the interval , corresponding to OR values of –. The dependence on parameters is somewhat non-trivial. To see this, note that is a lower bound on relative risk. Reducing would increase the relative risk, only making association easier. In other words if the disease incidence is low, and a causal variant is low frequency, then the presence of the causal variant is a strong indicator of disease status.
The results of the simulation are shown in Figure 2. For each choice of parameters (, and ), case-control studies were sampled as described in Methods. The data-set was tested using RareCover, collapsing, weighted-sum heuristics, as well as single-marker tests.
To enable a fair comparison, the methods were applied to randomizations of the same data-set, obtained by permuting cases and controls. The -value is similar to a False Discovery Rate calculation. The span of a typical human gene is about Kbp, and will contain about rare-variants, implying fewer than distinct windows per gene. If we assume candidate genes for a phenotype, we would have candidate windows. A -value, or FDR of could therefore be considered significant at the genome-wide level. A test score was considered significant if it was higher than each of the permutations, giving the 95% confidence interval of the p-value as . The power of a test for a specific choice of parameters is the fraction of () tests that had a significant score. Consider the sample point in Figure 2, with , and a sample of individuals. The power of RareCover is over , which can be contrasted with the low power of the weighted-sum [16], and collapsing heuristics. For any choice of parameters, RareCover shows better performance than the other methods.
Our phenotypic model differs somewhat from the one proposed by Madsen and Browning. In their model, the PAR for each causal variant is assumed to be equal, and is equal to the groupwise PAR divided by the number of causal variants. We also applied the tests to this model, using cases, and controls, and groupwise PAR values at . The power of the MB test at these values was computed to be respectively, while the power of RareCover on the data sets is (Supplemental Figure S1).
An advantage of the RareCover approach is that it does not depend upon MAF, or the density of RVs in a region. This is partly because it combines the effects of multiple associating RVs that associate, and discards the RVs that do not associate. By contrast, other methods combine all RVs, albeit with different weights. While RareCover does not recover all of the causal RVs, it always recovered a significant subset of the causal RVs in our simulations. See Figure 3, which summarizes the results for . Let correspond to the simulated, causal RVs, while corresponds to the set returned by RareCover. Thus, corresponds to the fraction of causal RVs recovered. With modest sample size, more than of the RVs are recovered, and help provide a direct interpretation of the genetic association. A somewhat unexpected aspect is that the number of causal RVs, , (and also, ) increases with an increasing sample size. For larger samples, we can recover a larger number of the low frequency variants, and the causal set has a larger mix of low frequency RVs. As we only consider RVs with MAF , the number saturates by K individuals.
The methods were also applied to the data sets provided by Kryukov et al., as explained in Methods. The cases and controls are chosen from the extremes of a population of phenotyped individuals to reflect current population cohorts. As the locus PARs are very small (–), we work with a nominal -value cut-off of . As before, power is defined as the fraction of 1000 simulations on which the test met the -value cut-off. In addition to the methods, we also plot the power of the true causal mutations to illustrate their small effect.
Figure 4 shows the results upon choosing the %, and % extremes for different levels of phenotypic association. RareCover outperforms other methods over the different tests, and is comparable to the results of selecting the true causal mutations. As suggested previously, increasing PAR values, and population sizes, increase the power of RareCover, as with all methods. However, the power of RareCover does not appear to be affected by the specifics of the demographic simulation.
The allele frequency spectrum of the CP and BRE models is shown in Figure 5. There are significant differences in allele frequency spectra in the two cases. In the CP case, there is a bias in the cases among alleles with lower frequency. High frequency causal variants represent an easier case, as they can be detected by single marker analysis. To eliminate these cases, we discarded high frequency causal variants from the simulations, which partly explains the bias in CP, relative to BRE. In the CP (respectively, BRE) models, the average number of variants per 5Kbp window was (respectively, ), with (respectively, ) causal variants. The performance of RareCover is robust against different demographic models, and depends mainly upon locus PAR, and sample size.
The running time of RareCover increases linearly with the number of SNPs, and the number of individuals, as shown in Figure 6. For a population of individuals, the running time time goes from 80ms to 311ms on a standard Linux desktop, as the number of SNPs in the window increases from to . The times shown here do not include the cost of reading and writing the data, which incurs a fixed additional cost (about ms. See Supplemental Figure S2). The total running time is at most that of a single marker test.
On the FAAH data ( individuals), the running time for a window of Kbp was computed to be seconds. Consider a genome-wide scan with windows. To achieve genome-wide significance, we would need randomizations for each window, which could be computationally intensive.
However, we run RareCover in two passes, using the Xcorr statistic on the union-variant as a filter (Figure 1). The permutation test is only applied to the fraction of the windows for which the Xcorr statistic exceeds a threshold ().
Therefore the RareCover computation is executed on windows. As discussed in the methods, . For (corresponding to in Figure 1), the total time isIf the number of candidate windows is larger (), and a conservative filter is chosen (corresponding to ), the running time increases to hrs., easily accomplished on a small cluster.
We described a novel method for Rare variant analysis with greatly improved power of detecting associations, relative to other published methods. RareCover utilizes the specific properties of RVs as compared to common variants, and applies a greedy approach to picking a subset of RVs, that best associate with the phenotype. It is a natural extension of previous methods, which either collapsed all RVs at a locus, or collapsed them after weighting different SNPs differently. Our algorithm is similar in orientation to the greedy solutions for the combinatorial problems of identifying set-cover and test-cover (See for example, Lovasz [38]). However, it is specifically designed for case-control analysis. The power of the method is extensively analyzed against different values of locus PAR, penetrance, and sample size. RareCover easily outperforms other methods which group, and collapse SNPs. The weighting approaches are reasonable, given that most causal RVs have functional significance, and likely to have moderately high penetrance, which one would not expect in a non-causal RVs. However, a large number of non-causal RVs, even with small weights, can dilute the association of the causal RVs. Also, it is difficult to identify different groupings of RVs, and to set appropriate weights for different groups.
Our application of the method on the CRESCENDO individuals, generates plausible hypotheses on the role of FAAH and MGLL in of obesity. The genetic association of FAAH with obesity is interesting because many previous studies with common variants have failed in identifying significant associations. We investigated the hypothesis that alleleic heterogeneity due to multiple RVs, influences the obesity phenotype. Second, the low LD between RVs and causal variants implies that if an RV is significantly associated, it is likely to have functional significance. Our simulations confirmed that RVs identified by RareCover were enriched in the causal RVs. In analyzing FAAH and MGLL, we identified exactly one small, functionally significant region, at each locus with significant association. This suggests that multiple rare variations help influence the regulation of the two genes. Recently, Sipe and colleagues collected metabolite expression levels on metabolites from severely obese subjects and 48 normal weight subjects [39]. Comparing against our FAAH data, we find that the levels of AEA (anandamide) are highest in obese individuals that carry an RV identified by RareCover, and lowest in individuals that are non-obese, and do not carry the causal RV. As FAAH helps metabolize AEA (anandamide), this result is consistent with the hypothesis of the RVs disrupting FAAH expression. The data on all metabolite expression will be published elsewhere (Harismendy et al., unpublished).
Nevertheless, our study also raises many methodological questions. Our approach is greedy, in that it selects the most discriminating RV at each step. Theoretically, it is possible that a collection of RVs, that are individually less discriminating, are jointly more strongly correlated. In that case, RareCover will not identify them. We implemented an approach based on simulated annealing to find an optimal subset of SNPs. However, in our simulations with the union model, the greedy method worked as well as the more complex optimization, and was significantly faster. Recall that in the simple Union model, the penetrance does not change upon inclusion of additional SNPs, but the PAR increases. Other, more complex models are possible. For example, we could have a threshold model, in which the penetrance increases with a minimum number of rare alleles. Or, we could have additive models, where the penetrance increases as a function of the number of rare alleles. As more re-sequencing data becomes available, these will be the focus of additional investigation. A second issue is that our definition of a locus is set arbitrarily as a window of fixed length, much like in other methods. However, empirical tests with a small range of window-sizes did not significantly change the results. It is possible that a dynamic assignment of the size of the locus could increase power, but at the cost of additional computations.
In this study, we analyze only the rare variants. While the RareCover algorithm can work unchanged with rare and common variants, a correct test for power of such an approach would require a biological model that combines the effect of RV and common variants. It is hard to speculate on such models in the absence of empirical data. However, preliminary results on comparing common and rare variants at the MGLL locus suggest an independent, additive effect.
GWA studies have shown that identifying the genetic basis of disease depends upon many factors. For this reason, algorithms have been devised to deal with population substructure issues, epistatic interactions between loci, as well as rare variant analysis. Our results indicate that RV analysis is useful in many contexts, and novel methods may have to be developed to include the effect of RVs in all of the above.
|
10.1371/journal.pntd.0004291 | Molecular Diversity of Trypanosoma cruzi Detected in the Vector Triatoma protracta from California, USA | Trypanosoma cruzi, causative agent of Chagas disease in humans and dogs, is a vector-borne zoonotic protozoan parasite that can cause fatal cardiac disease. While recognized as the most economically important parasitic infection in Latin America, the incidence of Chagas disease in the United States of America (US) may be underreported and even increasing. The extensive genetic diversity of T. cruzi in Latin America is well-documented and likely influences disease progression, severity and treatment efficacy; however, little is known regarding T. cruzi strains endemic to the US. It is therefore important to expand our knowledge on US T. cruzi strains, to improve upon the recognition of and response to locally acquired infections.
We conducted a study of T. cruzi molecular diversity in California, augmenting sparse genetic data from southern California and for the first time investigating genetic sequences from northern California. The vector Triatoma protracta was collected from southern (Escondido and Los Angeles) and northern (Vallecito) California regions. Samples were initially screened via sensitive nuclear repetitive DNA and kinetoplast minicircle DNA PCR assays, yielding an overall prevalence of approximately 28% and 55% for southern and northern California regions, respectively. Positive samples were further processed to identify discrete typing units (DTUs), revealing both TcI and TcIV lineages in southern California, but only TcI in northern California. Phylogenetic analyses (targeting COII-ND1, TR and RB19 genes) were performed on a subset of positive samples to compare Californian T. cruzi samples to strains from other US regions and Latin America. Results indicated that within the TcI DTU, California sequences were similar to those from the southeastern US, as well as to several isolates from Latin America responsible for causing Chagas disease in humans.
Triatoma protracta populations in California are frequently infected with T. cruzi. Our data extend the northern limits of the range of TcI and identify a novel genetic exchange event between TcI and TcIV. High similarity between sequences from California and specific Latin American strains indicates US strains may be equally capable of causing human disease. Additional genetic characterization of Californian and other US T. cruzi strains is recommended.
| Trypanosoma cruzi is a protozoan parasite that causes Chagas disease in humans and dogs and may eventually lead to mortalities related to cardiac failure. This parasite is most frequently transmitted by triatomine bug vectors, commonly called “kissing bugs.” Although Chagas disease is predominately acquired in Latin American countries, T. cruzi exists in wildlife and vectors in some parts of the United States of America (US), including regions of California. Within the US, occasional cases of locally acquired Chagas disease have been reported, and recent serological surveys indicate that T. cruzi exposure may be occurring more commonly than previously realized. However, relatively little molecular research has been performed on the T. cruzi strains present in the US, especially within California. In this study, we collected nearly 100 kissing bugs from regions of northern and southern California to determine the T. cruzi prevalence and genetic diversity for each region’s kissing bug population. We compared DNA sequences obtained in this study to those of several T. cruzi strains found in Latin America and the southeastern US. Based on our data, we conclude that Californian T. cruzi samples are closely related to strains found in Latin America known to be associated with human infections.
| Trypanosoma cruzi is a protozoan parasite that, in both humans and dogs, may cause an insidious onset of fatal cardiac disease[1]. Known as Chagas disease, T. cruzi is the most economically important parasitic infection in Latin America, where an estimated 8–9 million people are living with the chronic disease [2]. The parasite is most commonly transmitted vectorially, by numerous species of triatomine bugs (commonly called “kissing bugs”), distributed from Chile and Argentina in South America to approximately 42.5 degrees northern latitude of the United States of America [3, 4]. Only seven authochthonous clinical cases of Chagas disease in humans have been officially documented in the United States despite the fact that nine endemic Triatoma species are known to harbor T. cruzi [1]. The prevalence of infection varies among Triatoma species and across geographic regions [5, 6] and has been reported to be as high as 61% in Louisiana [7].
In the US, T. cruzi has been found in wild canids; numerous rodent species; and mesomammals such as raccoons, opossums and skunks [1]. The prevalence of T. cruzi in various wildlife species has ranged upwards from 50% in Texas and some southeastern states [8, 9]. Many of these mammals are peri-urban species that adapt well to human-modified landscapes and, if infected, can bring T. cruzi closer to humans and their canine companions. In turn, when triatomines are present in the local environment, there may be a subsequent increased risk of vectorial T. cruzi transmission to people, and an even greater transmission risk to dogs, who likely acquire T. cruzi via ingestion of infected vectors [10, 11]. In 2006, the Texas Veterinary Medical Diagnostic Laboratory reported 18.6% of 532 dogs presumably clinically ill with cardiac disease to be seropositive for T. cruzi [12]. In addition, canine serological surveys in states such as Tennessee, Louisiana and Texas indicate that T. cruzi infection is not an uncommon occurrence, even in apparently healthy domestic dogs [13–15]. Likewise, recent human serological surveys and Triatoma blood meal analyses suggest that human T. cruzi exposure may also occur more frequently than previously realized [16–18]. Physicians and veterinarians are not well-trained to recognize this disease in the US; treatment is not readily available [19]; and there are no drugs approved for veterinary use [1]. Understanding of the ecology of T. cruzi in the US, including vector and reservoir distribution, and of the molecular epidemiology of endemic strains will enable health and disease control professionals to better respond to the likely rising incidence of Chagas disease.
Trypanosoma cruzi taxonomy has been revised, with the most recent consensus classifying the organism into six subtypes or ‘discrete typing units’ (DTUs), designated TcI to TcVI [20]. Within each DTU fall numerous strains whose unique identities are generally determined via typing of several independent genetic loci. Very little T. cruzi molecular epidemiology research has been done in the US as compared to that accomplished in Latin American countries [21], despite the concern that Chagas may become an emerging disease in the country [19, 22, 23]. Most research on US T. cruzi has been restricted to typing to the DTU level, and to date, only TcI and TcIV have been detected in local vectors and wildlife [9, 24]. Researchers have recently begun to explore intra-DTU molecular diversity, focusing on isolates from the southeastern US [21]. However, data on genetic diversity in southwestern regions (e.g. California, Arizona, and New Mexico) are very limited [25].
California has the largest influx of migrants of any state in the US [26], with 53% of the immigrant population of Latin American origin [27]. Additionally, 2011 US census data indicates that more than 21% of the nearly 3 million South American migrants residing in the US live in California, with estimates of 75,000–399,000 living in Los Angeles alone [28]. It is therefore probable that many exogenous strains of T. cruzi enter California every year via human migration. It has been experimentally demonstrated that at least one virulent Honduran strain can be viable if introduced into Tr. protracta, the most common triatomine bug vector in California [29]. Thus, the pool of T. cruzi strains present in the US may potentially become more diverse. Additionally, with global climate change, it has been predicted that the human population at risk for T. cruzi transmission will increase in southern California due to increased triatomine activity associated with warmer temperatures [23]. Therefore, in addition to monitoring T. cruzi vector distribution, it is important to investigate the molecular genetics of endemic strains; how they compare to virulent strains in Latin America; and whether recently introduced strains may already exist in local vectors. To this end, the goals of this study were to: 1) compare the prevalence and DTUs of T. cruzi within triatomine bug populations from two regions of California and 2) further characterize the California T. cruzi samples via molecular genetics to assess whether there are regional differences and to determine how the California samples compare to those present in other regions of the US and Latin America.
Triatoma protracta specimens were actively collected from private residences in two study regions. All landowners consented to the collection of bugs from their properties. The first study area was located in southern California, in the town of Escondido (33.1247° N, 117.0808° W). This study site was chosen because previous research had identified T. cruzi in the resident triatomine bug population [25]. Abundant woodrat (Neotoma macrotis) nests were found, and much of the terrain was covered with large granite-based boulders and smaller rocks that provided crevices for triatomines. The second study area encompassed several residences in the town of Vallecito, situated in northern California (38.0903° N, 120.4736° W). This location was selected based on knowledge that multiple triatomine bugs collected there in 2011 were positive for T. cruzi (M. Niemela, pers comm). Woodrat nests at these properties varied by site but were generally less abundant than the Escondido location.
Black light traps were used in July and August 2012 to collect adult bugs from both study regions. Lights were turned on approximately 30 minutes before sunset and left on for at least two hours after sunset to coincide with the evening hours during which the adult bugs were flying (C. Conlan, pers comm) [30]. The bugs often did not fly the complete distance to the light trap; therefore, combing the surrounding area facilitated capture of bugs crawling on the ground nearby. This trapping method worked well in Escondido, where the trap was strategically placed at the top of a hill and the vegetation on the slope below consisted of small shrubs that did not obscure the emanating light. Light trapping was less successful in Vallecito. Hence, to augment the triatomine sample size from this region, we enlisted the help of property owners to collect bugs found in their homes. We also partially excavated several woodrat nests to obtain both adult and nymphal bugs. All bugs were placed in tubes and frozen at -20C° until laboratory processing.
In addition, we opportunistically obtained specimens from public health employees in southern California, who often received bugs from concerned citizens, especially if the bug had bitten someone within the home. These bugs were shipped to the laboratory either frozen or in ethyl alcohol during the months of April-July 2012 and June-August 2013.
The program Geneious was used for the assembly and alignment of maxicircle and nuclear sequences. Previously published COII-ND1, TR and RB19 partial sequences included in our alignments are listed in S1 Table. We prioritized the selection of the RB19 and TR isolates included in the alignment, based on the availability of sequences for both genes, overlap with isolates included in the COII-ND1 alignment when possible, as well as representation from a broad geographic range.
Two different approaches have been used to amplify the contiguous COII-ND1 maxicircle genes, which span a region of approximately 1,594 bp (CL Brener strain, GenBank #DQ343645). The first approach results in separate partial sequences for each gene, yielding short fragments of 417 bp and 369 bp for COII and ND1, respectively [42, 46]. The second approach, and the one applied in this study, generates a COII-ND1 combined partial sequence of approximately 1,272 bp in length [21, 35, 39]. The two shorter gene fragments obtained via the first method are completely nested within this longer combined sequence. The goal of our phylogenetic analyses was to maximize the number of unique T. cruzi sequences included, while ensuring that isolates represented a wide geographic range, especially within the TcI DTU. Therefore the alignment of our COII-ND1 sequences was not limited to published sequences of similar length, but also included the shorter separate gene sequences obtained in studies where the first approach was applied (S1 Table). Following alignment, all COII-ND1 sequences (n = 62) were trimmed and manually concatenated to a final length of 786 bp (369 + 417), representing the two separate gene fragments.
Phylogenetic trees for the RB19, TR and COII-ND1 gene sequences were re-constructed in MEGA6 via Neighbor-Joining (NJ) and Maximum Likelihood (ML) methods. In the NJ approach, the evolutionary distances were computed using the maximum composite likelihood method [47] with 2,000 bootstrap replicates. For the ML trees, the best fit model (as determined via the Model Test option in MEGA6) was run with 500 bootstrap replicates. The bootstrap support of the resulting NJ and ML phylogenies were compared for each genetic marker, and the best supported tree was selected. For trees displaying similar topology, both NJ and ML bootstrap values were included at appropriate nodes. In all cases, the trees were outgroup rooted with T. cruzi marinkellei.
The discovery of an apparent TcI/TcIV hybrid was further evaluated via the comparison of pairwise-distances between this sample and representative samples of TcI and TcIV for each genetic marker (i.e. T. cruzi sequences included in the reconstruction of the respective phylogenetic trees). The uncorrected p-distances were calculated in MEGA6 using pairwise deletion and transitions/transversions as the substitution types. The program Dna-SP version 5.10 [48] was used to calculate diversity indices for the TR, RB19 and COII-ND1 TcI sequences obtained in this study. The TcIV sequences were not analyzed due to their limited number. Haplotype diversity (Hd), nucleotide diversity (Pi), G+C content and the number of segregating sites (singleton + parsimony informative polymorphic sites) were calculated for all genes. The number of synonymous and non-synonymous mutations, as well as the ratio of number of nonsynonymous substitutions per site to synonymous substitutions per site (dN/dS), were calculated for the TR and RB19 genes but were omitted from the COII-ND1 analysis due to the putative RNA editing that occurs within the maxicircle gene [49].
A total of 29 triatomine bugs were collected from the Vallecito study area, of which 24 (two adults and 22 nymphs) were found within woodrat houses. The five remaining bugs were adults obtained from either light traps or within a resident’s home. All identified bugs were Triatoma protracta. The two PCR-based screening assays targeting different T. cruzi genomic loci showed a high degree of concordance in the DNA extracted from all bugs. The 121/122 kinetoplast minicircle assay was marginally more sensitive, detecting 16 positive bugs, whereas the TcZ1/TcZ2 nuclear assay only identified 15 of these same bugs as positive. Kinetoplast DNA sequences obtained from the single discordant sample confirmed the presence of T. cruzi DNA. Thus 55.2% of bugs at the Vallecito site were infected with T. cruzi.
At the Escondido study site, 53 bugs were collected, all of which were adult Tr. protracta drawn to light traps. Thirteen bugs were positive for both T. cruzi PCR screening assays; however, positive amplification was detected for an additional six bugs using only the kDNA minicircle assay. Five of these discordant samples were successfully cloned and sequenced to confirm the presence of T. cruzi DNA, yielding the conclusion that 18 bugs (34%) were T. cruzi positive at the Escondido location.
There were 15 Tr. protracta bugs submitted from public health employees in southern California, three of which were positive for parasite DNA on both screening assays (20%; there were no discordant results for screening assays among this set of bugs). With the exception of one specimen from San Diego, these bugs represented a range of locations within the Greater Los Angeles Area: Agoura Hills, Altadena, Los Angeles, Northridge, Oak Hills, Santa Clarita, Simi Valley, Tarzana, and Thousand Oaks (S1 Fig). Of the 11 bugs for which addresses were provided, area visualization via GoogleEarth revealed that the homes primarily abutted natural canyon areas designated as parks or were within housing tracts interspersed with parcels of undeveloped land. A summary of the T. cruzi positive bugs is shown in Table 2. Together these data confirm that wild populations of Tr. protracta at multiple sites in California are frequently infected with T. cruzi.
We next aimed to investigate which T. cruzi subtype/DTUs were present using lineage-specific genotyping on a subset (n = 29) of the positive bugs, as described in Fig 1. Within this subset, DTU determination was successful only for those samples that were positive for both of the T. cruzi screening assays described above (n = 22). Samples that were parasite positive only for the more sensitive 121/122 assay (n = 7) likely had insufficient DNA to amplify the lower copy number DTU gene targets. We detected 13 and 7 TcI samples from the Vallecito and Southern California locations, respectively. We found only two TcIV samples, both of which were from the Escondido location. Thus, the T. cruzi TcI and TcIV DTUs are both endemic in California.
The T. cruzi DTUs that we identified are known to contain substantial genetic diversity [35, 50–53]. We therefore generated nucleotide sequence data to investigate our sample diversity at the intra-DTU level and to enable comparison with strains from other studies. Sequences from two nuclear genes (TR and RB19) consistently classified our Californian (CA) samples into TcI (n = 10) and TcIV (n = 2) DTUs, confirming our previous genotyping results (Figs 2 and 3). The TR gene demonstrated greater sequence diversity across CA samples than did the RB19 gene: thirteen vs. two unique haplotypes identified respectively. The RB19 gene sequences for the 10 CA TcI samples were identical and indistinguishable from a single TcI sequence from an opossum isolate [38] obtained from the US state of Georgia (Fig 2). Likewise, the TcIV RB19 sequences were identical for the two CA samples (Esc19 & Esc26), as well as for two other US samples in GenBank (from a dog of unknown origin and a raccoon from Georgia). In contrast, for the TR gene, 2 to 8 single nucleotide polymorphisms (SNPs) were detected among both the TcI and TcIV-positive CA samples (Fig 3). The CA samples were distinguished from two GenBank TcIV sequences from Guatemala and Brazil by 11 to 14 SNPs. Within the TcI group, all the sequences from this study were closely related. The most closely related database sequences were from the US and northern South America (Colombia and Venezuela). Three of the southern CA sequences (Esc2, SoCal1 allele 1 and SoCal3) were identical to each other, as well as to an isolate obtained from a bug collected in the state of Florida (GenBank #AF358970) [35]. For the TR gene, the phylogenetic reconstruction between the NJ and ML trees was very similar, and bootstrap values for both trees are presented at congruent nodes (Fig 3). In contrast, the NJ and ML topology for the RB19 gene varied within major clades, and only the NJ tree is represented (Fig 2).
For the maxicircle COII-ND1 genes, the NJ and ML tree topologies were very similar within the TcI clade, but the NJ tree provided better support within the TcIV clade. We therefore present the NJ tree with both NJ/ML bootstrap values indicated at congruent nodes (Figs 4 and 5). Eleven of the twelve samples in this study were categorized as TcI based on the analysis of the maxicircle COII-ND1 genes (Table 2 and Fig 4). Phylogenetic analysis of the concatenated COII-ND1 sequences revealed that the CA TcI sequences obtained in this study were grouped with strong bootstrap support (96%) in a subclade (Fig 5, subclade 1) with other North American isolates (US and Mexico), as well as several isolates from Central America (i.e. Guatemala and Honduras). In addition, Colombian and Venezuelan isolates previously classified within either “sylvatic” or “domestic” genetic populations [42] were also included in this subclade. Thus, the composition of sequences within subclade 1 closely corresponded to that of the group described elsewhere as TcI-Dom, which contains a high proportion of TcI strains associated with human infection across the Americas [46, 51, 53].
Esc19 was the only sample classified as having a TcIV maxicircle sequence, varying by only 2–3 SNPs from the southeastern US isolates (Fig 4). Interestingly, the Esc26 COII-ND1 sequence, which was defined as TcIV via the RB19 and TR nuclear gene sequences, as well as the DTU assays, was classified as TcI and was identical to those sequences obtained for Esc2 and Esc46, both of which were typed as TcI by all other markers tested. These data are most consistent with the Esc26 sample being the product of a genetic exchange event between TcI and TcIV ancestors, leading to TcI mitochondrial introgression into a TcIV nuclear genomic background.
The p-distances presented in Table 3 highlight the genetic exchange between TcI and TcIV observed in sample Esc26. With respect to the RB19 and TR markers, Esc 26 was more closely related to TcIV than TcI by an order of magnitude. In contrast, for COII-ND1, the reverse finding was apparent. Table 4 provides values for the diversity indices calculated in Dna-SP. As seen in the phylogenetic analyses, no diversity was observed within the TcI sequences for the RB19 gene, whereas the TR and COII-ND1 genes are more genetically diverse.
The overall T. cruzi prevalence of 55.2% (16/29) in the Vallecito triatomine population is the highest reported for Tr. protracta. This infection level is comparable to that found in Triatoma gerstaeckeri, a US triatomine species that was implicated in a 2006 case of acute Chagas disease acquired in Texas [6]. Furthermore, if only the adult Tr. protracta Vallecito specimens collected in this study are considered, the T. cruzi prevalence increases to 71.4% (5/7). In contrast, this study’s prevalence of T. cruzi in adult Tr. protracta in Southern California (34% for Escondido specifically; 27.9% across all Southern California samples) is consistent with previous infection levels for this species, which have ranged from 20–36% in Southern California [25, 54]. The only published case of locally acquired human Chagas disease in California occurred in Tuolumne County [55, 56], just south of Calaveras County where our Vallecito study site was located. Our data show that Tr. protracta populations in both northern and southern California have high frequencies of T. cruzi infection, indicating that the risk for transmission to people and domestic animals is widespread in these regions.
In this research, we successfully amplified T. cruzi DNA of both mitochondrial (COII-ND1) and single-copy nuclear genes (TR and RB19) directly from Tr. protracta DNA extracts. Direct testing is commonly done for T. cruzi screening purposes using highly sensitive assays that target genes possessing thousands of copies (i.e. minicircle kinetoplast targets) or nuclear tandem repeat regions. However, to our knowledge, previous research on the COII-ND1, TR and RB19 genes have only used DNA extracted from pure T. cruzi cultures [21, 36, 40, 42]. Thus it is valuable to note that these assays are sensitive enough for analysis of triatomine bug extracts that have tested positive for T. cruzi via the TcZ1/TcZ2 assay.
We found greater genetic diversity in Escondido, where both TcI and TcIV DTUs were present as compared to Vallecito where only TcI was detected. Of particular interest is the determination that one of the Escondido samples (Esc26) belonged to TcIV based on DTU and nuclear phylogeny analyses, but that this same sample was grouped with TcI isolates based on the mitochondrial maxicircle COII-ND1 sequence analysis. To our knowledge, this is the first report of T. cruzi possessing TcIV nuclear genes and TcI mitochondrial genes. The reverse incongruency, first reported by Machado and Ayala [35] and interpreted as evidence of rare genetic exchange events, has been documented in multiple opossum TcI T. cruzi isolates in the southeastern US [21].
In the only other molecular typing study performed on T. cruzi in California [25], samples were obtained from Tr. protracta specimens collected from two southern sites, including the same suburban property in Escondido from which our samples were collected. In this earlier research, Hwang et al. [25] used the D71/D72 primer set employed in our DTU algorithm to obtain partial sequences of the 24sα ribosomal RNA gene for two samples. Although the authors did not identify these two samples as TcI, alignment of their two sequences (GU594186 & GU594187) with sequences from three of our TcI Escondido samples (KT879367-KT879369) differed by only 2–3 SNPs. It can therefore be inferred that the T. cruzi samples obtained in the earlier research were also TcI.
With the exception of two samples, our T. cruzi sequences were all classified as TcI. By contrast, in the southeastern states, Roellig et al. [21] identified twice as many TcIV as TcI isolates. However, this finding may simply reflect host sampling bias and culture success, as isolates were cultured and examined from 21 raccoons versus only nine opossums. In the southeast, raccoons are predominately infected with TcIV, whereas opossums have only been found to be infected with TcI [1, 21]. In California and southwestern states, woodrats are the presumed primary reservoir of T. cruzi. Research in Texas found that southern plains woodrats (Neotoma micropus) were hosts to both TcI and TcIV [9], and it is therefore not surprising that both DTUs were present in our Escondido study site where woodrat nests were extremely abundant. Our study represents the most northerly site of T. cruzi genotyping to date and therefore extends the known range of TcI. The fact that we did not detect TcIV in our northern site may be due to our smaller sample size in this region or reflect a true absence of this DTU. A geographical distribution for TcIV that extends as far north as Calaveras County remains a possibility; indeed skunks (Mephitis mephitis) and raccoons, known hosts of TcIV are present in this region and would be interesting to target for future research.
Our phylogenetic analysis of the concatenated maxicircle genes indicated that the TcI haplotypes from this study clustered within a subclade containing isolates from Venezuela, Colombia, Central America (Guatemala and Honduras) and North America. Included in this subclade were strains from Venezuela (strains 10462P2C3 and 11541) representing a domiciliated T. cruzi genetic population (VENdom/TcIdom) that has been associated with human infections [42, 46]. In a phylogenetic analysis based on nine concatenated maxicircle sequences, the TcIdom strains were similarly nested among North and Central American TcI strains [46]. The authors suggested that these results provided evidence for initial human contact with TcI occurring in North-Central America, with subsequent southerly movement of TcIdom as early colonizing Amerindians migrated south. However this previous analysis included T. cruzi sequences from only four US isolates, all of which originated in the southeastern US. Thus the placement of all our CA TcI samples in the same COII-ND1 subclade as TcIdom strains provides further phylogenetic support for the close relationship of US TcI isolates to TcIdom and so strengthens the conclusions proposed by these authors.
The gene sequences examined in this study only represent a fraction of the T. cruzi maxicircle and nuclear DNA. Genome scale analysis would enable a better understanding of the genetic relationships between strains found in the US and Latin America. Genetic analyses of US T. cruzi sequences obtained in this study and a limited number of US isolates included in other studies [42, 46] have revealed that US TcI haplotypes are similar, or even identical, to partial gene sequences of several Latin American T. cruzi strains associated with human illness. Therefore, although some researchers have questioned whether local T. cruzi strains are as infective or virulent as strains found in Latin America [5, 57, 58], our data suggest there is little genetic basis for considering US strains to have any particular unique characteristics that would distinguish them from Central American or most domestic South American TcI strains. In fact, US T. cruzi strains have been demonstrated to infect and cause clinical symptoms and pathology in dogs [59–61], non-human primates [62, 63] and humans [56, 64], including a Texas prisoner who contracted acute Chagas disease as a result of an unethical experimental study conducted in 1940 prior to the identification of the first case of autochthonous human Chagas disease in the US [65]. Nevertheless, there are few documented cases of locally acquired human Chagas disease in the US, and it has been proposed that the apparent rarity is most likely due to the infrequency of triatomine colonization within US homes [1, 5, 66] and reduced vector-transmission efficiency [1].
In some regions of Latin America, rural houses are constructed of materials (i.e. adobe brick walls, thatch roofs) that facilitate the invasion and exploitation of breeding niches by triatomines [67, 68]. In contrast, US standard housing construction likely hinders triatomines from establishing breeding colonies once adult bugs have entered homes [69]. However, it is conceivable that outdoor pet enclosures or substandard housing might be subject to colonization events if these structures were adjacent to natural areas with triatomine bug populations, especially if preferred breeding sites and sylvatic hosts had recently diminished. For example, adult Tr. protracta have been reported to disperse into human residences following the destruction of woodrat nests associated with construction activities [70]. Likewise, following environmental changes in Louisiana spurred by the wake of hurricane Katrina, infestations of Tr. sanguisuga in human dwellings may have been related to the triatomines’ search for new bloodmeal sources [7].
Emphasis has been placed on T. cruzi vector-human transmission being less efficient in the US due to the delayed defecation habits of some US triatomine bug species [1]. Yet experimental studies conducted on immobilized mice have demonstrated that two southwestern US species of triatomines, Tr. protracta and Triatoma rubida, may defecate upon repletion, that is, immediately after terminating a blood meal and disengaging from the host [54]. The relevance of these experimental studies with respect to the natural feeding behavior of Tr. protracta in human homes is not clear; however, given that 75% (6/8) of the observed Tr. protracta defecations occurred either before or upon repletion, the risk of T. cruzi vector transmission by this triatomine species cannot be disregarded. Furthermore, although the close ecological associations of some US triatomine species (i.e. Tr. protracta and woodrats) suggest host feeding preferences, bloodmeal analysis has indicated that vectors are not truly host-specific. In studies performed in California and Arizona, human blood in addition to other host species (e.g. chickens, pigs and wildlife), were identified from three triatomine species: Tr. protracta, Triatoma recurva and Tr. rubida [17, 71]. Thus, the risk of vectorial T. cruzi transmission from Tr. protracta and other US triatomine species is likely to be greater than has been previously assumed.
In fact, recent T. cruzi screening of blood donors identified 16 asymptomatic cases of T. cruzi infection most likely acquired from local vector transmission in the US [16]. Data extrapolated from this four-year screening study led the authors to conclude that vectorial transmission is not common within this country. However, one must consider the derivation of these data, which is potentially subject to large sampling bias and thus may not represent the general population. Specifically, T. cruzi screening was performed on asymptomatic blood donors who presumably felt healthy at the time of blood donation. In contrast, it is probable that people who feel unwell (some of whom may have chronic Chagas disease) do not choose to donate blood. If these people do not seek health care, or if their physicians are unaware of the local risk of Chagas infection, then the disease would go undiagnosed. Currently, Chagas disease is a reportable disease in four states [72], with Texas just having recently listed Chagas as reportable in late 2012 [73]. As these states and others begin to more closely track the incidence of human T. cruzi infections, the true risk of vectorial transmission should become more apparent.
Our research is one of only two molecular studies on T. cruzi in California and the first to investigate this parasite’s genetic diversity in the northern portion of the state. While this study’s prevalence of T. cruzi in Tr. protracta populations in southern California (~30%) was similar to earlier findings, an even higher prevalence was detected in our northern California study region. The genetic markers employed in this study allowed us to demonstrate the close similarities between T. cruzi strains in California and those present in other US states, as well as some Latin American countries. Thus, vectors across California present a clear transmission risk to humans and dogs. Additionally, experimental studies have already proven that Tr. protracta can sustain a Honduran T. cruzi isolate [29], and Triatoma infestans and Rhodnius prolixus, two vectors from Latin America, have been shown capable of harboring US T. cruzi isolates [4]. Although the divergence and migration of T. cruzi strains occurred over a period of millions of years [50, 53, 74], in this new era of global connectivity, vectors and the pathogens they carry may unknowingly be transported between countries [75, 76]. Therefore if triatomine vectors efficient in T. cruzi transmission, and perhaps more readily able to colonize human homes, were to be unwittingly introduced to the US, potential mixing of T. cruzi strains could occur within and among vector species.
Despite the fact that T. cruzi has been known to exist in the US for at least 80 years, only four states consider Chagas disease to be reportable. In late 2012, Texas became the fourth state to declare Chagas a reportable disease, a decision preceded by a series of in-state studies and clinical case reports on the disease in both canines and humans. Consistent with Texas, our research implies that some areas of California may have a similar risk for T. cruzi transmission and suggests that California physicians and veterinary practitioners should consider Chagas disease as a potential cause of cardiac illness in regions where Tr. protracta populations are evident.
|
10.1371/journal.pgen.1002423 | Substitutions in the Amino-Terminal Tail of Neurospora Histone H3 Have Varied Effects on DNA Methylation | Eukaryotic genomes are partitioned into active and inactive domains called euchromatin and heterochromatin, respectively. In Neurospora crassa, heterochromatin formation requires methylation of histone H3 at lysine 9 (H3K9) by the SET domain protein DIM-5. Heterochromatin protein 1 (HP1) reads this mark and directly recruits the DNA methyltransferase, DIM-2. An ectopic H3 gene carrying a substitution at K9 (hH3K9L or hH3K9R) causes global loss of DNA methylation in the presence of wild-type hH3 (hH3WT). We investigated whether other residues in the N-terminal tail of H3 are important for methylation of DNA and of H3K9. Mutations in the N-terminal tail of H3 were generated and tested for effects in vitro and in vivo, in the presence or absence of the wild-type allele. Substitutions at K4, K9, T11, G12, G13, K14, K27, S28, and K36 were lethal in the absence of a wild-type allele. In contrast, mutants bearing substitutions of R2, A7, R8, S10, A15, P16, R17, K18, and K23 were viable. The effect of substitutions on DNA methylation were variable; some were recessive and others caused a semi-dominant loss of DNA methylation. Substitutions of R2, A7, R8, S10, T11, G12, G13, K14, and P16 caused partial or complete loss of DNA methylation in vivo. Only residues R8-G12 were required for DIM-5 activity in vitro. DIM-5 activity was inhibited by dimethylation of H3K4 and by phosphorylation of H3S10, but not by acetylation of H3K14. We conclude that the H3 tail acts as an integrating platform for signals that influence DNA methylation, in part through methylation of H3K9.
| DNA methylation is a common feature of eukaryotic genomes. Methylation is typically associated with silenced chromosomal domains and is essential for development of plants and animals. Although the control of DNA methylation is not well understood, recent findings with model organisms, including the fungus Neurospora crassa, revealed connections between modifications of histones and DNA. DNA methylation is dispensable in Neurospora, facilitating genetic studies. Isolation of mutants defective in DNA methylation revealed that a histone H3 methyltransferase, DIM-5, is required for DNA methylation. DIM-5 trimethylates H3K9, which is then recognized by Heterochromatin Protein 1 (HP1), which recruits the DNA methyltransferase DIM-2. We investigated the possibility that H3 provides a platform to integrate information relevant to whether the associated DNA should be methylated. Indeed, we found that DIM-5 is sensitive to methylation of H3K4 and phosphorylation of H3S10. Our analyses further revealed that H3K14 is critical in vivo, but not because acetyl-K14 inhibits DIM-5. We also found that H3R2 is essential for DNA methylation in vivo but not important for DIM-5 activity. Interestingly, we found H3 mutants that show recessive defects in DNA methylation and others with dominant effects. We also defined a set of H3 mutations that are lethal.
| The primary structures of histones, the small basic proteins that are complexed with DNA to form chromatin in eukaryotes, are highly conserved but not invariant [1], [2]. For example, comparisons between a sea urchin and the filamentous fungus Neurospora crassa reveal that the two most highly conserved histones, H3 and H4, have 16/135 and 9/102 amino acid differences, respectively [3], [4]. Pioneering studies with the yeasts Saccharomyces cerevisiae and Schizosaccharomyces pombe revealed that mutation, or even deletion, of some histone residues is not lethal, allowing for genetic studies of structure/function relationships of these prominent proteins [5]–[9].
Histones are subject to a variety of posttranslational modifications including methylation, acetylation, phosphorylation, ubiquitylation, sumoylation, as well as ADP-ribosylation, deimination and proline isomerization [1], [10]. An explosion of findings has implicated these modifications in various fundamental cellular processes, including transcription, alternative splicing, replication, chromosome condensation, recombination, DNA repair and DNA methylation [1], [10], [11]. One approach to investigate potential involvement of modifications of a particular histone residue is to test for effects of substitutions that prevent the modification or mimic the modified state. For example, because acetylation of lysines reduces the positive charge of this residue, it can be informative to test the effect of substituting lysines with non-modifyable residues that are either neutral (e.g. glutamine) or positively charged (e.g. arginine) [12]. To gain insight into residues involved in DNA methylation and other cellular processes, we carried out a systematic analysis of the heavily modified N-terminus of N. crassa histone H3 [13].
Neurospora is an excellent organism to study effects of histone H3 mutations in part because the wild-type genome contains a single H3 gene, hH3 [3], [4]. A direct link between H3 modifications and DNA methylation was originally demonstrated in Neurospora when DNA methylation was found to depend on trimethylation of histone H3 lysine 9 (H3K9me3) [14], [15]. In this organism, DNA methylation is primarily found at sequences that are products of a genome defense mechanism called RIP (Repeat Induced Point mutation), which is triggered by sequence repeats [16], [17]. Duplicated sequences are efficiently detected and mutated during the sexual phase of the Neurospora life cycle in the period between fertilization and nuclear fusion. Both copies of duplicated sequences are peppered with numerous C:G to T:A mutations, rendering the resulting sequences AT-rich, and the remaining cytosines are typically methylated [18], [19]. A single DNA methyltransferase (DIM-2), guided by Heterochromatin Protein 1 (HP1) [20], [21], is responsible for all known DNA methylation in Neurospora [22]. HP1 binds to H3K9me2 and even more tightly to H3K9me3, the predominant form found in Neurospora [23], [24]. Mutation of the dim-5 gene, which encodes an H3K9 methyltransferase responsible for most if not all H3K9me3 in Neurospora, results in complete loss of DNA methylation [14]. DIM-5 is part of the DCDC (DIM-5/-7/-9/Cul4/DDB1 Complex), and depends on DIM-7 for recruitment to future heterochromatin domains [25]–[27]. All components of the DCDC are required for normal H3K9 methylation [26]. It is also known that methylation of some regions requires dephosphorylation of H3S10 by PP1 [28]. The current study was carried out to decipher which residues of H3 play a role in DNA methylation. We expected to identify interactions with DIM-5 as well as interactions with other components of the methylation machinery, such as HP1. In plants [29]–[31] and animals [32], H3K9 methylation also directs some DNA methylation, raising the possibility that our findings will provide insights into shared mechanisms operating in a variety of organisms.
Initial experiments demonstrated that introduction of an ectopic H3 gene bearing substitutions at lysine 9 (e.g. hH3K9R) can cause dominant loss of DNA methylation and can reactivate a methylation-silenced transgene, hphme [14]. Structural studies demonstrated intimate contacts between the H3 tail (residues A7-G12) and the DIM-5 backbone [33], suggesting that additional residues may be important for DIM-5 activity and DNA methylation. We wished to test both for possible dominant effects, as initially detected for K9 substitutions [14], and for possible recessive effects on DNA methylation, H3K9me and DIM-5 activity. We therefore first transformed a strain harboring the methylation-silenced transgene (hphme) to screen for substitutions that cause a dominant loss of DNA methylation. We next constructed substitution strains to reveal recessive mutations. Our current results suggest that multiple H3 tail residues, or their modifications, are involved in DNA methylation. We found both recessive and dominant effects.
Roughly half of the approximately 40 amino acid residues in the highly conserved N-terminal tail of histone H3 are subject to covalent modifications (Figure S1) [1], [2], [10]. To improve our understanding of the role of H3 in DNA methylation, we carried out mutational analyses of these and other residues of potential interest, such as those around lysine 9 (Figure 1A). Acetylatable lysines were changed to arginines (K→R) or to glutamines (K→Q) to simulate the hypoacetylated or acetylated states, respectively. Lysines and arginines that might be methylated were changed to leucines (K/R→L), and potentially phosphorylated serines and threonines were changed to alanines (S/T→A) to prevent modification. We also tested for a constraint on the spacing of residues by inserting an extra glycine at GG12,13 (+G13). We took advantage of a simple genetic test for effects on DNA methylation. Mutant alleles were introduced into a hygromycin-sensitive reporter strain, N644, in which the hygromycin phosphotransferase gene, hph, was reversibly silenced by DNA methylation [34]. In our initial experiments, we co-transformed N644 with the mutant constructs along with a dominant selectable marker, BenR, which confirms resistance to benomyl [14]. Loss of hph methylation caused by dominant or semi-dominant effects of the hH3mutant alleles would render the strain resistant to hygromycin (Figure 1B). Although co-tranformation is extremely efficient in Neurospora (typically 50–90% of transformants incorporate the non-selected DNA), representative transformants were tested to verify that the H3 construct had been integrated in a fraction of the tranformants that did not produce hygromycin resistant strains (data not shown).
We found that substitutions of any residue between positions 8 to 14 caused hygromycin resistance (Figure 1A, Table 1 and Table S1). Similarly, insertion of an extra glycine at GG12-13 (+G13) caused hygromycin resistance. Transformation experiments were repeated 4–8 times with similar results (Table 1). Some substitutions, notably G12P and R2L, gave variable or weaker hygromycin resistance; transformations with constructs bearing these substitutions only gave rise to hygromycin resistant colonies in ∼50% of the transformations. Overall, phenotypic analyses of transformants suggested that silencing depends on R2 and residues around K9 (R8 to K14) and that the spacing between residues around GG12-13 is important.
The methylated hph allele is somewhat unstable; its reversion frequency is ∼10−5 [25], [34]. To confirm that H3 mutations caused loss of DNA methylation, we directly examined the methylation status of genomic DNA by Southern hybridization using 5-methylcytosine-sensitive restriction endonucleases. Blots were probed for several genomic regions that are typically methylated [18]. DNA from the parental strain (N644) and from a DNA methyltransferase mutant (dim-2) served as positive and negative controls, respectively, and DNA from several independent transformants were analyzed for each hH3 construct. Consistent reductions in DNA methylation were observed for all transformants with S10A, T11A, G12P, +G13, G13M, K14R, K14Q, K9L and K9R mutations (Figure 1C, Figure S2A and S2B, Table 1 and Table S1). Curiously, about half of the hygromycin-resistant colonies that came from the R8A transformations exhibited loss of DNA methylation. None of the few hygromycin-resistant colonies that were obtained with the T6A, A15M, R17L, K18R, K18Q and K23R constructs showed changes in DNA methylation. In the case of R2L, although hygromycin-resistant colonies were obtained in half of the trials, only one of twelve mutant strains analyzed showed a significant reduction in DNA methylation. This variability might result from differences in copy number and/or chromosomal location of the integrated mutant alleles. Indeed, we found that some hygromycin-resistant transformants had single ectopic integrations of hH3mutant genes while others had multiple integrations (Figure S2C). Multiple integrations might result in a dose-dependent loss of DNA methylation. Of course, hygromycin resistance may also result from unknown genetic or epigenetic effects. To address these possibilities, we developed a system to construct strains with single copies of mutant alleles inserted at a particular location in the genome, as described below.
H3 substitutions may cause loss of DNA methylation through effects on site recognition or catalysis by the histone methyltransferase, DIM-5. Alternatively, H3 substitutions might affect the DNA methylation pathway downstream of H3K9me. In one approach to distinguish between these possibilities, we tested the in vitro activity of DIM-5 on various H3 substrates. We constructed modified forms of a GST-H3 (residues 1–57) fusion corresponding to the amino acid replacements that we tested in vivo. The fusion proteins were incubated with DIM-5 and radio-labeled methyl-group donor (S-adenosyl methionine), fractionated by SDS-polyacrylamide gel electrophoresis and tested for incorporation of 3H-methyl groups by fluorography (Figure 2A). As expected from our previous finding that H3K9 is the only target for DIM-5 [14], the K9L substitution abolished DIM-5 activity. The R8A and G12P mutations also completely inhibited DIM-5 activity. The S10A and T11A mutations greatly reduced, but did not abolish, DIM-5 activity. Slightly reduced activity was observed with the G13M protein in most assays performed. In contrast, K4L, A7M, +G13, K14R, A15M and P16A mutations had no significant effect on DIM-5 activity. The K14R mutation was not inhibitory in vitro while this change caused striking hygromycin resistance and loss of methylation in vivo (Figure 1). We conclude that the effects of mutations in H3 residues 8–12 are likely due to direct effects on DIM-5/H3 interactions while the inhibitory effects on DNA methylation of K14R is presumably due to some unknown downstream effects (e.g. on HP1 or DIM-2 action).
The H3 tail is subject to a variety of post-translational modifications (Figure S1) that could potentially play a role in controlling DIM-5 action. Indeed, there is precedence for effects of modifications on the activity of histone methyltransferases related to DIM-5. In particular, SUV39H1, SETDB1 and Clr4 have been shown to be sensitive to phosphorylation of H3S10 (H3S10ph) [35]–[37]. We were most interested in the possibility that H3K4me, a mark common in euchromatic regions, might interfere with DNA methylation in Neurospora. In addition, we wished to test the effect of acetylation of H3K14 (H3K14ac) because this modification is apparently countered by histone deacetylase 1, which is important for DNA methylation in Neurospora [38], and because substitution of K14 inhibited DNA methylation in vivo but did not noticeably affect DIM-5 activity in vitro (Figure 1A, 1C and Figure 2A). We therefore tested the effect of methylation of H3K4, acetylation of H3K14 and phosphorylation of H3S10 on in vitro DIM-5 activity with peptides covering residues 1–20 of H3 (Figure 2B). As expected, a peptide that had a substitution at the target residue, H3K9, completely abolished DIM-5 activity. Robust activity was found with a H3K14ac peptide. This contrasts results obtained with the H3K9 methyltransferase SETDB1 [37] but is consistent with findings with the SUV39H1 and Clr4 H3K9 methyltransferases [35], [36]. Essentially no DIM-5 activity was detected with H3S10ph or H3K4me2 peptides, whereas a peptide bearing the K4L substitution showed full activity. H3K4me2 did not affect H3K9 methylation by SETDB1 or Clr4 [36], [37]. Our results support the idea that H3K4me2 and H3S10ph impact DNA methylation by preventing DIM-5 activity. In contrast, the importance of H3K14 presumably results from an effect downstream of H3K9 methylation by DIM-5.
In theory, Neurospora transformants of strain N644 may have resulted either from ectopic insertion of hH3mutant genes or from gene replacement events. In practice, only ectopic insertions were obtained, and most transformants had single integrations (Figure S2C), which is unusual for transformation of Neurospora. This raised the possibility that transformants bearing exclusively, or even predominantly, altered H3 alleles may not be viable. To test this possibility, and to investigate if some H3 substitutions might be recessive and thus show defects that were masked by the simultaneous presence of wild-type H3, we developed a system to test hH3 constructs as single copies at a defined genomic location, in the presence or absence of a wild-type allele. The system takes advantage of a non-functional hH3 allele at the endogenous locus (hH3RIP1) that we generated by RIP when we crossed two strains homozygous for an hH3S10E allele at the his-3 locus [28]. The resulting hH3RIP1 strain with hH3S10E at his-3 was crossed with transformants bearing other hH3 substitution alleles integrated at his-3 to test whether progeny with substitution alleles at his-3 would be viable in the absence of a functional allele (Figure 3A; Text S1). For most of our substitutions, including those at R2, A7, R8, S10, A15, P16, R17, K18 and K23, we were able to build such (hH3RIP1; hH3mutant) strains. Thus the entire cellular pool of H3 carries the relevant substitution in these strains. Although viable, all these strains showed defects in vegetative development (Figure 3B, Figure S3) and were female sterile (data not shown). For substitutions of other residues (K9, G12, G13, K14, K27, S28 and K36), we were only able to obtain strains with the substitution allele at his-3 when the wild-type allele was present at the native locus (hH3WT; hH3mutant), suggesting that these mutations are lethal in the absence of wild-type H3. We also found evidence that substitutions of K4 and T11 are not tolerated in the absence of a wild-type allele (data not shown). Although it was conceivable that some or all of these mutations simply affected protein stability, their semi-dominant effects argued against this.
Analyses of DNA methylation in the targeted H3 mutant strains was informative (Figure 3). Where comparisons could be made with results obtained with the strains bearing one or more copies of mutant constructs at undefined sites, the strains with precisely one wild-type and one mutant H3 gene gave similar results (Figure 3C and Figure S4; Table S4). For example, mutation of K14 gave equivalent results with both experimental schemes. [Incidentally, it is noteworthy that the extent of loss of DNA methylation caused by substitutions in K14 was similar to that caused by substitutions in K9 even though the K14 mutations, unlike the K9 mutations, did not substantially interfere with methylation by DIM-5 in vitro (Figure 2) or in vivo (data not shown)]. We did observe some variability between experiments and among methylated regions tested (Table S4). R8A, which caused a partial (∼25%) reduction in DNA methylation in cotransformation experiments (Figure 1 and Figure S2), showed comparable reduced methylation at the 2:G12 region when the mutant allele was at the his-3 locus but did not show significant loss of methylation at 8:A6 or 2:B3. The K9L mutation, which resulted in clear loss of methylation in cotransformants at all regions tested also showed obvious reduced methylation at 8:A6 and 2:B3 but not at 2:G12 when targeted to his-3. Substitutions of S10 with A or E resulted in partial loss of DNA methylation at all sites examined but S10G did not cause loss of methylation at any region tested, when there was a wild-type H3 gene in the same strain. Mutations of R2, A7, A15, K18, K23, K27 and S28, which did not reveal loss of silencing in the cotransformation experiments also showed no loss when the mutant allele was targeted to his-3 in the presence of the native H3 gene. Mutations of P16 and R17, which also showed no effect in the cotransformation experiments, caused modest reduction of DNA methylation, but only in the 2:B3 region. In contrast, substitution of residues G12 showed significant loss of methylation in the contransformation experiments but not in the targeted transformants and mutation of G13 showed a modest reduction of methylation in the latter experiments and a greater reduction in the contransformants.
The results summarized above unequivically indicated that a number of mutations of the H3 gene are semi-dominant. Interestingly, we also found several mutant constructs that showed dramatic effects on DNA methylation only when they provided all of the H3, i.e. they were recessive. The most striking results were obtained for the R2L mutation. Although this change showed little, if any, effect in the presence of normal H3 (Figure 1 and Figure 3), the mutant allele, when alone, caused a nearly complete loss of DNA methylation (Figure 3C, Figures S4 and S5) Interestingly, methylation of H3 K9 and HP1 binding appeared normal (Figure 4), suggesting that R2 plays a role downstream of these events.
The A7M and R8A substitution alleles also showed marked reductions in DNA methylation alone, but little or no reduction when present along with the wild-type allele (Figure 3C). Of these two mutants, R8A gave the stronger effect, consistent with its marked inhibition of H3K9 methylation in vitro (Figure 2). We confirmed that the R8A substitution strain, but not the A7M strain, also shows greatly reduced H3K9me3 and mislocalization of HP1 in vivo (Figure 4).
We took advantage of the single-copy status of the histone H3 gene in Neurospora to explore the possible involvement of particular amino acid residues of H3 in DNA methylation. The study was facilitated by a methylated hph allele that allowed detection of partial loss of DNA methylation. Previous work revealed that transformation of the reporter strain with alleles of hH3 bearing substitutions in H3K9 caused almost complete loss of DNA methylation, while retaining a wild-type allele of hH3, i.e. the mutations were semi-dominant, as if the altered H3 “poisoned” the methylation machinery [14]. This may reflect a requirement for the HP1 dimer to bind pairs of methylated H3K9 residues within one nucleosome or between adjacent nucleosomes [39]. Failure to isolate transformants bearing multiple copies of the mutant allele further suggested that a preponderance of H3 with this substitution might not be tolerated by Neurospora [14].
Here we explored the generality of these observations by extending our analysis to include all residues in the N-terminal tail that are thought to be subject to post-translational modification [1], [10]. In addition to testing various residues with our system pioneered for K9, we also developed a scheme to test directly for viability of strains bearing mutant alleles in the absence of a wild-type copy of hH3 [28]. This approach identified a number of residues that are essential for viability and revealed nonlethal mutations that resulted in recessive DNA methylation defects (Table 2). The most striking recessive defects were due to R2L, A7M, R8A and S10G substitutions. S10A and S10E showed semi-dominant defects that became much more pronounced in the absence of the wild-type allele while some other changes, including P16A and R17L, showed modest defects either in the presence or absence of the wild-type allele. Most, but not all, residues identified as important for DNA methylation in vivo were also important for DIM-5 histone methyltransferase activity in vitro. In particular, R8, K9, S10, T11, G12 and G13 were required for H3K9me3 by DIM-5 in vitro, while substitutions in every residue from A7 through R17 caused loss of DNA methylation in vivo. We also found that the spacing between residues can be important since introduction of an extra glycine at GG12-13 caused a loss of DNA methylation. Finally, we found that DIM-5 is sensitive to certain H3 modifications (H3K4me2 and H3S10ph) but not to others (H3K14ac).
Our results provide clues to the role of histone H3 in the control of DNA methylation. No comparable study has been carried out in a system with DNA methylation. Extensive mutational analyses of histones have been carried out in the budding yeast S. cerevisiae [6]–[8], however, and more limited studies have been carried out in the fission yeast S. pombe [9], which unlike S. cerevisiae, sports H3K9 methylation. We found that point mutations in K4, K9, T11, G12, G13, K14, K27 and S28 could not be tolerated by Neurospora in the absence of a normal allele. Interestingly, unlike yeasts that have been examined, K27 is subject to methylation in Neurospora, but this modification is not essential (K. Jamieson and E. Selker, unpublished observation). Substitutions in K4, K9, S10 and K14 are tolerated in fission yeast [9], [40] although they do interfere with heterochromatin function, as in Neurospora. A curious result from studies in budding yeast is that effects of histone mutations appear to depend to some extent on strain background [7]. Our observation that a mutant allele containing an extra residue (+G13) is tolerated only in presence of a sheltering wild-type allele implies that the proper spacing between H3 N-terminal tail residues is critical for viability. This provides a possible explanation for the observed lethality of a deletion of residues 4 through 10 in budding yeast, even though substitutions in individual residues from 1 through 39 were tolerated [7].
Phosphorylation of T3, T6, S10, T11 and S28 in the H3 tail has been found to be required for chromatin-dependent processes including gene silencing in several systems [28], [41]–[47]. Based on the crystal structure of DIM-5, the presence of a phosphate on H3S10 should prevent the interactions between the hydroxyl group of S10 and residues in DIM-5 that are essential for optimal catalytic activity (Y283 and D209) [33]. The expected importance of the dephosphorylated state for H3K9 methylation by DIM-5 was supported by findings with strains deficient in the H3S10 phosphatase PP1; an increase in global levels of H3S10ph caused reduced levels H3K9me3 in regions of the genome that loose DNA methylation [28]. This was supported by the observed complete loss of H3K9me and DNA methylation in hH3S10A, hH3S10G, hH3S10E substitution strains (Figure 3C, Figures S4 and S5 and [28]). Information from other systems suggests that phosphorylation of the neighboring residue, T11, may be also important. H3T11ph is associated with chromosome condensation during mitosis and meiosis in plants [48]. In mammals, phosphorylation of H3T11 by PRK1 in response to androgen-dependent stimulation accelerates demethylation of H3K9me3 by the JMJD2C demethylase and activates transcription of target genes [43]. Protein kinase Chk1 also phosphorylates H3T11 and its dissociation from chromatin in response to DNA damage causes a decrease in H3T11ph and release of the H3K9 acetyltransferase GCN5, leading to transcriptional repression [42]. We found that replacement of this residue with alanine caused a semi-dominant loss of DNA methylation (Figure 1C and Figure S2). The main chain carbonyl group of H3T11 forms a hydrogen bond with the DIM-5 backbone (Q285) and the alanine substitution would disrupt this interaction, presumably resulting in the observed loss of DIM-5 activity in vitro (Figure 2A and [33]). H3T11ph, like H3S10ph, may impact H3K9 methylation and DNA methylation
It is interesting that substitutions of H3K14 caused a semi-dominant loss of silencing and DNA methylation, especially considering that neither mutation nor acetylation of this residue appreciably affected DIM-5 activity in vitro (Figure 2; [49]). Similarly, mammalian SUV39H1 and fission yeast Clr4 are not affected by H3K14ac [35], [36]. One possibility is that H3K14 is important for an activity downstream of methylation of H3K9 by DIM-5. In Neurospora, H3K9me3 is recognized by HP1, which then directly recruits the DNA methyltransferase DIM-2 to the heterochromatin domains [20], [21]. The fact that the H3K14 mutations can only be studied while effectively heterozygous confounded attempts to determine whether HP1 binding is reduced but it is interesting that we found little if any change in H3K9 methylation and HP1 binding in the K14Q mutant (Figure S6). It is noteworthy, that acetylation of this K14 does not affect binding of the S. pombe HP1 homologue, Swi6 [50], in vitro while mutation of H3K14 in S. pombe causes mislocalization of Swi6 [9]. Alhough H3K14ac does not affect H3K9 methyltransferase activity in vitro, it remains possible that it interferes with the establishment of H3K9 methylation in vivo. In both S. pombe and N. crassa, the H3K9 methyltransferases are components of multi-protein complexes, the CLRC and DCDC, respectively [25], [26], [51], [52]. Although there are important mechanistic differences between these two complexes, K14 mutations may influence H3 interaction with proteins that guide the H3K9 methyltransferases. The possible importance of H3K14ac in regulation of heterochromatin formation is consistent with the observed requirement of histone deacetylases (HDACs) for DNA methylation in N. crassa and for heterochromatin formation in S. pombe [38], [50], [53]–[55].
An important new finding from our study is that mutation of H3R2 disrupted DNA methylation. Information from other organisms suggests that this arginine is one of several in H3 that are subject to methylation, namely R2, R8, R17 and R26 [56]. Thus it will be of interest to determine if methylation of one or more of the arginines in the H3 N-terminal tail (Figure S1) impact methylation of H3K9 and DNA. Substitutions of both R8 and R17 (hH3R8A and hH3R17L) caused semi-dominant loss of DNA methylation, whereas substitution of R2 (hH3R2L) caused recessive loss of DNA methylation (Figure 3C, Figures S4 and S5). Presumably this reflects different roles of these residues in heterochromatin formation. The structure of DIM-5 complexed with an H3 peptide shows intimate contacts between R8 and the DIM-5 backbone [33]. Methylation of R8 is expected to prevent these interactions, consistent with the lack of DIM-5 activity in vitro on a substrate with the R8A substitution (Figure 2A and [49]) and with the observations on strains bearing this substitution, namely reduced H3K9me3, mislocalization of HP1 and loss of DNA methylation (Figure 4A and 4B). In contrast to the situation with R8, R2 lies outside of the region bound by DIM-5 (A7-G12) [33] and substitutions in R2 have little effect on DIM-5 activity in vitro [49]. Conceivably, the loss of DNA methylation observed in the R2L mutant is mediated by H3K4 methylation. It is known that H3R2me2 indirectly blocks H3K4me3 in both budding yeast and mammals [57]–[59]. Though H3K4me2 inhibits DIM-5 activity in vitro (Figure 2B and 2C), we observed wild-type levels of H3K9 methylation in the R2L mutant in vivo and apparently normal localization of HP1 (Figure 4). These observations suggests that the recessive loss of DNA methylation caused by R2L substitution results from disruption of a step downstream of H3K9 methylation and HP1 binding.
Our finding that DIM-5 activity is inhibited by methylation of H3K4 (Figure 2B) raises the possibility that effects on DNA methylation may reflect effects on methylation of H3K4. Indeed, methylated forms of H3K4 and H3K9 appear mutually exclusive in the Neurospora genome [19]. The mammalian H3K9 methyltransferase G9a is similarly sensitive to methylated forms of K4 [60] but, surprisingly, other H3K9 HMTases including S. pombe Clr4, mammalian SETDB1 and dSU(VAR)3-9, have been reported to methylate H3 peptides bearing K4me2 [35], [36], [61]. The establishment of H3K9me in these organisms apparently requires the concerted activity of various H3K4 demethylases, specifically SU(VAR)3-3 in Drosophila and Lid2 in fission yeast [62], [63]. In mammals, the DNA methyltransferase regulator DNMT3L directly interacts with the H3 tail and this interaction is abolished by methylation of H3K4me [64]. Additionally, Cfp1, a component of the H3K4 methyltransferase complex Setd1, binds unmethylated CpGs and helps establish H3K4me3 domains free of DNA methylation [65], [66].
In summary, our observations in Neurospora suggest that H3 amino-terminal tail residues and their covalent modifications regulate methylation of H3K9, binding of HP1, and one or more downstream events required for the establishment, maintenance and propagation of DNA methylation. Strains generated in this study should be useful to elucidate the role of H3 in other processes in Neurospora including the gene silencing mechanisms, RIP, meiotic silencing and quelling, as well as in other chromatin-based processes including transcription, DNA repair, recombination and the genomic arrangement of histone variants.
Strains used in this study are listed in Table S2. Standard procedures were followed to grow strains and to perform crosses [67]–[69]. Primers used in this study are listed in Table S3.
Genomic DNA was prepared as described previously [70], [71] from strains grown with shaking in Vogel's medium N with appropriate supplements at 32°C for 3 days. For Southern blotting, approximately 1 µg DNA was digested overnight with 3–5 units of the desired restriction enzyme, fractionated on 0.8 or 1.0% agarose gels and transferred to nylon membrane [72]. Hybridizations were performed with probes prepared by random hexamer priming [73], as previously described [74].
DIM-5 assays were carried out in a volume of 20 µl at 10°C for 1 h with 5 µg of the appropriate recombinant GST-H3 variant (H3 residues 1–57) and 2.75 µCi of S-adenosyl [methyl-3H]-L-methionine (NEN), as previously described [14]. The reaction products were fractionated on SDS-PAGE gel (16%, acrylamide/bis, 29∶1) and incorporation of the radioactive methyl group was analyzed by fluorgraphy using ENTENSIFY (DuPoint). To test the effect of covalent H3 modifications or amino acid substitutions on DIM-5 activity, similar assays were performed with 0.5 µg peptide substrates (modified or unmodified H3 residues 1–20) [15]. The reaction products were analyzed by fluorography or precipitation with 20% TCA, filtration (Millipore GF/F filter), washing and liquid scintillation counting [75].
Nuclei were isolated as previously described [76], but with minor modifications. The following enzyme inhibitors were added to all buffers: 1 mM sodium butyrate, 1 µM Trichostatin A, 1 µM PMSF, 3 mM DTT, 10 mM sodium fluoride, 1 mM sodium vanadate, 0.1% phosphatase inhibitor cocktail (Sigma, P2850) and 1 µM each of leupepsin, pepstatin and E-64. Western blotting was performed with ∼100 µg of nuclear protein as described previously [15]. The following antibodies were used: α-H3K4me2 (Upstate, 07-030), α-H3K9me3 [15] and α-H3 C-terminal (Active Motif/LP Bio, AR-0144-200). Modifications on histones are represented according to the nomenclature proposed by Turner [77]. All antibodies were used at a dilution of 1∶5000. Horseradish peroxidase (HRP)-conjugated goat antibody against rabbit IgG was used to detect antibody-peptide complexes by chemiluminescene (Thermo Fischer Scientific Inc, USA).
Dilute suspensions of vegetative spores (conidia) were germinated on solidified Vogel's N medium with 1.5% sucrose and the required supplements for 2 hrs at 30°C. Square pieces of agar with the germinating conidia were cut out from plates and placed on glass slides. Agar pieces were flooded with a few drops of liquid Vogel's N medium and then overlayed with coverslips [78]. Bright-field and fluorescence images were collected on a Zeiss Axioplan 2 Imaging System with a EBQ 100 isolated light source, Endow GFP (S65T) filter (excitation 470, emission 525) and Plan-APOCHROMAT 100×/1.46 N.A. objective. Images were processed with Axiovision (version 4.6.3) and Adobe Photoshop CS (version 8) software.
|
10.1371/journal.ppat.1005304 | Evolutionary Analyses Suggest a Function of MxB Immunity Proteins Beyond Lentivirus Restriction | Viruses impose diverse and dynamic challenges on host defenses. Diversifying selection of codons and gene copy number variation are two hallmarks of genetic innovation in antiviral genes engaged in host-virus genetic conflicts. The myxovirus resistance (Mx) genes encode interferon-inducible GTPases that constitute a major arm of the cell-autonomous defense against viral infection. Unlike the broad antiviral activity of MxA, primate MxB was recently shown to specifically inhibit lentiviruses including HIV-1. We carried out detailed evolutionary analyses to investigate whether genetic conflict with lentiviruses has shaped MxB evolution in primates. We found strong evidence for diversifying selection in the MxB N-terminal tail, which contains molecular determinants of MxB anti-lentivirus specificity. However, we found no overlap between previously-mapped residues that dictate lentiviral restriction and those that have evolved under diversifying selection. Instead, our findings are consistent with MxB having a long-standing and important role in the interferon response to viral infection against a broader range of pathogens than is currently appreciated. Despite its critical role in host innate immunity, we also uncovered multiple functional losses of MxB during mammalian evolution, either by pseudogenization or by gene conversion from MxA genes. Thus, although the majority of mammalian genomes encode two Mx genes, this apparent stasis masks the dramatic effects that recombination and diversifying selection have played in shaping the evolutionary history of Mx genes. Discrepancies between our study and previous publications highlight the need to account for recombination in analyses of positive selection, as well as the importance of using sequence datasets with appropriate depth of divergence. Our study also illustrates that evolutionary analyses of antiviral gene families are critical towards understanding molecular principles that govern host-virus interactions and species-specific susceptibility to viral infection.
| Evolutionary analyses have the potential to reveal not only biochemical details about host-virus arms-races but also the nature of the pathogens that drove them. Primate MxB was recently shown to restrict the replication of primate lentiviruses, including HIV-1. However, we find that positive selection in primate MxB is incongruent with known molecular determinants of lentiviral restriction. This suggests that MxB has antiviral activity against a broader range of viruses than is currently appreciated. We also identified multiple losses of MxB in mammals, as well as rampant recombination between Mx paralogs, which has distorted gene orthology. Our study illustrates the importance of evolution-guided functional analyses of antiviral gene families.
| Ancient, pathogenic viruses have played a major role in shaping the extant host innate immune repertoire. Understanding how pathogen-driven evolution has shaped host-virus interfaces can reveal insights into the molecular basis of cross-species transmission, including human susceptibility to zoonoses [1]. The genetic signature of diversifying (positive) selection distinguishes many host antiviral genes, indicating their involvement in a long-standing genetic conflict with viral pathogens. Such genetic conflict has also driven gene copy number expansion in several mammalian antiviral genes, allowing further diversification of pathogen defense. For instance, the TRIM5 antiviral gene is present in one copy in primates but has expanded to 6–7 copies in mice and other mammals [2,3]. Similarly, primates encode seven members of the APOBEC3 antiviral gene family whereas mouse genomes encode only one [4,5].
Unlike the TRIM5, APOBEC3 or other antiviral gene families, the copy number of myxovirus resistance (Mx) gene appears to be relatively static, with two copies in both primate and mouse genomes. Mx proteins are interferon-inducible dynamin-like large GTPases. They are composed of a highly conserved GTPase domain (GD), which is connected to a helical stalk by a hinge-like bundle-signaling element (BSE) [6]. Previous work has shown that human MxA and both murine Mx1 and Mx2 proteins have broad and potent activity against a diverse range of RNA and DNA viruses [7,8]. In contrast, the antiviral activity of human MxB appears to be much more narrow, only recently having been shown to restrict HIV-1 and other primate lentiviruses [9–12].
In this study we employed a detailed evolutionary approach to address the basis for the apparent stasis of Mx gene copy number and the discrepancy in antiviral breadth of Mx homologs. Using maximum-likelihood approaches, we found strong evidence of diversifying selection in the N-terminal region of primate MxB genes in contrast to the previously observed diversifying selection in loop L4 of MxA [13]. Surprisingly, signatures of MxB diversifying selection do not overlap with previously-mapped lentiviral-restriction determinants. We therefore conclude that simian lentiviruses have not driven the rapid evolution of primate MxB. Our analysis instead suggests that MxB plays a central and conserved role in the interferon response to a broader range of pathogens than is currently appreciated. Extending our analysis to other mammalian Mx genes, we find that multiple, lineage-specific exchanges have occurred between Mx paralogs throughout mammalian evolution. These gene conversion events have led to both the preservation of key enzymatic and structural features of Mx GTPases, as well as the acquisition of new antiviral specificity via the complete conversion of MxB-like genes to a MxA-like state. Our findings highlight the impact of diversifying selection and gene conversion on the functional repertoires of antiviral gene families.
We wished to investigate whether the primate MxB gene has been subject to pathogen-driven diversifying selection. A previous analysis reported that the primate MxB gene had not evolved under diversifying selection, even though it did find evidence for positive selection for some individual sites (see below) [14]. In contrast, our previous analysis of primate MxA found strong evidence of positive selection at both the gene and codon level [13]. Although this variance could reflect genuine differences in selective pressures that have acted on the two paralogs, we also considered the possibility that lower sampling of MxB sequences in the previous report may have led to reduced power to detect selection [15].
We, therefore, cloned and sequenced MxB from 21 hominoid, Old World monkey and New World monkey species for a total of 32 MxB sequences after including sequences from public databases. Maximum likelihood tests were implemented using the PAML [16] and HyPhy [17] suite of programs to detect whether rates of non-synonymous changes (dN) exceeded synonymous changes (dS) (dN/dS > 1 implies positive selection). Recombination can yield false signatures of positive selection [18]. We used a genetic algorithm for recombination detection (GARD) [19,20] to show that Mx genes did indeed undergo recombination during their evolutionary history. We therefore carried out selection analyses only on MxB gene segments for which evolutionary history was determined to be uniform by GARD (Fig 1A). For all analyzed GARD segments, we found strong evidence for positive selection in primate MxB (M7 vs. M8, P < 0.001) (Fig 1A). Our findings support a role for MxB in a long-standing and recurrent host-virus conflict during primate evolution.
Our analyses also revealed six codons in primate MxB that showed strong evidence of diversifying selection (model M8, Bayes Empirical Bayes (BEB) posterior probability (PP)>0.95) (Fig 1A and 1B). Four of these six rapidly evolving residues are located in the disordered N-terminus of MxB (amino acids 1–83) (Fig 1B). The MxB N-terminus determines both its localization to the nuclear pore as well as its antiviral specificity against simian lentiviruses [14,21–24]. The positive selection in MxB contrasts with the positive selection in MxA, which is concentrated in the loop L4 (S1B Fig) [13]. Although our previous analysis also found evidence of positive selection in the MxA N-terminus, we identified no sites in the loop L4 of MxB in primates that had a high posterior probability of having evolved under diversifying selection (Fig 1, S1A Fig). Our findings appear to be at odds with a recent analysis of MxB evolution in mammals, which concluded that positive selection is centered on the MxB L4 [25]. However, this previous analysis was based on a broad range of mammals and did not account for the possibility of MxA-MxB recombination, which might have confounded the analysis (see below). Despite the high similarity of MxA and MxB GTPases [26], our finding that diversifying selection is centered on distinct surfaces (L4 versus N-terminus), together with their divergent cellular localization (MxA is cytoplasmic whereas MxB is nuclear) [22,27,28], suggests that different pathogens have uniquely shaped MxA and MxB antiviral surfaces during primate evolution.
The N-terminus of MxB has been shown to be essential for its antiviral activity against lentiviruses [14,21–24,29]. We considered whether simian lentiviruses could be responsible for driving positive selection in primate MxB. If so, we would expect that the amino acids that govern anti-lentiviral specificity of MxB would also be the amino acids that are under diversifying selection as has been previously observed for APOBEC3G and TRIM5 [30–33]. We therefore compared the residues that evolved under diversifying selection with two regions in the MxB N-terminus previously identified as molecular determinants of MxB anti-lentivirus activity (Fig 1C), a triple-arginine motif (RRR11-13), and residues 37–39. To our surprise, we found no overlap between either of these molecular determinants and positively selected sites. Mutation of the RRR11-13 motif abolishes MxB anti-lentiviral activity [21]. Moreover, anti-HIV-1 activity can be conferred on the highly diverged canine MxB by restoring the RRR11-13 motif [21]. We find that the triple arginine motif arose in the common ancestor of simian and prosimian primates and has since been strictly conserved, except for in New World monkeys in which it has degenerated multiple times (Fig 1C). Residues 37–39 are known to dictate MxB's differential activity against different lentiviruses. Specifically, the group O HIV-1 chimeric CMO2.5 strain is sensitive to human but not African green monkey (AGM) MxB. Similarly, the HIV-1 P207S capsid mutant is sensitive to rhesus macaque but not AGM MxB. Functional differences between AGM and rhesus macaque MxB map to N-terminal residues 37–39, especially residue 37 [14]. We find no evidence that recurrent diversifying selection has acted on either the RRR motif or residues 37–39 (Fig 1). Formally, target recognition of the lentiviral capsid may span a broader region of the MxB N-terminus although alanine-scanning mutagenesis of these sites did not appear to disrupt anti-HIV-1 activity [21]. Given that Old World monkeys are the primary primate lineage infected by lentiviruses, we also restricted our analysis to Old World monkey sequences; analysis of only these species might therefore be expected to better reveal lentiviral-driven selection. Again, we found positive selection in MxB but not in residues shown to confer lentiviral specificity (Fig 1). Likewise, if we exclude Old World monkeys and analyze only hominoids and New World monkeys, we still observe positive selection (Fig 1). Analysis of New World monkeys alone is also weakly suggestive of positive selection in the GARD 2 segment (Fig 1), but our analysis (only 7 NWM species) lacks the statistical power for confident interpretation of this result; a denser sampling of this clade is needed. Thus, despite the occurrence of intense diversifying selection in the MxB N-terminus, our results strongly imply that the only known antiviral activity of MxB, i.e., towards primate lentiviruses, cannot explain the rapid evolution of primate MxB.
Our analysis of positive selection in MxB is discrepant with two previous findings, which reported that (A) MxB has not evolved under positive selection in primates [14], and that (B) the sites under positive selection in mammalian MxB are predominately located in loop L4 [25]. Here, we explore the sources of these discrepancies, which may help inform future analyses of antiviral genes.
Busnadiego et al. [14] performed three NSsites tests to detect positive selection in primate MxB, one of which was significant. However, the single significant analysis (model 0 vs. 3) tests variability in the ω ratio among sites and does not constitute a test of positive selection [34]. In contrast, both analyses that are designed to detect positive selection (i.e., model 1 vs. 2, or 7 vs. 8) were not significant. However, it should be noted that Busnadiego et al. based their analysis on 12 MxB sequences, of which 3 were macaque species and 5 were hominoids, making this dataset relatively shallow. Previous work has demonstrated that species representation strongly influences the robustness and inference of positive selection analyses [15]. In addition, although ten sites were reported to be under positive selection by both REL (posterior probability > 0.9) and M3 (probability > 0.99); (the former is a valid test of positive selection), false positives may arise from the analysis of shallow datasets [35]. Thus, the difference in our ability to detect positive selection in primate MxB is the result of the increased number and diversity of MxB sequences in our analysis wherein we included 32 primate species dispersed throughout the phylogenetic tree.
Sironi et al. [25] also identified positive selection in MxB based on an analysis of 29 eutherian mammals. However, while our analysis identifies the MxB N-terminus as the “hotspot” of diversifying selection, they identified the MxB L4 (Fig 1). It is formally possible that primates have experienced a unique evolutionary history relative to other eutherian mammal lineages, which might explain the incongruence between our results. However, the conclusion that MxB L4 has been a target of selection in mammals should be tempered for two reasons: i) Large divergences, such as those found across 29 mammals, are prone to dS saturation leading to the underestimation of dS (i.e., false positives) [34]. ii) The failure to account for recombination can also lead to false positives [18]. Recombination was not detected in their dataset despite our finding of recurrent recombination in a similarly sampled analysis (see below). Although we do not know the exact set of sequences assayed in Sironi et al., we note that 7 of 31 species listed in their S1 Table encode pseudogenized or gene converted copies of MxB. Our analyses highlight the merit of dense sampling within mammalian orders compared to broad sampling across orders, which is better suited to identify conserved features [36].
Many antiviral gene families have undergone dynamic, lineage-specific changes in copy number, presumably as a mechanism for gaining new antiviral specificities without losing existing functions [37]. In contrast to other antiviral gene families, previous studies have found that Mx gene copy number is relatively static [25,38]. However, this apparent stasis may be misleading. For instance, a recent report found that Mx genes have been lost in Odontoceti cetaceans (toothed whales) [39]. To more comprehensively determine the evolutionary dynamics of Mx genes in mammals, we performed phylogenetic analyses of mammalian Mx paralogs from at least one representative of all sequenced mammalian orders. We found shared synteny of the Mx locus throughout terrestrial vertebrates (Fig 2A). Both human and mouse genomes encode two Mx genes, but as previously described [38], rodents have lost the MxB-like gene and instead encode two MxA orthologs (Mx1 and Mx2). Platypus genomes encode two Mx genes that both appear ancestral to the eutherian MxA and MxB lineages; we term them here MxAB1 and MxAB2 (Fig 2). We were unable to identify any Mx genes in any of three marsupial genome sequences (tammar wallaby, opossum, Tasmanian devil) and suggest that Mx gene(s) were lost from the common marsupial ancestor. We therefore infer that MxA and MxB genes diverged at or just prior to the origin of the eutherian mammal lineage.
Within eutherian mammals, our phylogenetic analyses revealed surprisingly poor resolution; many nodes have less than 50% bootstrap support and some discordance at well-supported nodes of the mammalian phylogeny (Fig 2B) [40]. These discrepancies are not entirely unexpected for rapidly evolving antiviral genes, and likely reflect complex evolutionary histories of Mx genes in multiple mammalian lineages (see below). Nevertheless, we were able to conclude that most eutherian mammalian genomes encode both MxA-like (orthologous to human MxA) and MxB-like (orthologous to human MxB) genes. Further, elephants encode two closely related intact MxA genes in addition to MxB; this MxA duplication appears to have occurred since divergence from the sister order Macroscelidea (e.g., elephant shrew) (Fig 2). With the exception of the loss of both MxA and MxB genes in toothed whales [39], we found evidence for an intact MxA gene in all surveyed eutherian mammal species. In contrast, we found that MxB has been lost at least three additional, independent times in Rodentia, Felidae and Xenarthra (Fig 2 and S2 Fig). Therefore, MxB loss has been tolerated on multiple occasions during eutherian mammal evolution despite the fact that its importance as an antiviral factor has been experimentally demonstrated.
Phylogenetic and synteny analyses allowed us to propose a hypothetical scenario for the loss of the MxB gene in mouse. We found that the ancestral configuration of the mammalian Mx locus was Bace2(+);Fam3B(+);MxB(+);MxA(+);Tmprss2(-) (Fig 2A). In the rabbit genome (Lagomorpha, an outgroup to Rodentia), an inversion occurred in the Mx locus such that MxB and Fam3B have opposite orientations relative to the ancestral locus (Fig 2A). Distinct rearrangements appear to have taken place in Rodentia, represented by squirrel and mouse genomes, leading to a Bace2(+);Mx1(-);Fam3B(-);Mx2(+);Tmprss2(-) configuration (Fig 2A). Intriguingly, squirrel Mx2 is an MxB-derived pseudogene, whereas mouse Mx2 is an MxA-derived intact gene (S2 Fig). Based on this, we propose that mouse Mx2 originated as a result of complete gene conversion by Mx1 (MxA-like) (Fig 2A), possibly preceded by loss of the ancestral MxB-like gene in rodents.
We further investigated gene conversion of rodent Mx genes using GARD [19,20] to identify putative recombination breakpoints in a multiple sequence alignment. We found that recurrent gene conversion has occurred between Mx genes throughout rodent evolution such that various Mx gene segments have distinct, incongruous evolutionary histories (Fig 3A and S3 Fig). We used PHYML to generate bootstrapped phylogenies of each segment identified by GARD; this analysis confirmed that different segments have different phylogenies, with key nodes indicative of recombination supported by strong bootstrap values. For example, a phylogeny based on GARD segment A indicates that for the majority of rodent species the Mx1 and Mx2 genes are more closely related to each other than to the orthologous gene in related species (Fig 3B, white circles). We also use mVISTA [41] to determine that gene conversion tracts can extend into intronic sequences (S3 Fig). The phylogenetic grouping of mouse-like rodents also suggests that more ancestral conversion events have occurred. We estimate that at least eight independent gene conversion events have occurred in this region alone between Mx1 and Mx2 genes in the eight surveyed rodent species (indicated at specific nodes in Fig 3B). However, given that our sampling is limited to sequenced genomes, this is likely an underestimate of the degree of gene conversion/recombination during rodent Mx evolution (S3 Fig). Thus, a high frequency of gene conversion events has scrambled the phylogenetic relationships between rodent Mx genes.
We next extended our survey of possible gene conversion between Mx genes beyond rodents to other mammalian genomes. We found at least two Mx gene segments have distinct evolutionary histories among eutherian mammalian Mx paralogs (Fig 4). In contrast to the nearly complete gene conversion/recombination between rodent Mx paralogs (Fig 3 and S3 Fig), in other mammals recombination is more localized (S4 Fig). For instance, the regions that coincide with the second coding exon of human MxA and MxB are remarkably similar to each other (S4 Fig). These exons correspond to the Mx bundle-signaling-element BSE α1B (amino acids 84–116 in human MxA) and GTPase domain α1G (amino acids 117–147) (S1 and S4 Figs). α1G contains the highly conserved P-loop (G1), which is an essential GTP-binding element that interacts with the α- and β-phosphates of bound nucleotide [42].
Similarly, we also found that within primate, carnivore and ungulate mammals, the BSE-GTPase domain-encoding segments (GARD segment A, Fig 4) from MxA and MxB cluster with each other, instead of by species; independent phylogenetic analyses support GARD's finding of recombination with high bootstrap support. Based on these observations, we conclude that there has been recurrent gene conversion/recombination between the BSE-GTPase domain-encoding exons of MxA and MxB genes that occurred early after the separation of the different mammalian orders (Fig 4B, black circles). As in rodents, there have been additional recent exchanges in some mammalian lineages (Fig 4B, white circles). We note that this frequent intergenic recombination is the likely cause of the apparent poor resolution of the Mx gene phylogeny in mammals (Fig 2B). Therefore, gene conversion of the BSE-GTPase domains has contributed to the genetic exchange of critical functional elements between the two Mx genes over different evolutionary timeframes.
In our study, we have found that despite the apparent stasis of Mx gene copy number, both rapid evolution and recurrent gene conversion have led to a highly diversified complement of Mx genes in mammals. We find that MxB, like MxA, has evolved under positive selection, which argues that it has had a broader role in antiviral defense than is currently appreciated. MxB has been lost in some lineages, whereas in others, it has been converted into an MxA-like gene. Moreover, recurrent recombination has exchanged important enzymatic and structural motifs between MxA and MxB. Thus, both pathogenic pressure and recombination between paralogs has shaped the extant specificity of Mx antiviral proteins in different extant mammalian genomes. Discrepancies between our findings and previous studies of MxB evolution [14,25] provide a practical illustration of several important considerations when performing evolutionary studies: (a) it is necessary to use sufficiently dense phylogenetic sampling to have the power to detect positive selection; (b) recombination can confound evolutionary history and must be accounted for in studies of selective pressures; and (c) use of species sets that are too diverged can cause problems due to difficulties in accurately estimating synonymous rates.
Molecular arms races between simian lentiviruses and their infected primate hosts have been ongoing for at least 5 million years [31,32]. Evidence for lentivirus-driven evolution has been identified in other restriction factors [32] indicating that ancient, simian lentiviruses have imposed a dominant selective pressure on primate antiviral genes. Despite the coincidence of rapid evolution being centered on the MxB N-terminal tail, we find no evidence of diversifying selection having acted on either the RRR11-13 lentiviral-restrictive motif or residue 37, the two known molecular determinants of MxB anti-lentivirus activity. Previously, comprehensive triple-alanine-scanning mutagenesis of MxB’s N-terminal domain (amino acids 1–91) effectively ruled out any other determinants of antiviral activity against HIV-1 [21]. We therefore fail to find evidence that primate lentiviruses have driven diversifying selection in primate MxB. However, since only a subset of primate lentiviruses have been directly tested for MxB restriction, it is possible that some of the diversifying selection we have mapped might correspond to either primate lentiviruses that have not been tested, or to ancient lentiviruses that no longer exist.
On the other hand, since MxB anti-lentivirus activity does not require GTP binding or hydrolysis [9,10], if lentiviruses were the driving target of primate MxB we might have expected MxB GTPase motifs to have degenerated. Contrary to this expectation, we find that all GTP binding motifs are strictly conserved in all surveyed mammals that encode an intact MxB gene (S1C Fig). Preservation of the MxB GTPase domain may either imply a house-keeping function as previously suggested [43], or an antiviral requirement for GTPase activity against other (non-lentiviral) pathogens, as is seen with MxA [44,45]. The recurrent loss of MxB in various mammalian lineages suggests that an important housekeeping role for MxB is either unlikely or would have to be highly lineage-specific.
Despite the ability of MxB to inhibit primate lentiviruses, its evolution is inconsistent with a recurrent “arms race” scenario with that group of pathogens. The absence of overlap in MxB diversifying selection with previously-mapped lentiviral restriction determinants contrasts with other restriction factors such as TRIM5α and TRIMcyp, in which genetic innovation directly correlates with capsid-binding and antiviral restriction [33,46–50]. It is possible that the selective pressure exerted on lentiviruses by MxB in vivo may be weak compared to other restriction factors, thereby reducing the likelihood or intensity of an MxB-capsid arms race. In single round or spreading infections, MxB mediates 5–20 fold restriction [9,10], lower than the >100-fold restriction demonstrated for APOBEC3G, TRIM5α, and TETHERIN [12,51–53]. Another possibility is that the site of MxB restriction in capsid does not allow for viral escape. This possibility seems unlikely since MxB-resistant capsid mutants can be readily selected in vitro [14,54]. On the other hand, though MxB-escape mutant viruses might arise readily, it is possible that they are rendered more susceptible to interaction with other capsid-binding host restriction factors (e.g., TRIM5alpha, TRIMcyp). If this is true, MxB may play a crucial albeit indirect role in constraining capsid evolution as part of a multifaceted interferon response. Finally, it is formally possible that the molecular determinants of MxB anti-lentivirus activity do not mediate direct binding to lentiviral capsid but instead mediate its interaction with a cellular cofactor, leading to purifying rather than diversifying selection.
Our findings are reminiscent of previous studies in which we found diversifying selection of the primate Bst-2/Tetherin restriction factor at nef- but not vpu-interacting sites [55]. While such analyses cannot irrefutably prove that nef drove diversifying selection of primate Tetherin, it strongly argues that vpu did not. Using a similar rationale, we conclude that a different lineage(s) of viruses, distinct from currently known simian lentiviruses, have shaped MxB evolution in primates. We considered whether LINE-1 retrotransposons may have driven MxB evolution based on the recent discovery that they are also restricted by MxB [56]. However, LINE-1 restriction is independent of MxB's N-terminal tail [56], so it is unlikely that a conflict with retroelements drove diversifying selection in the N-terminus of primate MxB. Instead, we hypothesize that the evolutionary signatures of diversifying selection in MxB N-terminus, together with the evolutionary constraint acting on its GTPase domain, implicates its action against a widespread family (or families) of as-yet-identified pathogens. Interestingly, the signature of diversifying selection distinguishes the mode of MxA and MxB antiviral specificity. Just as positive selection in MxA is centered in the loop L4, which mediates its target recognition [13], we predict that individual changes in the MxB disordered N-terminus may also define MxB antiviral specificity.
MxA (cytoplasmically localized) and MxB (nuclear pore localized) proteins restrict different classes of viruses [38], which likely explains their stable retention as paralogs in most mammals. In light of this fact, it is surprising that gene conversion has resulted in the loss of MxB in rodents, followed by retention of an additional MxA-like gene in some. Interestingly, mouse Mx1 and Mx2 appear to have subfunctionalized, diverging in their antiviral range by localizing to different cellular compartments (cytoplasmic and nuclear) [22,27,28]. Therefore, the conversion of MxB to MxA in some rodents allowed for the refinement of MxA-like antiviral activity against cytoplasmic and nuclear replicating viruses, even at the expense of ancestral MxB-like activity. Although mouse Mx1 is nuclear, its localization is distinct from MxB (human MxB localizes to the nuclear pore whereas mouse Mx1 forms discrete nuclear foci). Furthermore, the evolutionary acquisition of a C-terminal NLS on Mx1 is unique to mouse-like rodents [57]. Unlike rodents, however, pseudogenization of MxB in other lineages (e.g. armadillo, felids and squirrel) does not appear to be have been driven by MxA diversification. Instead, these may reflect cases in which MxB-targeted pathogens went extinct, relaxing selective pressure to maintain MxB activity. A similar loss of constraint may also explain the dual loss of both MxA and MxB in toothed whales [39].
Rodents appear to be distinct from other mammals in their Mx gene conversion profiles. In other eutherian mammals, including primates, gene conversion between MxA and MxB appears to be largely restricted to a region encoding parts of the BSE and GTPase domains. Such localized gene conversion could be the result of an uncharacterized recombination hotspot. However, it is also possible that this is the indirect result of selective constraint. For example, a swap of the BSE-GTPase domain is least likely to deleteriously impact the antiviral repertoire of the Mx paralogs. As a result, gene conversions spanning the BSE-GTPase domain, but not other domains, might be more easily tolerated. In contrast to this ‘tolerated conversion’ model, it is also possible that the frequent gene conversion of the BSE-GTPase domain is instead directly favored by selection. We speculate that gene conversion may serve as an adaptive mechanism (akin to diversifying selection) to escape antiviral antagonism. For instance, if the MxB BSE-GTPase domain were the direct target of viral antagonism, replacing it with the diverged but functionally equivalent MxA BSE-GTPase domain might be a rapid means to escape antagonism without compromising function. In support of this latter idea, there are residues under diversifying selection in the GTPase domain of Mx proteins in primates [13], which might be the result of a genetic conflict to escape viral antagonists.
Many antiviral gene families have undergone extensive, lineage-specific copy number variation, reflecting distinct bouts of pathogen-driven evolution and highlighting genetic variation in the antiviral repertoires of even closely related species. At first glance, Mx antiviral genes appear to belie the highly dynamic nature of antiviral gene copy number variation. Instead, closer examination reveals the highly dynamic evolution of Mx antiviral genes, with both diversifying selection having driven changes in presumed viral specificity domains and recombination homogenizing the catalytic domains of Mx proteins. Due to both these evolutionary forces, there is little likelihood of Mx ‘orthologs’ maintaining functionally analogous antiviral repertoires. At the very least, our analysis raises the need for caution in functional assignments of Mx genes based on the established “Mx1" and "Mx2” nomenclature, especially for rodents. Indeed, similar whole or partial gene conversion events are likely to have shaped many mammalian antiviral multigene families in which paralogs are present in genetic proximity to each other [2,4,58].
Primate fibroblasts were purchased from the Coriell Cell Repository. Cells were cultured in DMEM (Gibco) supplemented with 10% FBS and 5% penicillin/streptomycin. RNA was isolated from cultured primate fibroblasts using the RNeasy kit (Qiagen). RT-PCR to obtain primate MxB cDNA sequences was performed using HiFi one-step RT-PCR (Invitrogen) with the following primers:
Hominoid
F 5’—ATGTCTAAGGCCCACAAGCCTTGG—3’
R 5’—GTGGATCTCTTTGCTGGAGAATTGACAGAGTG—3’
Old World monkey
F 5’—ATGTCTAAGGCCCACAAGTCTTGGC—3’
R 5’—RTGGATCTCTTCGCTGGAGAATTGACAGAG—3’
New World monkey
F 5’—ATGTCTAAGGCCCACAKGTCTTGGCC—3’
R 5’—CTCTGTAAATTCTCCAGTGAAGRGATCCACGACTACAAAGACGACGACAAATGA—3’
Gel extracted products were directly sequenced (Sanger method). Contigs were assembled using CodonCode Aligner. Detailed information on primate species and cell lines used in this study can be found in S2 Table. Sequences have been deposited in GenBank (accession numbers KT698228-KT698252).
Mx genes were retrieved from publically available databases (non-redundant nucleotide collection, reference genomic sequences, high-throughput genomic sequences and whole-genome shotgun contigs) using BLASTN or TBLASTN from at least one representative of all sequenced mammalian orders (S1 Table). After including sequences obtained through RT-PCR, multiple sequence alignments were conducted using CLUSTALW or MUSCLE, and adjusted manually. To determine the evolutionary relatedness of Mx genes, maximum-likelihood phylogenies (1000 replicates) were constructed using PhyML under the GTR substitution model either locally or through the phylogeny.fr website [59,60]. Analysis of Mx locus synteny was evaluated in Ensembl using the Comparative Genomics Alignment tool for standard reference genomes or by retrieving genomic sequences and using BLASTN analyses to determine order of Mx homologs and flanking genes.
We examined alignments for evidence of recombination using the GARD algorithm [20] from the HyPhy package [17], run on local computers. Briefly, for each in-frame alignment to be analyzed, we first used HyPhy's NucModelCompare to suggest the best-fitting nucleotide substitution model (using as input a maximum-likelihood tree generated from the entire alignment using PhyML [59] and the GTR model). We then supplied the alignment and best-fitting substitution model to GARD, using the general discrete model of site-to-site rate variation with 3 rate classes. Breakpoints assigned by GARD were independently assessed by generating bootstrapped (n = 1000) maximum-likelihood phylogenies in PhyML.
Maximum-likelihood tests were performed with CODEML implemented with the PAML software suite [16]. Input trees were generated from each alignment using PHYML [59] with the GTR+I+G nucleotide substitution model; using alignment-specific trees is more appropriate than using the species tree in this case, given that recombination has scrambled the orthologous relationships of these genes. Mx coding sequence alignments were fit to NS sites models that disallow (M7 or M8A) or allow (M8) ω > 1. Models were compared using a chi-squared test (degrees of freedom = 2) on twice the difference of likelihood values to derive P values reported in Fig 1A and S1A Fig). Analyses were robust to varying codon frequency models (F3x4 and F61); results from analyses using model F3x4, which we empirically find to be more conservative, are shown in all figures. The model 7 vs 8 comparison was also robust to the initial omega value used (0.4, 1 or 1.5); model 8a cannot be run with other initial omega values. In cases where a significant difference (P < 0.01) between M7 versus M8 (or M8A versus M8) was detected, the Bayes Empirical Bayes (BEB) analysis was used to identify codons with ω > 1 (reporting values with posterior probability ≥ 0.95). The depiction of positively selected sites on crystal structure representations was carried out in PyMol (pymol.org).
We also used the REL algorithm of the HyPhy package to detect selected sites [17]—we uploaded each alignment to the DataMonkey website, used the model selection tool to select the most likely evolutionary model, and used that model and the alignment as input to the REL algorithm.
|
10.1371/journal.ppat.1002536 | Efficient Capture of Infected Neutrophils by Dendritic Cells in the Skin Inhibits the Early Anti-Leishmania Response | Neutrophils and dendritic cells (DCs) converge at localized sites of acute inflammation in the skin following pathogen deposition by the bites of arthropod vectors or by needle injection. Prior studies in mice have shown that neutrophils are the predominant recruited and infected cells during the earliest stage of Leishmania major infection in the skin, and that neutrophil depletion promotes host resistance to sand fly transmitted infection. How the massive influx of neutrophils aimed at wound repair and sterilization might modulate the function of DCs in the skin has not been previously addressed. The infected neutrophils recovered from the skin expressed elevated apoptotic markers compared to uninfected neutrophils, and were preferentially captured by dermal DCs when injected back into the mouse ear dermis. Following challenge with L. major directly, the majority of the infected DCs recovered from the skin at 24 hr stained positive for neutrophil markers, indicating that they acquired their parasites via uptake of infected neutrophils. When infected, dermal DCs were recovered from neutrophil depleted mice, their expression of activation markers was markedly enhanced, as was their capacity to present Leishmania antigens ex vivo. Neutrophil depletion also enhanced the priming of L. major specific CD4+ T cells in vivo. The findings suggest that following their rapid uptake by neutrophils in the skin, L. major exploits the immunosuppressive effects associated with the apoptotic cell clearance function of DCs to inhibit the development of acquired resistance until the acute neutrophilic response is resolved.
| Prior studies in mice have shown that the inoculation of Leishmania major into the skin by sand fly bite or by needle provokes a massive recruitment of neutrophils that take up the parasite, and that this response somehow suppresses immunity since neutrophil depletion results in better control of the infection. We investigated how neutrophils recruited to the injection site might interact with and suppress the function of dendritic cells (DCs) in the skin. Infected neutrophils recovered from the skin expressed increased levels of apoptotic markers compared to uninfected neutrophils, and were efficiently taken up by dermal DCs when injected back into the skin. When dermal DCs were permitted to take up parasites in the absence of neutrophils, their expression of activation markers and their ability to present Leishmania antigens were enhanced. Neutrophil depletion also enhanced the activation of Leishmania specific CD4+ T cells in vivo. The results suggest that for insect borne pathogens like Leishmania that provoke a strong inflammatory response at the site of infection, the immunosuppressive effects associated with the apoptotic cell clearance function of DCs will inhibit the early development of immunity.
| Leishmaniasis is a vector-borne disease initiated by the bite of an infected sand fly. Based on exhaustive findings in the murine model of cutaneous leishmaniasis due to Leishmania major, the clinical course of disease is thought to depend on the balance of activating cytokines, produced largely by Th1 cells, and deactivating cytokines, produced largely by Th2 cells and subsets of regulatory T cells [1]. Even in genetically resistant C57BL/6 mice, however, that develop self-limiting lesions due to a strongly polarized Th1 response, the early growth of the parasite is unrestrained, suggesting that innate killing mechanisms and the development of acquired resistance are avoided or delayed [2]. There is evidence that the acute neutrophilic response is itself critical to the early establishment of infection in the skin [3], [4]. Inoculation of L. major by the bite of a sand fly, or by needle injection, induces an intense infiltration of neutrophils that phagocytose the majority of parasites but fails to kill them, and neutrophil depletion prior to sand fly challenge leads to more rapid parasite clearance [5]. The manner in which the acute neutrophilic response inhibits the development of immunity to L. major infection is not understood.
Neutrophils and DCs are normally located in distinct anatomical compartments, but converge at sites of inflammation in response to infection or tissue injury. The essential function of neutrophils in phagocytosis and killing of bacteria and in tissue repair is well described [6], [7]. Their additional role in modulating the adaptive response is suggested by their ability to release chemokines, cytokines, and anti-microbial peptides, [8], [9], and by more recent findings suggesting that activated neutrophils can deliver both activation signals and microbial antigens to DCs [10], [11]. By contrast, engulfment of apoptotic cells, including neutrophils, by DCs under steady state conditions has been shown to suppress DC maturation and is thought critical to the maintenance of peripheral tolerance [12]–[14]. Thus, the immunologic outcome of neutrophil - DC interactions may vary depending on the activation state of the neutrophils, their type of cell death, and the presence or absence of additional danger signals in the microenvironment in which these encounters occur.
Importantly, the cross-talk between neutrophils and DCs has not been investigated in the context of any vector borne pathogen for which the co-localization of these cells at the site of transmission by bite or injection by needle in the skin is apt to be especially pronounced. In the present studies, we have monitored the sequence of inflammatory events following infection with L. major in the mouse ear dermis. We provide clear evidence that dermal DCs are preferentially infected via their capture of parasitized neutrophils in the skin, and that the Leishmania specific CD4+ T cell response is compromised until the acute neutrophilic response is resolved.
We investigated the sequence of local inflammatory responses and identified the cells harboring L. major following injection of Lm-RFP metacyclic promastigotes (2×105) in the ear dermis of C57BL/6 mice. Myeloid populations were identified as CD11b+ cells, and further classified based on their expression of additional markers (Figure 1A) as follows: neutrophils (Ly6CintLy6G+, region 1); inflammatory monocytes (Ly6ChiLy6G−CD11c−MHCII−, region 2); monocytes/macrophages (Ly6ChiLy6G−CD11c−MHCII+, region 3); monocyte-derived DCs (Ly6ChiLy6G−CD11c+MHCII+, region 4); dermal DCs (Ly6C−Ly6G−CD11c+MHCII+, region 6); and dermal macrophages (Ly6C−Ly6G−CD11c−MHCII+, region 5). The cells in region 5 were uniformly F4/80+ (data not shown). The CD11b+ cells recovered from naïve ears included few neutrophils and inflammatory monocytes, and relatively greater numbers of dermal DCs and macrophages. The total number of CD11b+ cells recovered from the infected ears increased slowly over the first week, and expanded dramatically over the second week (Figure 1B). A prominent and transient neutrophil infiltrate accounted for the earliest increase in myeloid cells in the site, beginning at 1 hr, peaking at 12 hr, and dropping markedly between 1- 4 days (Figure 1C). Interestingly, neutrophils were found infiltrating the site again at day 7, and by day 14 their numbers exceeded the peak numbers observed during the first wave of the neutrophilic response. Comparison of L. major infected and sham injected mice demonstrated that at 1 hr the initial neutrophil infiltrate was induced, at least in part, by the tissue injury associated with the needle injection. At subsequent time points, however, the recruitment was dependent on the infectious status of the inoculum (Figure 1I).
The increase in the number of inflammatory monocytes (Figure 1D) lagged slightly behind the neutrophil response, beginning at 12 hr and peaking at 24 hr. Similarly to the neutrophils, their numbers dropped markedly by 4 days but began to increase again by day 7. Very few of the CD11b+Ly6ChiLy6G− cells recovered from the site during the first week of infection were MHCII+ or CD11c+ (Figure 1F), of note because of recent findings implicating monocyte-derived DCs formed at the infection site as crucial to the induction of protective immunity during the active stage of disease [15]. The number of macrophages and DCs remained relatively unchanged from steady state conditions until 7 days post-infection, marking the onset of their massive accumulation in the site (Figure 1G and 1H).
By analyzing the total population of RFP+ gated cells, we could follow the subsets of infected cells in the injection site over time (Figure 2A–C). Regions 1–6 define to the same subsets of myeloid cells as the corresponding regions in figure 1A, and in each case their CD11b expression was confirmed (data not shown). By contrast, many of the infected cells in region 7 were CD11b−, and their identity was not established using additional markers. Considering the total population of RFP+ cells (Figure 2D), low numbers were recovered at 1 and 4 hr which significantly increased between 4–12 hr and dramatically increased between 7–14 days (Figure 2D). In Figures 2E–L, the infected subsets are expressed both as a percentage of the total infected cells and their absolute numbers recovered from the ear dermis at each time point. Neutrophils were the predominant infected cells during the first 1–12 hrs (Figure 2E). At 12 hrs, 72% of the infected cells were neutrophils, with the remainder inflammatory monocytes, macrophages, DCs and other populations of CD11b+ and CD11b− cells. At 24 hr, neutrophils still represented approximately 32% of the total RFP+ cells. By day 4, the percentage of neutrophils in the RFP+ gate had dropped to fewer than 1%. Interestingly, their numbers began to increase again by day 7, and by day 14, the absolute number of infected neutrophils in the site exceeded the peak numbers observed during the first wave, although they remained <5% of the total population of infected cells. The inflammatory monocytes in the RFP+ gate also demonstrated two phases of recruitment, the first peaking at day 1 when they represented 22% of the total RFP+ cells, and the second at day 7 (Figure 2F). Their absolute numbers were greatest at day 14, again reflecting the massive expansion in the total number of infected cells at this time point. Very few of the RFP+ inflammatory monocytes recovered during the first 4 days were MHCII+ or CDllc+, while at 7 and 14 days, the majority of the RFP+ Ly6ChiLy6G− cells were MHCII+ and CD11c− (Figure 2G), reflecting the early stage of their differentiation to macrophages in the site. By day 14, appreciable numbers of infected monocyte-derived DCs (Figure 2H) were recovered, though they still represented only around 4% of the total RFP+ cells. By contrast, the infected macrophages (Figure 2I), expressed both as a percentage and absolute number of infected cells, started to increase at 4 days, and accounted for up to 20% of the total RFP+ cells at 14 days. The RFP+ dermal DCs remained few in number and <5% of the RFP+ cells over the first 24 hr, and while they remained a low percentage of the RFP+ cells at later time points, their absolute numbers increased markedly at 7 days and especially at 14 days post-infection (Figure 2J).
In summary, our detailed analysis of infected cells in the L. major loaded dermis confirmed that neutrophils rapidly infiltrating the site represent the vast majority of infected cells over the first 12 hr, with the infections transitioning to inflammatory monocytes, and finally to monocyte derived macrophages and DCs during the active stage of disease.
Given their predominance both as the earliest infiltrating and parasitized cells in the injection site, we investigated the influence of neutrophils on the subsequent program of infection and immune response. In vitro studies have suggested that macrophages can acquire L. major by phagocytosing infected, apoptotic neutrophils [16], [17]. To investigate the fate of infected neutrophils and their internalized parasites, Lm-RFP metacyclic promastigotes were injected into the ears of LYS-eGFP mice [18], in which neutrophils (CD11bhiGr-1hiF4/80−MHCII−), including those recovered from the skin, are eGFPhi [5]. eGFPhiRFP+ infected neutrophils were purified by cell sorting (Figure 3A) and injected into the ears of C57BL/6 mice. Analysis of a stained, cytospin preparation of the sorted cells just prior to injection indicated that approximately 30% of the parasites had already been released from the neutrophils during the 4–5 hr collection. Four hours after injection, the vast majority of the RFP+ cells recovered from the ear (90%) were found in an eGFP− population (Figure 3B), suggesting that in addition to the free parasites present in the inoculum, most of the remaining parasites were released from the infected neutrophils and available to be taken up by host cells in the skin. These cells were CD11clo, and their F4/80 and CD11b expression indicated that they were endogenous macrophages/monocytes or neutrophils (Figure 3D). Of the RFP+ cells that retained their eGFP fluorescence, approximately half appeared to be the injected population of intact, infected neutrophils (Figure 3C). The remaining RFP+eGFP+ cells were CD11c+, suggesting that the capture of infected neutrophils in the skin was largely accomplished by DCs. Of the total number of CD11c+RFP+ cells, 68% were eGFP+ (Figure 3E), suggesting that most of the DCs acquired their parasites via uptake of infected neutrophils. Of note, the eGFP fluorescence in these cells was reduced relative to the starting population of infected neutrophils. To rule out the possibility that the eGFPhiRFP+ infected neutrophils could have differentiated into CD11c+ cells, or that a small contaminating population of CD11c+ cells in the purified eGFPhiRFP+ infected neutrophils was responsible for the RFP+eGFP+CD11c+ cells observed in figure 3B, we sorted eGFPhiRFP+ infected neutrophils (donor, CD45.2), and injected them into the ears of B6SJL mice (host, CD45.1). Analysis of CD45.1 expression on the subpopulations of RFP+ cells indicated that virtually all of the RFP+eGFP+CD11c+ cells were CD45.1+, ruling out their donor origin (Figure S1).
To investigate whether DCs might favor engulfment of infected neutrophils over uninfected neutrophils in the skin, equal numbers of eGFPhiRFP− uninfected and eGFPhiRFP+ infected neutrophils (Figure 3A) were co-injected into the ears of C57BL/6 mice. The analysis of eGFP+ gated cells recovered four hours later confirmed that DCs are able to take up neutrophils in vivo, representing approximately 11% of the eGFP+ cells (Figure 3F). Importantly, and despite their exposure to equivalent numbers of infected and uninfected neutrophils, an average of 66% of the CD11c+eGFP+ cells were RFP+ (Figure 3F and G), indicating that the dermal DCs favored the uptake of the infected neutrophils.
The capture of mammalian cells by DCs and others phagocytes is a common event during tissue remodeling and at infection sites, where cells dying by apoptosis expose signals that are recognized for engulfment. The best studied ‘eat me’ signal on apoptotic cells is phosphatidylserine (PtdSer) whose outer membrane exposure can be quantified by staining with Annexin V. When neutrophils were recovered from the ear dermis of LYS-eGFP mice 12 hr after infection, 53% of the infected neutrophils compared to 19% of the uninfected neutrophils were Annexin V+ (Figure 4A and B). TUNEL staining confirmed a higher degree of apoptosis in the infected population of neutrophils recovered from the injection site (Figure 4C and D). These findings suggest that the uptake of L. major leads to accelerated apoptosis and earlier exposure of PtdSer on neutrophils infiltrating the injection site, which may favor their recognition and capture by DCs in the skin.
To investigate neutrophil - DCs interactions following injection of the parasite directly, we evaluated dermal DCs recovered from C57BL/6 mice 24 hr after infection with Lm-RFP parasites, and stained for neutrophil derived-myeloperoxidase (MPO) and elastase (NE). In addition, mice were treated with two neutrophil-depleting antibodies: the GR-1 specific antibody RB6-8C5, which recognizes an epitope shared by Ly6G and Ly6C, and the Ly6G specific antibody, 1A8. Administration of 1A8 one day before infection depleted 85% of the CD11b+GR1hiLy6Cint neutrophils present in the ear dermis 24 hr after infection (Figure 5A and C). The remaining neutrophils showed lower GR1 staining, likely due to competition with the surface bound 1A8 antibody. The CD11b+GR1intLy6Chi population was unaffected. By contrast, and consistent with the prior reports [19], [20], the RB6-8C5 antibody depleted both neutrophils and a population of inflammatory monocytes (Figure 5B and C). Furthermore, the neutrophil depletion achieved using RB6-8C5 was virtually complete (99%). Neither reagent affected the total number of DCs recovered from the ear at 24 hr, or the number RFP+ DCs as a percentage of the total population of RFP+ cells (Figure 5D and E). Gating on RFP+ or RFP− dermal DCs (Figure 5F), MPO staining on cells recovered from the control treated mice was observed in an average of the 58% of the RFP+ DCs, suggesting that the majority of the infected DCs acquired their parasites via uptake of infected neutrophils (Figure 5G and H). By contrast, only a low proportion (<5%) of the RFP−DCs were MPO+, although because far more RFP−DCs were recovered from the site compared to RFP+DCs (Figure 5F), the percentage of RFP−DCs staining for MPO was on average 60% of the total population of MPO+ DCs (data not shown). The intracellular MPO staining in the majority of the infected DCs was comparable to the MPO staining observed in the neutrophils themselves, and greater than the MPO staining observed in the inflammatory monocytes recovered from the site (Figure S2), reinforcing the conclusion that the acquisition of the MPO marker by infected DCs was due to their uptake of infected neutrophils. Importantly, the RFP+ DCs recovered from the RB6-8C5 treated mice were virtually all MPO−, confirming that the uptake of parasites in the absence of neutrophils or inflammatory monocytes does not upregulate the expression of MPO in the DCs. The number of RFP+ DCs recovered from the 1A8 treated mice that stained for MPO was also significantly reduced, though an average of 24% of the cells were still MPO+ cells, consistent with the incomplete neutrophil depletion using this antibody (Figure 5G and H). Staining for NE, while relatively weak compared with MPO, reinforced the MPO result in that the majority of the RFP+ DCs recovered from the non-depleted mice stained positive for NE (Figure S3). Finally, RFP+ DCs recovered from control treated mice 14 days after infection were mainly MPO− (Figure 5I), suggesting that following the resolution of the acute neutrophilic response, infected neutrophils were no longer the main source of parasite delivery for DCs in the skin. We further characterized the possible subsets of the RFP+ DCs recovered from the site based on their expression of Langerin and CD103. As reviewed [21], and confirmed in our analysis of the DCs recovered from the ear dermis 24 hr after infection, the DC subsets include Langerhans cells (LC) and migratory LC (CD11c+MHCII+Lang+CD103−), Langerin+ DC (CD11c+MHCII+Lang+CD103+), and Langerin− DC (CD11c+MHCII+Lang−CD103−) (Figure S4). The RFP signal was associated exclusively with the Langerin− DCs.
To address whether neutrophils might modulate the antigen presentation functions of DCs during the early stages of infection, the expression of activation markers on infected DCs recovered from the ear dermis 3 days after infection in neutrophil-depleted (RB6-8C5) or control treated C57BL/6 mice was compared (Figure 6A–C). Expression of MHC class II, CD86 and CD40, but not CD80, was increased on RFP+ DCs recovered from the neutrophil depleted mice (Figure 6B and C). Functional studies involving these infected DCs required pooling of dermal cells from 10 mice (20 ears) for each treatment group in order to obtain a sufficient source of antigen and antigen presenting cells for the co-culture assays. Using CD11c+ RFP+ cells that were normalized for their RFP signals by cell sorting (Figure 6D), the infected DCs from neutrophil depleted mice were more efficient than the infected DCs from the control treated mice in activating Leishmania-primed T cells from healed mice to secrete IFN-γ, observed in two independent experiments (Figure 6E).
To evaluate the influence of neutrophils on CD4+ T cell priming to L. major- derived antigen in vivo, B6.SJL congenic mice were depleted of neutrophils 24 hr prior to infection with L. major SP-OVA or control 3′NT transgenic parasites in the ear. CFSE-labeled, naïve OT-II CD4+ T cells specific for OVA were adoptively transferred into the same recipients. Draining lymph nodes were harvested on day 6 and dilution of CFSE fluorescence was determined on CD45.2+ and CD4+ gated cells. Infection of control treated mice with Lm SP-OVA failed to induce OT-II proliferation above the background levels (6–7%) observed in control treated or neutrophil depleted mice infected with Lm 3′NT (Figure 6E). By contrast, mice treated with 1A8 and RB6-8C5 had an average of 20% and 34% of the gated cells in division, respectively (Figure 6F). We also assessed the ability of CD45.2+ OT-II CD4+ cells to produce IFNγ, IL-10 and IL-17, following ex-vivo restimulation with PMA/ionomycin for 4 hours in the presence of brefeldin-A. A percentage of proliferating CD45.2+ OT-II CD4+ cells from 1A8 and RB6-8C5 treated mice (29% and 25%, respectively) produced IFNγ (Figure 6G). Neither IL-10- nor IL-17A-producing T cells were detected (data not shown). The influence of early neutrophil depletion on CD4 priming was no longer apparent when OT-II cells were transferred 14 days post-infection with Lm SP-OVA (Figure 6H), at a time when infected DCs no longer harbored neutrophil markers (Figure 5I). Taken together, these findings suggest that the favored uptake of infected neutrophils by dermal DCs effectively prevents the activation of Leishmania-specific CD4+ T cells until the acute neutrophilic response is resolved.
We have recently described the efficient capture of L. major metacyclic promastigotes by neutrophils at the site of needle inoculation or infected sand fly bite, and the powerful effects of early neutrophil depletion in promoting rather than compromising host resistance to sand fly transmitted infection [5], [22]. The current studies provide an underlying mechanism to explain the immunomodulatory role of neutrophils in the L. major loaded dermis. Under steady state conditions, DCs are strategically positioned in peripheral and lymphoid tissues to sense microorganisms and endogenous stress signals, including apoptotic cells. Neutrophils, by contrast, are present mainly within the blood, and circulate in a non-activated state with a half-life of 6–7 hrs. Following inoculation of L. major into the skin by needle or by the bite of an infected sand fly, the parasites are taken up by neutrophils that are rapidly recruited to and accumulate with DCs at the injured site. We observed that phagocytosis of L. major significantly accelerated the rate of neutrophil apoptosis, which was associated with the favored uptake of infected over uninfected neutrophils by DCs in the skin. More importantly, for the majority of infected DCs in the skin their initial encounter with the parasite occurred via capture of infected neutrophils, with a negative impact on CD4+ T cell priming.
These studies confirm the previous findings in L. major [5], [23], recently extended to L. infantum [24], that neutrophils are rapidly recruited to and accumulate in the inoculation site, and represent the predominant parasitized cell during the first 1–12 hours of infection in the skin. The inflammatory and infectious process induced by L. major in the skin may be regulated in a tissue specific manner, since recent observation by Gonçalves et al. [25] and confirmed by our own studies (data not shown) have revealed that when L. major metacyclics are introduced into the peritoneal cavity, neutrophils are neither the first infiltrating nor predominant infected cells. Our kinetic analysis of the L. major loaded dermis revealed that the rapid neutrophilic response is initiated in part by signals generated by the tissue injury produced by the needle injection itself, since a transient recruitment was observed in sham injected mice, and amplified by more durable signals derived from the parasite and/or from infected cells [26]. The fate of the infected neutrophils was followed by transfer of eGFPhiRFP+ cells into the ear dermis of C57BL/6 mice. By 4 hr, the majority of RFP+ cells recovered from the site were endogenous neutrophils and monocytes/macrophages that were eGFP−, consistent with our prior in vivo imaging results that readily captured infected neutrophils undergoing apoptosis and releasing viable parasites for subsequent uptake by other cells in the skin [5]. Thus, the ‘Trojan Horse’ hypothesis as originally proposed [17], in which neutrophils serve as a vector for silent entry of Leishmania into macrophages, has not been directly substantiated in these studies. We cannot, however, dismiss the possibility that phagosomal degradation of the eGFP signal occurred rapidly following engulfment of the infected neutrophils by macrophages. It is also possible that clearance of neutrophil-derived, apoptotic bodies by infected macrophages would still contribute to their deactivation and promote the intracellular survival and growth of the parasite, as proposed.
By contrast to macrophages, the evidence for the uptake of L. major infected neutrophils by DCs in the skin seems clear. Firstly, CD11c+ cells were the only endogenous cells associated with both the RFP and eGFP signals. Secondly, when the infections were initiated by RFP L. major metacyclics, the majority of the RFP+ DCs recovered from the injection site at 24 hr also stained positive for neutrophil-derived MPO and elastase. In studies by Ng et al. [27], two-photon imaging captured dermal DCs but not Langerhans cells taking up Leishmania promastigotes in the skin. We also found Langerin− dermal DCs as the major infected DC subset in the skin, but conclude based on their staining for neutrophil markers, and the absence of these markers in DCs that have taken up parasites in the absence of neutrophils, that the majority of the infected DCs acquired their parasites via engulfment of infected neutrophils. Favored uptake of infected over uninfected neutrophils was also observed, correlated with their accelerated expression of apoptotic markers that may have targeted them for more efficient recognition and clearance by DCs. Neutrophil ingestion of other microbial pathogens, notably E. coli [28], Str. pneumoniae [29], [30], C. albicans [31], Sta. aureus [32], and M. tuberculosis [33], has also been found to accelerate their apoptotic program. The findings involving Leishmania are inconsistent on this point, with delayed or enhanced expression of PtdSer observed on neutrophils obtained from human blood or the mouse peritoneal cavity and exposed to Leishmania in vitro [34]–[37]. The current studies are the first to compare the apoptotic profile of tissue infiltrated neutrophils that have taken up parasites, or not, in the inflamed dermis.
Apoptosis is an active process to regulate cellular homeostasis. Efferocytosis refers to the capture of apoptotic cells by phagocytes, primarily macrophages and immature DCs (iDC), and is itself thought to be a homeostatic mechanism to resolve inflammation and to maintain peripheral tolerance [13]. Recognition and engulfment of apoptotic cells, including apoptotic neutrophils, by DC is known to inhibit their production of pro-inflammatory cytokines, expression of costimulatory molecules, and their ability to stimulate T-cell proliferation [14], [38], [39]. The exploitation of these inhibitory signals by microbial pathogens is suggested by in vitro studies showing that M. tuberculosis-induced activation of human iDC can be inhibited by their co-culture with apoptotic neutrophils [40], and that Plasmodium falciparum-infected erythrocytes can inhibit the maturation of mouse DCs by binding to CD36, a known recognition receptor for apoptotic cells [41]. The present studies are the first to demonstrate efferocytosis involving neutrophils and DCs in an infection driven inflammatory setting in vivo. The sequestration of Leishmania antigens within apoptotic neutrophils would seem an especially efficient process to exploit the immunosuppressive signals conferred by the clearance of dying cells by DCs. Removing host neutrophils as a source of apoptotic cells was sufficient to reconstitute the immune function of infected DCs. It should be noted that in contrast to recent studies [42], we did not observe a reduction in either the total number of DCs nor infected DCs recovered from the ear following neutrophil depletion (Figure 5D and E). We would offer that while the prior study was confined to cells migrating out of the ear dermis ex vivo, our analysis was based on the greater recovery of cells following enzymatic digestion of the tissue. By comparing the ex vivo APC function of infected DCs recovered from the skin of mice depleted or not of neutrophils, and normalized for their RFP signals, the inhibitory effects of neutrophil uptake on DC maturation and Leishmania specific T cell activation could be formally demonstrated. The consequence of this inhibition in effectively delaying the onset of Leishmania specific T cell priming in vivo was directly supported by the enhanced, early OT-II priming to Lm-derived OVA in the neutrophil depleted mice.
The neutrophil - DC interactions that inhibit T cell priming following needle challenge with L. major might be relevant to more general vaccination protocols in which an acute neutrophilic infiltrate accumulates at the site of antigen deposition. A recent report by Yang et al. [43] described the negative influence of neutrophils on the T and B cell responses to protein antigens administered by needle in the footpad. It is clear, however, that apoptotic neutrophils can also provide a source of immunogenic molecules to DC, especially for cross-priming, and especially if accompanied by extrinsic maturation signals [44], [45]. The relative paucity of activation signals associated with the phagocytosis of Leishmania promastigotes by neutrophils is suggested by the fact that the parasite traffics to a non-lytic compartment, avoids activation of the NADPH oxidase, and survives capture by these cells [5], [37]. It should be noted that PtdSer exposure on the parasites themselves has been suggested to facilitate their silent entry into macrophages, [46], [47], and may be especially relevant to their initial survival in neutrophils. Following neutrophil depletion, or the resolution of the first wave of neutrophils in the site, the majority of the infected DCs recovered from the skin lacked neutrophil markers, and are presumed to have taken up the parasite directly. By contrast to the absence of activation signals associated with the direct uptake of L. major metacyclic promastigotes by macrophages, the activation of human and mouse DCs following their phagocytosis of these organisms in vitro is well described [48]. Direct uptake might allow for parasite antigens to be more accessible to the MHC class I and II processing machinery, for parasite encoded TLR agonists to more efficiently engage their respective receptors, and for activation pathways to proceed in the absence of the inhibitory signals induced by apoptotic cell clearance. By two weeks, the priming conditions had clearly improved, and neutrophil depletion did not further enhance the CD4+ T cell response, despite the reappearance of neutrophils in the site. In contrast to the initial wave, however, the infected neutrophils recovered at two weeks represented a small percentage of the total population of infected cells, and the majority of infected DCs no longer harbored neutrophil markers. It is likely that the conditions of neutrophil recruitment to and activation in the skin during the active stage of disease, possibly Th17 driven at this later time, are distinct from those associated with the acute infiltrate, and that the influence of these respective neutrophil populations on the anti-leishmanial response will also be distinct.
In the current studies, there was a significant difference in the effects of the neutrophil depleting antibodies, 1A8 and RB6-8C5, in potentiating the early OT-II response to infection with Lm SP-OVA in the skin. The 1A8 treatment critically confines the enhanced priming observed to specific depletion of Ly6G+ neutrophils. The more powerful effects observed with the RB6-8C5 antibody is consistent with the more efficient depletion of neutrophils that was achieved, although the removal of an additional population of GR-1+ myeloid cells with suppressor activity [49], [50] cannot be discounted.
While our studies have employed a relatively high dose, needle challenge in order to recover a sufficient number of infected cells from the ear dermis for analysis, it should be emphasized that the initial wave of neutrophil recruitment to the infected sand fly bite site is more massive, localized, and sustained compared to the needle injection site [22]. This may explain why the ablation of the early neutrophilic response had such a strong effect in promoting protection against sand fly transmitted infection as compared to needle challenge [5], [51]–[53]. Thus, the impact of the early neutrophil - DC interactions described in these studies may be especially relevant to the inflammatory conditions elicited by natural sand fly transmission, as well as to that of other vector borne pathogens, in promoting the early establishment of infection and the progression of disease.
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Animal Care and Use Committee of the NIAID, NIH (protocol number LPD 68E). All mice were maintained at the NIAID animal care facility under specific pathogen-free conditions. Female C57BL/6 and B6.SJL congenic mice, and RAG1-deficient OT-II CD4+ TCR transgenic mice were purchased from Taconic Laboratories. C57BL/6 LYS-eGFP knock-in mice [18] were a gift from T. Graf (Albert Einstein University, NY) and were bred at Taconic Laboratories.
Experiments were carried out using different lines of L. major: L. major Friedlin strain FV1 (MHOM/IL/80/FN); a stable transfected line of L. major FV1 promastigotes expressing a red fluorescent protein (Lm-RFP), L. major FV1 promastigotes expressing a portion of the ovalbumin gene encoding amino acids 139 to 386 containing the class II restricted epitope recognized by OT-II TCR transgenic CD4+ T cells (Lm- SP-OVA), and L. major V1- transfected with the control plasmid expressing Leishmania donovani 3′ nucleotidase-nuclease (Lm-NT). Transfected lines were generated as described previously [54]–[55].
Parasites were grown at 26°C in medium 199 supplemented with 20% heat-inactivated FCS (Gemini Bio-Products), 100 U/ml penicillin, 100 µg/ml streptomycin, 2 mM L-glutamine, 40 mM Hepes, 0.1 mM adenine (in 50 mM Hepes), 5 mg/ml hemin (in 50% triethanolamine), 1 mg/ml 6-biotin (M199/S), and 50 µg/ml of Geneticin (Gibco). Infective-stage, metacyclic promastigotes of L. major were isolated from stationary cultures (4–5 days old) by negative selection using peanut agglutinin (PNA, Vector Laboratories Inc). For flow cytometric studies of dermal and draining lymph node cells, mice were infected with the specified number of metacyclic promastigotes in the ear dermis by i.d. injection in a volume of 10 µl. In parallel, sham mice received i.d. injection of DMEM in a volume of 10 µl. To obtain chronically infected mice, animals were infected 16–20 weeks previously with 104 L. major FV1 metacyclic promastigotes in the left hind footpad.
Ear tissue was prepared as previously described [2]. Briefly, the two sheets of infected ear dermis were separated, deposited in DMEM containing 100 U/ml penicillin, 100 µg/ml streptomycin, and 0.2 mg/ml Liberase CI purified enzyme blend (Roche Diagnostics Corp.), and incubated for 1 h and 30 min at 37°C. Digested tissue was placed in a grinder and processed in a tissue homogenizer (Medimachine; Becton Dickenson). Retromaxillary (ear) lymph nodes were removed, and mechanically dissociated using tweezers and a syringe plunger. Tissue homogenates were filtered through a 70 µm cell strainer (Falcon Products).
Single-cell suspensions were incubated with an anti-Fc-γ III/II (CD16/32) receptor Ab (2.4G2, BD Biosciences) in RPMI without phenol red (Gibco) containing 1% FCS and stained with fluorochrome-conjugated antibodies. The following antibodies were used: APC- anti-mouse CD11c (HL3, BD Biosciences), PE-Cy7- anti-mouse CD11c (N418, eBioscience), PerCP-Cy5.5 or PE-Cy7- anti-mouse CD11b (M1/70, eBioscience); PerCP-Cy5.5- anti-mouse Ly6C (HK1.4, eBioscience); FITC- anti-mouse Ly6G (1A8, eBioscience); FITC- anti-mouse GR-1 (RB6-8C5, BD Biosciences); eFluor anti-mouse F4/80 (BM8, eBioscience), Alexafluor-700 anti-mouse MHC II (M5/114.15.2, eBioscience), APC- anti-mouse CD103 (M290, eBioscience), A488- anti-mouse Langerin (929F3.01, Dendritics), APC- anti-mouse CD40 (1C10, eBioscience), FITC- anti-mouse CD80 (16-10A1, eBioscience), PerCP-Cy5.5- anti-mouse CD86 (GL-1, BioLegend), APC- anti-mouse CD4 (RM4–5, eBioscience), PerCP-Cy5.5- anti-mouse CD45.2 (104, eBioscience); APC-eFluor 780 anti-mouse CD45.1 (A20, eBioscience), FITC- anti-mouse myeloperoxidase (MPO) (8F4, Hycult), anti-human neutrophil elastase (NE) (H-57, Santa Cruz), FITC conjugated using and amine reactive probe (Sigma-Aldrich). The isotype controls used (all obtained from BD Biosciences) were rat IgG1 (R3–34) and rat IgG2b (A95-1). The staining of surface and intracytoplasmic markers was performed sequentially: the cells were stained first for their surface markers, followed by a permeabilization step with BD Cytofix/Cytoperm (BD Biosciences) and staining for Langerin, MPO or NE. For intracellular detection of cytokines, cells were first stimulated with Leukocyte Activation Cocktail, plus GolgiPlug (BD Biosciences) according the manufacturers' instructions for 4 h in vitro. Following surface staining and permeabilization, cells were then stained with a combination of anti-mouse antibodies: PerCP-Cy5.5 anti-IL17A (eBio17B7, eBioscience) APC anti-IFN-g (XMG1.2, eBioscience), PE anti-IL-10 (JES5-16E3, BD Bioscience) in Perm/Wash buffer (BD Bioscience). Intracellular staining was carried out for 30 minutes on ice. The data were collected and analyzed using CELLQuest software and a FACScalibur or FacsDIVA software and a FacsCANTO flow cytometer (BD Biosciences). Neutrophils, dendritic cells, macrophages and monocytes from the ear dermis were identified based on size (forward scatter) and granularity (side scatter) and by surface phenotype as indicated in the text and figure legends.
Infected DCs (CD11c+RFP+) were purified using a FACSVantage or a FACsAria (BD Biosciences) cell sorter on cells recovered from the ear dermis 3 days after infection with 2×106 L. major-RFP. For the analysis of the capacity of infected, dermal DCs to induce the secretion of IFNγ by L. major specific T cells, 4×104 (Exp. 1) or 4.5×104 (Exp. 2) infected dermal DCs pooled from 10 mice (20 ears) for each treatment group were co-cultured with 1×105 T cells purified by negative selection (Miltenyi Biotec) from draining lymph nodes (dLNs) of B6 mice with a healed, primary infection with L. major FV1. After 3 days, culture supernatants were analyzed for IFN-γ production by ELISA (eBioscience). For adoptive transfer experiments, CD4+ T cells were purified from spleens and lymph nodes of RAG1-deficient OT-II CD4+ TCR transgenic mice by negative selection (Miltenyi Biotec). Purified CD4+ T cells were incubated at 2.5–5×107 cells/ml in PBS with 0.5 µM CFSE (Invitrogen) for 10 min at 37°C. The reaction was stopped with 10% normal mouse serum, and the cells were washed twice with cold PBS/0.1% BSA. B6.SJL congenic mice received intravenously (i.v.) 2–5×105 CFSE-labeled, purified CD4+ OT-II T cells either the same day or 14 days after challenge in the ear dermis with 105 metacyclic promastigotes. Six days after adoptive transfer, the dLNs were removed and analyzed by flow cytometry. To obtain neutrophils recruited to the site of infection in the skin, LYS-eGFP mice were inoculated in the ear dermis with 2×106 Lm-RFP. Twelve hours later the ear tissue was prepared as described above and infected (RFP+eGFPhi) and uninfected (RFP−eGFPhi) neutrophil populations were sorted from dermal tissue using a FACSVantage or a FACsAria (BD Biosciences) cell sorter. Sorted populations were washed once and immediately analyzed for apoptosis or injected into the ear dermis of C57BL/6 and B6SJL recipient mice in a volume of 10 ul.
Sorted, infected (RFP+eGFPhi) and uninfected (RFP−eGFPhi) neutrophil populations were stained with Annexin-V-APC and 7-AAD (BD Biosciences) as recommended by the manufacturer. For TUNEL assays, neutrophil populations were fixed in 4% paraformaldehyde, and then labeled with the Beckman Coulter Mebstain Apoptosis kit using biotinylated dUTP. Cells were then incubated with streptavidin-conjugated APC (BD Pharmingen) for 30 min at room temperature. Cells were analyzed by flow cytometry.
Neutrophils were depleted employing a single i.p. injection of 0.5 mg RB6-8C5 (anti-Gr-1), or 1 mg of 1A8 (anti-Ly6G, BioXCell), or GL113 (control IgG, BioXCell), 1 d prior to parasite injection. The efficiency and specificity of the depletions were evaluated on dermal cell preparations, and on heparinized whole blood.
Statistical significance between groups was determined by the unpaired, two-tailed student's t test using Prism software (GraphPad).
|
10.1371/journal.ppat.1005779 | How Quorum Sensing Connects Sporulation to Necrotrophism in Bacillus thuringiensis | Bacteria use quorum sensing to coordinate adaptation properties, cell fate or commitment to sporulation. The infectious cycle of Bacillus thuringiensis in the insect host is a powerful model to investigate the role of quorum sensing in natural conditions. It is tuned by communication systems regulators belonging to the RNPP family and directly regulated by re-internalized signaling peptides. One such RNPP regulator, NprR, acts in the presence of its cognate signaling peptide NprX as a transcription factor, regulating a set of genes involved in the survival of these bacteria in the insect cadaver. Here, we demonstrate that, in the absence of NprX and independently of its transcriptional activator function, NprR negatively controls sporulation. NprR inhibits expression of Spo0A-regulated genes by preventing the KinA-dependent phosphorylation of the phosphotransferase Spo0F, thus delaying initiation of the sporulation process. This NprR function displays striking similarities with the Rap proteins, which also belong to the RNPP family, but are devoid of DNA-binding domain and indirectly control gene expression via protein-protein interactions in Bacilli. Conservation of the Rap residues directly interacting with Spo0F further suggests a common inhibition of the sporulation phosphorelay. The crystal structure of apo NprR confirms that NprR displays a highly flexible Rap-like structure. We propose a molecular regulatory mechanism in which key residues of the bifunctional regulator NprR are directly and alternatively involved in its two functions. NprX binding switches NprR from a dimeric inhibitor of sporulation to a tetrameric transcriptional activator involved in the necrotrophic lifestyle of B. thuringiensis. NprR thus tightly coordinates sporulation and necrotrophism, ensuring survival and dissemination of the bacteria during host infection.
| Bacillus thuringiensis is an entomopathogenic bacterium used worldwide as biopesticide. Its life cycle in insect larvae, which includes virulence, necrotrophism and sporulation, is regulated by cell-cell communication systems involving sensor proteins directly regulated by re-internalized peptide pheromones. After toxaemia caused by pore-forming Cry toxins, the PlcR sensor activates the production of virulence factors leading to insect death. B. thuringiensis then shifts to a necrotrophic lifestyle preceding sporulation. Previously, we showed that this process is regulated by the sensor NprR, which, in the presence of its cognate signaling peptide NprX, adopts a tetrameric conformation allowing its binding to specific DNA sequences and transcription of genes involved in survival of the bacteria in insect cadavers. Here, we demonstrate that, in the absence of NprX, NprR is a dimer, which negatively controls sporulation, independently of its transcription factor activity. We show that NprR prevents the phosphorylation of the phosphoprotein Spo0F and inhibits the phosphorylation cascade regulating sporulation. This demonstrates that NprX binding switches the bifunctional sensor NprR from a dimeric sporulation inhibitor to a tetrameric transcription factor. By establishing a close coordination between cell density, necrotrophism and sporulation, this communication system benefits a pathogenic bacterium feeding on death matter like B. thuringiensis. NprR is found in all strains of the B. cereus group, including B. anthracis and B. cereus involved in food poisoning. Our results may provide new insights for controlling the development and the survival of these undesirable bacteria.
| Sporulating bacteria have developed a number of sophisticated mechanisms devoted to the temporal and spatial coordination of cell fate and gene expression. Regulatory mechanisms, as two-component systems, contribute to maintain this tight coordination by tuning gene expression and protein activation in response to a large variety of environmental stimuli including nutrient limitation and population density [1]. In Gram-positive bacteria, quorum sensing is a mode of cell-cell communication involving the secretion of diffusible signaling peptides recognized by the responder bacteria [2, 3]. These peptides elicit a response either directly, by binding to effector proteins or indirectly, by triggering a two-component phosphorelay.
The effectors of direct quorum-sensing systems are grouped in the RNPP protein family [4] named according to the first identified members: the Rap proteins of Bacillus subtilis [5, 6], the NprR and the PlcR transcriptional activators of Bacillus cereus and B. thuringiensis [7, 8] and the Enterococcus faecalis sex pheromone receptor PrgX [9, 10]. Recently the RNPP family was extended with the characterization of a new member identified in Streptococci, the SHP pheromone receptor Rgg [11]. These regulators are characterized by a conserved peptide-binding domain consisting of 6 to 9 copies of tetratricopeptide repeats (TPRs). Individual TPR motifs are degenerated sequences of ~34 residues composed of two anti-parallel α-helices. Their assembly results in a right-handed superhelix creating the peptide-binding pocket [12, 13]. Binding of the signaling peptides regulates the activity of their cognate quorum sensor [8, 10, 14, 15]. Except for the Rap proteins, RNPPs also contain an N-terminal helix-turn-helix (HTH)-type DNA-binding domain [16] and act as transcriptional regulators. The Rap proteins bind and regulate target proteins via an N-terminal alpha-helical domain replacing the HTH motif. In the complex with the target protein, this N-terminal domain displays a 3-helix bundle conformation, whereas it forms two additional TPR motifs in the peptide-bound form [17–20].
Interestingly, NprR contains both the HTH DNA-binding domain and the two Rap-like additional TPR motifs, and was thus proposed as an evolutionary intermediate in the RNPP family [4, 8]. We already demonstrated that in the presence of its cognate signaling peptide NprX, NprR acts as a major transcriptional regulator. The NprR-NprX complex regulates at least forty-one genes encoding proteins involved in the necrotrophic lifestyle of Bacillus thuringiensis, allowing the bacterium to survive in the insect cadaver [21]. A recent structure-function analysis demonstrated that the NprR-NprX complex adopts a tetrameric structure, suggesting that NprR binds simultaneously to two DNA-binding sites [22]. In the absence of bound peptide, NprR dissociates into dimers and loses its ability to bind DNA. Two types of interfaces respectively involved in dimer and tetramer formation were identified.
The sequence similarity with the Rap proteins led us to hypothesize that the dimeric form of apo NprR could act as a Rap-like indirect regulator of the sporulation process. The Rap family has been characterized in B. subtilis and B. anthracis and several members have been shown to prevent sporulation initiation by stimulating the dephosphorylation of the phosphotransferase Spo0F-P [5, 23, 24]. This phosphatase activity thus interrupts the phosphorelay regulating Spo0A, the key transcription factor, which controls early stationary phase and sporulation gene expression [25, 26].
It has been recently proposed that NprR is a positive activator of sporulation initiation and that the binding of the signaling peptide NprX activates the sporulation function of NprR [27]. These results, suggesting a different mechanism for Rap and NprR, were very surprising and in sharp contradiction with our data. As we show here, NprR acts as a Rap phosphatase, negatively affecting the commitment to sporulation in the absence of bound peptide and independently of its transcriptional activator function. NprR prevents expression of Spo0A-regulated genes and delays the initiation of the sporulation process by dephosphorylating Spo0F. In addition, we show that the dimeric apo form of NprR is highly flexible, as observed in the Rap proteins [18]. Finally, we propose a molecular mechanism for the NprR effect on sporulation. We explain how NprX binding promotes dissociation of Spo0F, formation of the NprR tetramer and DNA binding. Taken together, these results thus demonstrate that NprX regulates the switch from an indirect sporulation inhibitor function to a transcriptional activator function. By integrating an enzymatic activity and a regulatory function in a single molecule, NprR links adaptation properties and cell fate, thus allowing an accurate synchronization of important functions of the cell.
In view of the similarity between the NprR and Rap proteins [4, 22], we investigated whether NprR could play a role in the sporulation process. We compared the sporulation rates of the parental B. thuringiensis 407 wild-type strain with the NprR/NprX-deficient strain (ΔRX), the NprR-deficient strain (ΔR) and the NprX-deficient strain (ΔX) (Fig 1). In the growth conditions used in this study (LB medium at 30°C), the OD600 at the onset of stationary phase (t0) was similar for all the strains (S1 Table). However, the percentage of sporulation of the ΔRX mutant strain was lower (46 ± 2.6%) than that of the wild type (60 ± 1.3%). A similar sporulation rate (48 ± 2.8%) was measured in the ΔR mutant strain demonstrating that NprX alone does not affect sporulation. In sharp contrast, the sporulation rate of the ΔX mutant strain was significantly reduced to 7 ± 1.5%, and the total number of viable bacteria of this mutant strain was 300-fold lower than in the wild-type strain. These results indicate that the significant proportion of bacteria, which did not enter into the sporulation process was unable to survive starvation. It results that the production of heat-resistant spores was reduced by 1200- and 2500-fold compared to the ΔRX and wild-type strains, respectively. These results suggest that NprR acts like a Rap-like protein negatively affecting sporulation in the absence of NprX peptide. The sporulation efficacy of a ΔX strain carrying the X7i gene (designated ΔX amy::X7i), which encodes the heptapeptide SKPDIVG corresponding to the minimal active form of NprX [8] was fully restored when the expression of X7i was induced (Fig 1). These results thus demonstrate that NprX inhibits the Rap-like function of NprR on sporulation and activates its transcriptional function.
To determine whether the role of NprR on sporulation is independent of its transcriptional activator function, we constructed an nprR* gene with a non-functional HTH domain. Based on the weight matrix established by Dodd and Egan for the evaluation of potential HTH motifs [28] we introduced the M20A, T21A and Q22A substitutions in the N-terminal HTH domain of the protein (NprR*). Moreover, sequence alignment with the HTH domain of PlcR and analysis of the PlcR/DNA interactions (PDB ID 3U3W) [29] showed that the MTQ motif is conserved and belongs to PlcR helix α2, which interacts with the sugar-phosphate backbone of the bound DNA, thus participating to stabilization of the complex (S1 Fig). The nprR* and nprR*-nprX genes were inserted at the amy locus of the ΔRX mutant and these strains were designated ΔRX amy::R* and ΔRX amy::R*X, respectively. No ß-galactosidase production was detected in the strain ΔRX amy::R*X harbouring PnprA’Z (S2A Fig), thus demonstrating that the mutations introduced in the HTH domain of the NprR* protein result in the complete loss of its transcriptional activation function. We then tested the effect of the NprR* protein on sporulation efficiency (Fig 1). As described above for the ΔX strain, a sporulation-defective phenotype was observed with the ΔRX strain harbouring the nprR* gene (sporulation rate reduced from 46 ± 2.6% to 5 ± 0.6%). Moreover, the production of viable spores in the ΔRX amy::R* strain was 5-log lower than in the wild-type strain. As with native NprR, the sporulation efficiency of the strain producing NprR* was restored when NprX was produced. Taken together, these results demonstrate that the NprR protein with a non-functional HTH domain is unable to activate nprA expression but preserves its negative effect on sporulation in the absence of NprX. Therefore, the two NprR functions are dissociated: the effect of NprR on sporulation is not due to a pleiotropic effect of its functions as transcriptional activator.
The initiation of sporulation is regulated by the phosphorylated state of Spo0A (Spo0A-P), the master regulator of sporulation [30]. To determine whether NprR, as the Rap phosphatases in B. subtilis, affects the phosphorylation of Spo0A in B. thuringiensis, we examined the activity of the spoIIE promoter, known to be under the direct control of Spo0A-P in B. subtilis [31]. A Spo0A box is present in the spoIIE promoter of B. thuringiensis and the spoIIE’-lacZ transcriptional fusion was not expressed in a 407 Δspo0A mutant strain (Fig 2A), thus suggesting that spoIIE transcription is directly controlled by Spo0A in B. thuringiensis as in B. subtilis. We measured PspoIIE-directed ß-galactosidase expression in the wild type, ΔRX and ΔX strains. In the ΔX strain as in the Δspo0A mutant strain, no spoIIE expression was observed whereas the ΔRX mutant behaved as the wild type (Fig 2A), suggesting that NprX inhibits the negative effect of NprR on spoIIE expression. We also verified that spoIIE expression was inhibited in the ΔRX strain complemented with nprR*, and that the spoIIE expression pattern was restored when NprX was produced (Fig 2B). Altogether, these results show that, in the absence of NprX, NprR inhibits sporulation independently of its transcriptional activator function by negatively affecting the transcription of Spo0A-regulated genes.
To investigate whether the role of NprR on sporulation depends on the phosphorylated state of Spo0A we used the Spo0ASad67 mutant (designated 0A67), which is constitutively active independent of its phosphorylation state in B. subtilis [32]. Thus, the 0A67 bypasses the requirement of the phosphorelay to trigger sporulation. Due to the high level of conservation between the B. thuringiensis and B. subtilis spo0A genes [33], we assumed that such a mutation would have the same effect in these two Bacillus species. We verified the activity of the 0A67 protein by measuring the sporulation efficiency of the Δspo0A strain harbouring the pHT-0A67. In the Spo0A-deficient strain, no heat-resistant spores were detected and the sporulation phenotype of the Δspo0A harbouring the pHT-0A67 was partially restored when 0A67 was expressed (Fig 2C). The sporulation rate and the total production of heat-resistant spores were fully restored in the ΔX strain harbouring the pHT-0A67 plasmid (Fig 2C). These results show that the 0A67 protein, which bypasses the requirement of the phosphorelay, is able to trigger sporulation despite the absence of NprX. These results clearly demonstrate that the inhibition of sporulation by NprR relies on an action on the sporulation phosphorelay.
In order to confirm that NprR acts like a Rap-like protein by targeting the phosphotransferase Spo0F, we used microscale thermophoresis to test the interaction of NprR with Spo0F from B. thuringiensis (Fig 3A). Our results confirmed direct binding between NprR and Spo0F with an apparent Kd value of about 5 μM. This interaction was lost in the presence of NprX, thus demonstrating the specificity of the NprR-Spo0F interaction and the regulatory role of the NprX peptide. The Rap proteins preferentially bind Spo0F-P and act as phosphatases [5]. Because no sporulation kinase has been characterized in the B. thuringiensis genome, we used the KinA-Spo0F system from B. subtilis to test if NprR displays the same mechanism, as already done to test the activity of Rap proteins from B. anthracis [24]. These results demonstrate that the KinA-dependent phosphorylation of Spo0F is reduced in the presence of NprR. Indeed, densitometry analysis revealed that, after 20 minutes of incubation in the presence of NprR, the phosphorylation of Spo0F was at least 3-fold lower compared to the control (band volume reduced from 64202 to 20287 arbitrary unit). Furthermore, addition of NprX restores the phosphorylation signal (Fig 3B). Taken together, these results thus demonstrate that NprR prevents the phosphotransfer from KinA to Spo0F and that peptide binding inhibits this activity.
Our previous structural analysis demonstrated that the peptide-bound tetrameric crystal structure of NprR displayed a compressed conformation of the TPR-superhelix whereas the SAXS analysis of NprR suggested that the apo dimer would adopt a more extended conformation similar to the 3-helix bundle conformation of the Rap proteins [22]. To confirm this hypothesis and go further in the understanding of the sporulation inhibitor function of NprR, we needed to solve the crystal structure of the protein alone. As already observed in our previous crystallisation trials of the NprR-NprX complex [22], no crystals could be obtained using full-length NprR. We thus used the truncated NprR(ΔHTH) form of the protein, deleted of the HTH domains, which are known to be highly flexible in the absence of DNA. Because crystallisation trials are performed at very high protein concentration favouring the tetrameric form of NprR, we also used the (Y223A/F225A) mutant that has been shown to impair tetramer formation [22]. Crystals were finally obtained with the truncated double mutant protein NprR(ΔHTH)(Y223A/F225A). They diffracted up to 3.25Å resolution in space group P 21 21 21 with 2 molecules per asymmetric unit. The crystallographic phase problem was solved by the SAD method using a selenomethionine-labelled form of the protein. Atomic coordinates and structure factors have been deposited in the Protein Data Bank (PDB ID 5DBK).
In both polypeptide chains contained in the asymmetric unit, the 18 α-helices and the connecting loops forming the 9 TPR motifs of NprR are well defined. Only some residues (up to seven) at the N- and C-terminal extremities of NprR(ΔHTH) as well as the C-terminal His-tag were not visible in the electron density. The two subunits (called apoA and apoB) display the right-handed superhelix conformation characteristic of TPR domains [13]. They are associated into a stable dimer (Fig 4A) characterized by an overall interface area of 1113Å2 and a solvation free energy gain ΔiG of -18 kcal/mol. The contacts involve 5 H-bonds and 2 salt bridges as well as hydrophobic interactions. In particular, the dimerization interface involving the C-terminal residues N407 and Y410, which was observed in the NprR/NprX complex, is conserved in the apo dimer. Interestingly, the NprR mutant N407A/Y410A, which has been shown to be monomeric and inactive as a transcription factor [22], has also lost its sporulation inhibitor activity (Fig 4B), demonstrating that NprR dimerization is essential for Spo0F binding.
Comparison of the two subunits of the apo NprR dimer demonstrated that, due to distinct crystal packing constraints, the pitch of the superhelix formed by the 9 TPR motifs of the protein is larger in apoB than in apoA (Fig 4C). This discrepancy, characterized by a Z-score of 7.9 and an rmsd distance of 1.99Å over 293 aligned Cα atoms, reflects the elasticity of the superhelix and suggests that, in solution and in the absence of NprX, the protein dynamically oscillates between distinct conformations. Comparison with the subunits of the symmetrical NprR-NprX tetramer (PDB ID 4GPK) (Fig 4C) showed that both apoA (Z-score = 7.7; rmsd = 2.31Å over 301 Cα atoms aligned with 4GPK-J) and apoB (Z-score = 5.1; rmsd = 2.74Å over 284 Cα atoms aligned with 4GPK-K) display a more extended conformation than the peptide-bound structure.
This flexibility of the TPR superhelix is also observed in the available structures of the Rap proteins. Peptide binding stabilizes a compressed form of the Rap superhelix similar to that observed in the NprR-NprX complex (PDB-ID 4GPK) [22] (Z-score = 9.3; rmsd = 2.26Å over 290 Cα atoms of chain 4GPK-F aligned with chain 4I9C-A of the RapF-PhrF complex [20]) (Z-score = 8,5; rmsd = 2.33Å over 288 Cα atoms of chain 4GPK-K aligned with chain 4GYO-A of the RapJ-PhrC complex [19]). On the other hand, binding of the Rap protein target stabilizes the 3-helix bundle conformation (structures of the RapH-Spo0F, PDB ID 3Q15 [18] and RapF-ComA, PDB ID 3ULQ [17] complexes). Interestingly, the available structures of the apo form of Rap proteins display distinct conformations. Apo RapI (PDB ID 4I1A, [19]) displays an extended TPR superhelix conformation similar to NprR apo-B (Z-score = 6.4; rmsd = 3.30Å over 307 aligned Cα atoms aligned with chain 4I1A-B). In turn, apo RapF displays the 3-helix-bundle conformation (PDB ID 4I9E, [20]) (S3 Fig). This demonstrates that the conformational change is not induced by protein binding but intrinsic to the flexibility of the TPR-superhelix.
The structural similarity between NprR and the Rap proteins, and in particular the conserved flexibility of the TPR superhelix, strongly suggests that NprR could adopt a 3-helix bundle-like conformation to carry its Rap-like phosphatase activity on Spo0F-P. NprR would thus follow the same molecular mechanism than the Rap phosphatases, despite the additional presence of the HTH domain at the N-terminus. We thus propose that NprR apoA and apoB represent two intermediate conformations of the protein, between the compressed TPR superhelix of the peptide-bound complex and the 3-helix bundle conformation of the Spo0F-bound form.
Structure based-sequence alignments (S4 Fig) showed that the RapH residues Q47, F58, E137 and Y175, which have been shown to be directly involved in Spo0F binding [18], correspond to NprR residues D107, Y118, E188 and Y223, respectively. We mutated these NprR residues into alanine and tested the ability of the resulting mutant proteins to inhibit sporulation. The four point mutations abolished the negative effect of NprR on sporulation (Fig 5A), confirming a Rap-like Spo0F binding mode. Because Y223, together with residue F225, is part of the interface stabilizing the tetrameric conformation of the NprR-NprX complex [22], we also mutated F225, which is not conserved in RapH, and measured its activities. Both single mutants NprR(Y223A) and NprR(F225A) had lost their transcriptional activity (S2B Fig) but the NprR(F225A) mutant retained its sporulation inhibitor activity (Fig 5A). Taken together these results demonstrate that the NprR residue Y223 is alternatively involved in tetramer formation and in Spo0F interaction, depending on the presence or absence of bound peptide, respectively, whereas F225 is only involved in stabilization of the DNA-binding tetramer. We thus propose that: i) Spo0F binding most probably stabilizes apo NprR in a Rap-like 3-helix bundle conformation, and ii) NprX binding stabilizes the compressed conformation of the TPR superhelix compatible with tetramer formation and DNA binding [22]. However, since only the truncated form of the protein could be crystallized, the position of the additional N-terminal HTH-domain of NprR remains unknown.
In our previous structural study of the NprR-NprX complex we showed that peptide binding involves residues from TPR-2 (residue R126) to TPR-7 (residues N306, D309). In particular, the importance of a stacking interaction between the K residue of the bound SSKPDIVG octapeptide and the NprR residues H201 and W167 has been demonstrated [22]. We show here that this interaction further induces an additional stacking interaction between residues Y165 from TPR-3 and R343 from TPR-8 (Fig 5B). Dissociation of the peptide releases this constraint, resulting in a dynamic flexibility of the TPR superhelix. In the relaxed conformation of the superhelix observed in NprR apoB, residues Y165 and R343 are about 11Å apart (Fig 5C). The role of the Y165-R343 interaction has been investigated by mutational analysis. A ß-galactosidase assay showed that these residues are necessary for the transcriptional activity of NprR (S2C Fig), thus confirming their role in stabilization of the closed transcription factor conformation of the protein. Interestingly, the sporulation inhibitor activity of the NprR mutant proteins (Y165A) and (R343A) was also abolished (Fig 5A), suggesting that these residues could also play a role in regulating Spo0F binding.
Furthermore, mutating into alanine residue R126, which had been shown to be essential for peptide binding [22], abolishes the inhibitory effect of NprR on sporulation (Fig 5A), suggesting that this residue is also alternatively involved in Spo0F binding.
Unlike the results published by Cabrera and colleagues [27] claiming that NprR has a positive effect on sporulation, our results demonstrate that NprR is a bifunctional regulator repressing sporulation in the absence of its cognate peptide NprX and acting as a transcriptional activator in the presence of peptide. These findings are in agreement with previous studies suggesting that NprR is an evolutionary intermediate between the Rap and the transcriptional activators of the RNPP family [4, 8].
We show that the mechanism involved in the negative control of sporulation is due to a second regulatory function of NprR, independent of its transcriptional activator function. Moreover, expression of the Spo0A-regulated genes spoIIE and spoIIG is prevented in the NprX-deficient strain, demonstrating that NprR affects Spo0A phosphorylation in the absence of signaling peptide. This new function of NprR is similar to the function of the Rap phosphatases, which interact with Spo0F-P in the absence of the Phr peptides [19, 34]. Dephosphorylation of Spo0F-P disrupts the phosphorelay therefore blocking the activation of Spo0A and the commitment to sporulation. Here we demonstrate that NprR binds to Spo0F and prevents its KinA-dependent phosphorylation, thus repressing the expression of Spo0A-regulated genes and finally inhibiting sporulation. This activity is relieved by NprX binding. According to these results, the NprR-NprX system thus functions as the Rap-Phr systems. However, the sporulation efficiency of the NprR/NprX-deficient strain was slightly lower than the wild type strain contrary to the Rap/Phr-deficient strains in B. subtilis and B. anthracis [24, 35]. The transcriptional activity of the NprR/NprX complex regulating genes coding for several degradative enzymes providing the nutrients required for bacterial survival and sporulation [21] could explain this result.
In the Rap proteins, the 3-helix bundle conformation has been shown to be required for binding of the target protein, Spo0F or ComA [19]. Our structural analysis suggests that NprR, despite the presence of the additional HTH domain at the N-terminus, most probably undergoes a similar conformational change for Spo0F binding.
It is also noteworthy that, in the Rap proteins where dimerization is not universally observed in solution [17, 23], the analysis of the assemblies and interfaces using PISA [36] revealed that all available crystal structures display a dimerization mode similar to NprR (S3B Fig). This interface could be less important and weaker in the Rap proteins than in the other RNPP regulators, which require dimerization for DNA binding. However, detailed analysis of the dimerization interface of RapF revealed that it involves residues equivalent to NprR residues, which have been shown to play important roles in dimerization. In particular, the NprR residues H205 and H206 are equivalent to RapF residues Y158 and F159. In the 3-helix bundle conformation of the RapF dimer (PDB ID 4I9E), symmetrical F159 residues are in stacking interactions and Y158 forms a sandwich interaction involving Y117 and R115 from the neighbouring subunit (S5A Fig). In the peptide-bound TPR conformation of RapF (PDB ID 4I9C), the interaction network between symmetrical Y158 and F159 residues is modified (S5B Fig) and Y117 now interacts with the bound peptide and Y153 (S5C Fig). This situation is reminiscent of what is observed in the NprR-NprX complex with residues W167 and H201 (Fig 5A). In addition, R115 interacts with D297, thus stabilizing the TPR conformation of RapF (S5C Fig), as observed in the NprR-NprX complex between residues Y165 and R343 (Fig 5C). Moreover, the conserved tryptophan residue corresponding to RapH W17 that has been shown to be important for the folding of the 3-helix bundle conformation [18] is conserved and corresponds to NprR W79 (S4 Fig). This structural comparison further supports the hypothesis that NprR and the Rap proteins would share the same mechanism and that NprR could adopt the 3-helix bundle conformation for Spo0F binding.
However, understanding how this bifunctional quorum sensor accommodates the presence of additional HTH domain at the N-terminus will require further investigations. We propose that NprR residue R343, which has been shown to be essential for the sporulation inhibitor function of the protein, could be involved in stabilization of the HTH domain in the complex with Spo0F. The drastic conformational change resulting in the 3-helix bundle conformation indeed brings the equivalent residue RapF-D297 in the vicinity of helix α1 N-terminus, suggesting that NprR-R343 could interact with the N-terminal HTH-domain of NprR missing in the crystal structure. In the meantime, we propose a molecular regulatory mechanism for NprR (Fig 6A) in which important residues are directly and alternatively involved in the transcription factor and sporulation inhibitor functions of NprR. In particular, Y223 is alternatively involved in tetramerization and Spo0F binding. Similarly, we propose that residues Y165 and W167 alternatively stabilize the NprX-induced tetrameric conformation and the 3-helix bundle dimer compatible with Spo0F binding.
From a physiological point of view, we propose (Fig 6B) that during the transition phase between the end of exponential growth and the onset of sporulation, NprR is predominantly non-associated with NprX and present in a flexible dimeric form compatible with the conformational change required for Spo0F binding. The sporulation kinase is thus unable to phosphorylate the NprR-bound Spo0F. The phosphorylation cascade is inhibited and Spo0A is kept in an inactive state. As a consequence, the PlcR regulon is maintained, resulting in the production of extracellular virulence factors such as haemolysins, degradative enzymes and enterotoxins [37]. This allows the bacteria to be maintained in a virulence state ending by host death [3]. At this stage, the bacterial density in the host cadaver increases to a plateau and the intracellular concentration of matured-NprX is sufficient to lock NprR in the tight superhelix conformation incompatible with Spo0F binding, thus activating the sporulation phosphorelay. Concomitantly, the NprR-NprX complex tetramerizes and binds to DNA, thus activating the transcription of the NprR-regulated genes involved in the necrotrophic lifestyle of the bacteria [21].
Recently, a study on the dynamics of cell differentiation in B. thuringiensis population during growth in various conditions demonstrated that commitment to sporulation depends on the activity of NprR as a transcriptional activator [38]. This is in agreement with the new role of NprR-NprX in the modulation of the Spo0A phosphorylation rate suggesting that the Spo0A-P concentration required to initiate the sporulation process is reached only in bacteria engaged in the necrotrophism. This mechanism coupling necrotrophism with spore formation would ensure the survival and the dissemination of the bacteria during the infection process.
Our work thus revealed unexpected properties of NprR and NprX concerning their role in the developmental program of bacteria belonging to the B. cereus group. Altogether our results indicate that the NprR-NprX regulation system functions as a typical Gram-positive quorum-sensing system in which NprX is the signaling peptide. However, a striking difference between NprR-NprX and other known systems, including PlcR-PapR and Rap-Phr, is the dual function of the regulator protein: 1) In the absence of NprX, NprR prevents Spo0A to be phosphorylated by inhibiting phosphorylation of Spo0F; and 2) In the presence of bound NprX, NprR acts as a transcriptional regulator. Therefore, NprX acts as a switch, which toggles NprR from one activity (NprR = developmental function) to another (NprR-NprX = transcriptional activator). These two regulatory activities control completely distinct pathways: NprR combined to NprX activates the transcription of a set of genes involved in the necrotrophic lifestyle whereas NprR alone inhibits sporulation events. Bifunctional regulators, also designated as moonlighting proteins, have been described in prokaryotes and eukaryotes [39, 40]. The RapH phosphatase acts both on Spo0F to inhibit the sporulation phosphorelay and on ComA to inhibit its binding to DNA [41]. However, NprR is the first example of a quorum-sensor that recognizes DNA for its transcription factor activity and a protein for its sporulation inhibitor activity. This bifunctional regulator may have been selected in B. cereus group because it provides benefits to a pathogenic bacterium feeding on host proteins. The pathogenic lifestyle may require close coordination between cell density, proteolytic activities and sporulation events.
The B. thuringiensis strain 407 Cry- (hereafter referred to as Bt 407 strain) is an acrystalliferous strain cured of its cry plasmid [42]. Escherichia coli K-12 strains TG1 was used as host for the construction of plasmids and cloning experiments. Plasmid DNA for Bacillus electroporation was prepared from the Dam- Dcm- E. coli strain ET12567 (Strata gene, La Jolla, CA, USA). E. coli and B. thuringiensis cells were transformed by electroporation as described previously [42, 43]. E. coli strains were grown at 37°C in Luria Broth (LB). Bacillus strains were grown at 30°C in LB or at 37°C in HCT, a sporulation-specific medium [44].
The following concentrations of antibiotic were used for bacterial selection: ampicillin 100 μg/ml for E. coli; tetracycline 10 μg/ml, spectinomycin 200 μg/ml and erythromycin 10 μg/ml for B. thuringiensis. Bacteria with the Lac+ phenotype were identified on LB plates containing 100 μg/ml X-gal. The xylA promoter in B. thuringiensis was induced by adding 20 mM xylose to the culture medium.
The sporulation efficiency of the B. thuringiensis strains was determined in LB medium after 3 days of growth. Numbers of viable cells were counted as total colony-forming units (cfu) on LB plates. The number of spores was determined as heat-resistant (65°C for 30 min) cfu on LB plates. The percentage of sporulation was calculated as 100 × the ratio between the numbers of heat-resistant spores ml−1 and viable bacteria ml−1. The OD600 at the onset of the stationary phase is similar for all tested strains (ranging from 2.1 to 2.6). Results are given as mean ± standard error of the mean (SEM). The experimental values are given in S1 Table.
Chromosomal DNA was extracted from B. thuringiensis cells using the Puregene DNA Purification Kit (QIAgen, France). Plasmid DNA was extracted from E. coli by a standard alkaline lysis procedure using QIAprep spin columns (QIAgen, France). DNA polymerase, restriction enzymes and T4 DNA ligase (New England Biolabs, USA) were used in accordance with the manufacturer’s recommendations. Oligonucleotide primers (S2 Table) were synthesized by Sigma-Proligo (France). PCRs were performed in an Applied Biosystem 2720 Thermak cycler (Applied Biosystem, USA). Amplified fragments were purified using the QIAquick PCR purification Kit (QIAgen, France). Digested DNA fragments were separated on 1% agarose gels and extracted from gels using the QIAquick gel extraction Kit (QIAgen, France). Nucleotide sequences were determined by Beckman Coulter Genomics (Takeley, U K).
The plasmid pRN5101 [45] was used for nprX disruption. The plasmid pMAD-amy::spc [38] was used for complementation experiment at the amy locus. Transcriptional fusions for the spoIIE promoter region was constructed in pHT304-18’lacZ [46]. The low-copy-number plasmid pHT304.18-PxylA [15] was used for complementation experiment with the modified spo0Asad67 gene under xylose-inducible promoter. The vectors pQE60 and pQE30 (QIAgen, France) were used to overproduce NprR-6His and 6His-Spo0F from Bt 407, respectively. The pQE30 was also used to overproduce 6His-Spo0F and 6His-KinA from Bacillus subtilis for NprR Spo0F-P dephosphorylation assays. All the constructed plasmids used in this study are described in S3 Table.
The thermosensitive plasmid pRN5101ΩnprX::tet (S3 Table) was used to disrupt the chromosomal wild-type copy of nprX by homologous recombination as described previously [47]. The recombinant strain, designated Bt 407 ΔX, was resistant to tetracycline and sensitive to erythromycin. The Bt 407 nprR-nprX::tet (Bt 407 ΔRX), Bt 407 nprR::tet (Bt 407 ΔR) and Bt 407 spo0A::kan (Bt 407 Δ0A) mutant strains were described previously [8, 33].
ß-Galactosidase activities were measured as described previously [8], and specific activities are expressed in units of ß-galactosidase per milligram of protein (Miller units). Each assay was independently repeated at least three times and a representative graph was shown for each experiment.
The (ΔHTH)(Y223A/F225A) mutant of the nprR gene from B. thuringiensis strain 407 has been cloned into a pQE60 plasmid and expressed in E. coli strain M15 [pRep4] with a C-terminal 6xHis tag. The recombinant protein was purified as already described for the full-length wild-type NprR [8]. The purified protein was aliquoted, flash frozen and stored at -20°C in 20 mM Tris-HCl pH8 and 100 mM NaCl.
Selenomethionine-labelled NprR(ΔHTH))(Y223A/F225A) was produced in E. coli strain M15 [pREP4, pQE60ΩnprRΔHTH-Y223A-F225A] grown at 37°C in the presence of 50 μg/ml ampicillin and 50 μg/ml kanamycin in M63 medium. At OD600nm of 1 the M63 medium was supplemented with 10 g/l of Studier amino acid and 2.5 g/l of selenomethionine. Finally, expression was induced at OD600nm of 1.5 by adding 1 mM IPTG and the culture was incubated for 4 hours at 37°C. The labelled protein was purified following the same protocol as for the unlabelled protein.
Spo0F from B. thuringiensis was produced as an N-terminal His-tagged recombinant protein in E. coli strain NMA522 [pQE30Ωspo0F] grown at 37°C in the presence of 50 μg/ml of ampicillin. Expression was induced at OD600nm of 0.6 by adding 1 mM IPTG and the culture was incubated for 3 hours at 37°C. The protein was produced in inclusion bodies, inclusion bodies were resuspended in 25 mM Tris-HCl pH 8, 500 mM NaCl, 3 M Guanidinium Chloride and 10% Glycerol and incubated for 3 hours at 4°C. Finally Spo0F was renatured on the IMAC column following the protocol of the manufacturer. The purification was completed by a size exclusion chromatography step using a S75 16/60 column equilibrated in 20 mM Tris-HCl pH 8, 200 mM NaCl, 10% Glycerol.
Spo0F and KinA from B. subtilis were produced as N-terminal His-tagged recombinant proteins in the E. coli M15 [pRep4] strain grown at 37°C in the presence of 50 μg/ml ampicillin and 50 μg/ml kanamycin. Expression was induced at OD600nm of 0.6 by adding 1 mM IPTG and the cultures were incubated overnight at 15°C. The Bs Spo0F and KinA proteins were then purified using an IMAC column following the protocol of the manufacturer (QIAGEN). The purification was completed by a size exclusion chromatography step using a S75 16/60 and a S200 16/60 column, for Spo0F and KinA respectively. The samples were stored frozen in 50 mM Tris-HCl pH 7.5, 50 mM Na2SO4, 15% Glycerol, 5 mM MgCl2 for Spo0F and KinA.
Microscale thermophoresis (MST) [48] was performed using the Monolith NT.115 apparatus and MST Premium coated capillaries from Nanotemper Technologies. NprR was fluorescently labelled with the blue fluorescent dye NT-495-NHS according to the manufacturer’s protocol. The Bt Spo0F solutions were serially diluted from about 260 μM to 8 nM in the presence of 33 nM labelled NprR and 5% glycerol in MST buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 10 mM MgCl2, 0.05% Tween-20). Data analyses were performed using the Nanotemper Analysis software.
Phosphorylation of B. subtilis Spo0F was determined in a reaction buffer (50 mM Tris-HCl, pH 7.4, 20 mM MgCl2, 0.1 mM EDTA and 5% glycerol). KinA (0.1 μM) was first activated by pre-incubation (1.5 hours at 37°C) in the presence of 1 mM ATP, containing 20 μCi mmol-1 [γ-32P]-ATP. Spo0F (5.4 μM), NprR (10 μM) and NprX-8 (57 μM) were then added as indicated, and mixtures were further incubated at 37°C. Aliquots were withdrawn at different times (2, 10 and 20 minutes). Samples were analyzed by 15% SDS-Tris-glycine PAGE. The gel was dried and the radioactivity detected with a PhosphorImager and the ImageQuant software (Molecular Dynamics Corp.).
In order to determine the conformation of NprR carrying out the sporulation inhibitor function of the protein, we performed a large-scale screening of crystallization conditions. The truncated double mutant protein NprR(ΔHTH)(Y223A/F225A) crystallized at 18°C in 1.0 M Na citrate, 0.1 M Hepes pH 7.6. Initial needle clusters were improved using micro- and macro seeding. The crystals were flash frozen in the crystallization solution supplemented with 30% glycerol for data collection.
A native data set was collected at 3.25Å resolution on beamline ID29-1 (ESRF, Grenoble, France). A selenomethionine (SeMet) labelled form of the protein was used to collect anomalous diffraction data on beamline Proxima-1 (SOLEIL, Gif-sur-Yvette, France). The derivative data set was collected at a wavelength of 0.97911 corresponding to the maximum of anomalous dispersion f”. Both data sets were processed with the XDS package [49]. Sub-structure determination was performed with the SHELX program suite [50] using the HKL2MAP interface [51]. The positions of 16 from the 22 expected selenium atoms could be identified and helped determine initial crystallographic phases using the program PHASER [52]. The PHENIX wizard [53] was then used for iterative model building, density modification and structure refinement. The final model was refined against the 3.25Å resolution native data set and manually optimized using COOT [54]. Data processing and refinement statistics are given in S4 Table.
We used the Protein structure comparison service PDBeFold at European Bioinformatics Institute [55], to superimpose and compare crystal structures. We used the Protein Interfaces, Surfaces and Assemblies service PISA at the European Bioinformatics Institute [36] to analyze interactions. We used the PyMOL Molecular Graphics System [56] to analyze the 3D structures and prepare Figs 4A, 4C, 5B, 5C, S3 and S5.
The genome of Bacillus thuringiensis 407 used in this study is accessible in the NCBI Reference Sequence (RefSeq) Database under number NC_018877.1.
|
10.1371/journal.pntd.0004016 | Proteomics-Based Characterization of the Humoral Immune Response in Sporotrichosis: Toward Discovery of Potential Diagnostic and Vaccine Antigens | Sporothrix schenckii and associated species are agents of human and animal sporotrichosis that cause large sapronoses and zoonoses worldwide. Epidemiological surveillance has highlighted an overwhelming occurrence of the highly pathogenic fungus Sporothrix brasiliensis during feline outbreaks, leading to massive transmissions to humans. Early diagnosis of feline sporotrichosis by demonstrating the presence of a surrogate marker of infection can have a key role for selecting appropriate disease control measures and minimizing zoonotic transmission to humans.
We explored the presence and diversity of serum antibodies (IgG) specific against Sporothrix antigens in cats with sporotrichosis and evaluated the utility of these antibodies for serodiagnosis. Antigen profiling included protein extracts from the closest known relatives S. brasiliensis and S. schenckii. Enzyme-linked immunosorbent assays and immunoblotting enabled us to characterize the major antigens of feline sporotrichosis from sera from cats with sporotrichosis (n = 49), healthy cats (n = 19), and cats with other diseases (n = 20).
Enzyme-linked immunosorbent assay-based quantitation of anti-Sporothrix IgG exhibited high sensitivity and specificity in cats with sporotrichosis (area under the curve, 1.0; 95% confidence interval, 0.94–1; P<0.0001) versus controls. The two sets of Sporothrix antigens were remarkably cross-reactive, supporting the hypothesis that antigenic epitopes may be conserved among closely related agents. One-dimensional immunoblotting indicated that 3-carboxymuconate cyclase (a 60-kDa protein in S. brasiliensis and a 70-kDa protein in S. schenckii) is the immunodominant antigen in feline sporotrichosis. Two-dimensional immunoblotting revealed six IgG-reactive isoforms of gp60 in the S. brasiliensis proteome, similar to the humoral response found in human sporotrichosis.
A convergent IgG-response in various hosts (mice, cats, and humans) has important implications for our understanding of the coevolution of Sporothrix and its warm-blooded hosts. We propose that 3-carboxymuconate cyclase has potential for the serological diagnosis of sporotrichosis and as target for the development of an effective multi-species vaccine against sporotrichosis in animals and humans.
| Sporotrichosis is a neglected fungal disease of humans and animals that remains a serious public-health problem. Sporothrix infections persist in cats, leading to continued transmission via cat-to-cat and cat-to-human contact. Cats are the major source of transmission of Sporothrix brasiliensis to the human population. We stress the importance of implementing health policies aimed at detecting Sporothrix infection in cats as an attempt to reduce massive zoonotic transmission to humans. Early diagnosis of feline sporotrichosis is critical to recognize outbreaks areas and effectively tackle future spread of the disease among humans. We explored the diversity of molecules that are expressed by S. brasiliensis and S. schenckii and that are recognized by immunoglobulin G. Upon infection, the cat delivers an IgG-mediated response against Sporothrix antigens, similar to the response in murine and human sporotrichosis. We detected remarkable cross-reactivity among S. brasiliensis and S. schenckii antigens, supporting the hypothesis that antigenic epitopes may be conserved among closely related species. One protein, 3-carboxymuconate cyclase, was prominent in immune profiles from infected animals, using both types of Sporothrix antigens. Knowledge of the immune response in feline sporotrichosis is critical to advancing techniques for serological diagnosis, developing vaccines, and improving our understanding of Sporothrix evolution.
| Sporothrix schenckii was originally described in 1898 as the causal agent of a subcutaneous disease in humans in the Mid-Atlantic USA [1]. Subsequently, the disease was reported in rats naturally infected in southeastern Brazil [2] and later in a wide range of animals including dogs, cats, horses, cows, camels, dolphins, goats, mules, birds, pigs, and armadillos. Several Sporothrix spp., previously reported to be closely related to S. schenckii, are known to establish infections in various hosts, with dissimilar virulence traits [3, 4] and responsiveness to antifungal treatment [5].
The S. schenckii complex consists of at least four closely-related species [6, 7], ranging from geographically restricted agents such as S. brasiliensis [8, 9] to cosmopolitan pathogens such as S. schenckii s. str. and S. globosa [7, 10, 11]. Sporothrix spp. are endowed with an extraordinary ecological diversity [12–15]; they are frequently recovered from soil, plants debris, and insects (Coleoptera: Scolytidae). Phylogenetic data support a recent habitat shift within Sporothrix from plants to cats [9] that culminated in the largest epizootic transmission in southeastern Brazil [16–19]. Feline sporotrichosis emerged in the 1990s, with S. brasiliensis recovered from many outbreaks [8, 20]. More recently, S. brasiliensis has been recognized as a threat to humans [21–23] due to the massive zoonotic transmission in southeastern Brazil that affects thousands of patients regardless of whether they are immunocompetent or immunocompromised [9, 24–26].
Cats have been a source of Sporothrix spp. infection transmitted to humans and other animals [18, 19, 27]. Most human cases occurred in housewives and professionals who had contact with infected animals and a history of scratches or bites [21, 28]. The largest epidemic of sporotrichosis due to zoonotic transmission was reported in the State of Rio de Janeiro, Brazil [18, 19, 21, 23, 28]; since then, the incidence of sporotrichosis among animals, particularly cats, has increased [8, 28, 29]. More than 4,000 humans and 4,124 cats were diagnosed at Instituto Nacional de Infectologia (INI) Evandro Chagas /Fundação Oswaldo Cruz by 2012 [30]. Pereira et al. [29] observed that the majority of cats with sporotrichosis in Rio de Janeiro between 2005 and 2011 were male, mongrel, and unneutered, had a median age of 24 months, and presented with three or more cutaneous lesions in non-adjacent locations. This mycosis in cats is hard to treat and generally requires a long period of daily care; these animals do not always respond well to antifungal treatment [30].
Sporothrix is widely distributed in nature, and traumatic inoculation of plant organic matter is classically associated with infection [31]. Feline habits render cats more susceptible to the fungal agent because they are constantly in contact with soil, plant debris, and other cats that may be infected. In cats, a broad spectrum of clinical presentation ranges from single lesions to fatal systemic forms. After monitoring the feline epidemic for 4 years, Schubach et al. [18] observed that the lymphocutaneous form occurred less frequently than did involvement of the mucous membranes of the respiratory tract and upper digestive tract and multiple cutaneous lesions. Some animals present involvement of skin and various internal organs [32].
Cats are susceptible to a variety of fungal infections, including blastomycosis [33], histoplasmosis [34], cryptococcosis [35], candidiasis [36], dermatophytosis [37], aspergillosis [38], coccidioidomycosis [39], and sporotrichosis. Misdiagnosis results in ineffective treatment, which further worsens outcome. Major contributors toward misdiagnosis include the small number of affordable and effective treatment techniques as well as other social issues. Definitive diagnosis of feline sporotrichosis is based on mycological culture, micromorphological characterization, and mold-to-yeast conversion. Histopathological methods and cytopathological examination are useful tools for the presumptive diagnosis of Sporothrix infection in cats [40]. Detection of specific anti-Sporothrix antibodies offers a rapid alternative for accurate diagnosis [41–43]. We recently proposed a serological approach that employs an enzyme-linked immunosorbent assay (ELISA) to diagnose feline sporotrichosis [44] using purified antigen (the S. schenckii ConA binding fraction) and crude antigen, with high sensitivity and specificity.
There is a lack of information about feline sporotrichosis and the antigenic components involved in infection; therefore, the present study aimed to explore the diversity of molecules expressed by closely related species (S. brasiliensis and S. schenckii) and that are recognized by immunoglobulin G (IgG) in sera from cats naturally infected with S. brasiliensis. We found remarkable cross-reactivity among S. brasiliensis and S. schenckii antigens, and we identified an immunodominant molecule that is an important biomarker in feline sporotrichosis, irrespective of clinical manifestation. Here, we show that, although S. brasiliensis and S. schenckii may infect different warm-blooded hosts, infection result in highly similar IgG-mediated response in cats compared to humans, what is important for serodiagnosis and to the development of prophylactic or therapeutic vaccine against the enormous health burden of sporotrichosis in endemic areas. This knowledge may enable selection of potential biomarkers that can be used in seroepidemiological studies, diagnosis, and vaccine development, and may contribute to understanding of the pathogenesis of this infection in cats and humans.
This study was performed in strict accordance with recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Ethics in Research Committee of the Fundação Oswaldo Cruz, Rio de Janeiro, Brazil, under license number L-041/06. This study was also approved by the Institutional Ethics in Research Committee of the Federal University of São Paulo under protocol number 0244/11.
S. brasiliensis and S. schenckii s. str. from cats and humans were used for protein extraction (Table 1). The dimorphic nature of Sporothrix spp. was demonstrated by converting the fungus to its yeast form at 36°C in brain-heart infusion medium (Difco) and observing typical oval multibudding yeast cells. Molecular identification was performed and confirmed via DNA sequencing of the gene encoding calmodulin [25].
Isolates were selected because they had been previously characterized at the molecular level [3, 9, 45, 46]; crude exoantigen (CBS 132974 = Ss118; S. schenckii s. str.) was successfully used to diagnose feline sporotrichosis via ELISA [44]. Sporothrix spp. was grown on Sabouraud medium (Difco) agar slants at room temperature (20–25°C) for 7 days. Approximately 2x106 conidia (≥85% viable cells) were used to inoculate 500-mL flasks containing 150 mL of brain-heart infusion broth. Cultures were incubated at 36°C in a rotary shaker (Multitron II, Infors HT) with constant orbital agitation at 110 rpm for 7 days.
Whole extracts of Sporothrix yeast cells were obtained as described elsewhere [47]. Briefly, yeast cells (4 mL of each culture) were collected via centrifugation at 5000 x g for 10 min at 4°C and washed three times in ultrapure water. Pellets were frozen in liquid nitrogen and disrupted by gridding. Cells were macerated with a pestle until a fine powder was obtained. This cellular powder was vortexed for 30 min at 4°C in Tris-Ca2+ buffer (20 mM Tris-HCl pH 8.8, 2 mM CaCl2) containing a commercial cocktail of protease inhibitors (1:1000; GE Healthcare), RNAse and DNAse (1:1000; GE Healthcare), and 600-μm glass beads (1:1; Sigma). Cell debris and glass beads were removed via centrifugation at 5000 x g for 10 min, and dithiothreitol (final concentration 20 mM) was added to the supernatant [48]. Protein concentrations were determined using the Bradford method [49]. Cell extracts were kept at -80°C until use.
Sera from cats with definitive diagnoses of sporotrichosis (via S. brasiliensis isolation in culture; n = 49) were obtained from INI/Fundação Oswaldo Cruz, Rio de Janeiro, Brazil. The distribution and number of skin lesions of the cats were classified as L1 (cutaneous lesion in only one place), L2 (cutaneous lesion in two non-adjacent places), or L3 (cutaneous lesions in three or more non-adjacent places). During examination, blood was collected via vein puncture and stored in an incubator for 1 h; serum was obtained via centrifugation and stored at -20°C until use. Sera from 19 cats with no evidence of sporotrichosis or other diseases (the control group) were obtained from São Paulo as described elsewhere [44]. Sera from 20 cats with other diseases were also studied to verify cross-reactions with feline infectious peritonitis/coronavirus (5 sera), feline leukemia virus (3 sera), feline immunodeficiency virus (2 sera), leptospirosis (3 sera), rickettsiosis (2 sera), erlichiosis (3 sera), and leishmaniasis (2 sera) as previously described [44] (S1 Diagram). All sera were stored at -20°C until use.
Sera from cats with confirmed sporotrichosis and from cats from the control group were tested via ELISA. To determine the best protein concentration for microplate sensitization, whole cellular proteins from S. brasiliensis (CBS 132990 and CBS 132021) and S. schenckii s. str. (CBS 132974 and CBS 132984) were tested and examined by checkerboard titration for antibody detection. Afterward, all microplates were sensitized with concentrations of 3.6μg/mL (100μL per well in 0.1 M carbonate-bicarbonate buffer, pH 9.6). High binding microtiter plates (Corning Costar, Corning) were sensitized for 2 h at 37°C and overnight at 4°C in a refrigerator. The remaining binding sites were blocked with phosphate-buffered saline containing 0.1% Tween 20 (PBST) and 5% non-fat dry milk (200 μL/well) for 4 h at 37°C. After washing three times with PBST, diluted serum (1:800 in PBST, 100μL/well) was added in duplicate for 1 h at 37°C. Afterward, 100μL horseradish peroxidase-conjugated goat anti-feline IgG (1:1000; Southern Biotech) were added to each well and incubated for 1 h at 37°C. After three washes with PBST, 100μL substrate solution (5 mg of o-phenylenediamine in 25 mL of 0.1 M citrate-phosphate buffer pH 5.0 plus 10 μL 30% H2O2) were added to each well, and the reaction was interrupted after 8 min in the dark by adding 50 μL 4 N H2SO4. Optical density was read at 492 nm with a Tecan Sunrise 96-well Microplate Reader (Tecan).
S. brasiliensis and S. schenckii protein extracts (2 μg) were analyzed via SDS-PAGE with 10% gels [50] and silver-stained [51]. The relative molecular weights of the fractions were estimated using standard broad-range molecular weight markers (Protein Benchmark, Invitrogen).
For immunoblotting, proteins (10 μg) from strains CBS 132990, CBS 132021, CBS 132974, and CBS 132984 were resolved with SDS-PAGE and transferred onto 0.45-μm polyvinylidenedifluoride membranes (Bio-Rad) at 20 V for 30 min with transfer buffer (25 mM Tris base, 192 mM glycine, 20% methanol, pH 8.3) [52] using a Trans-Blot SD semi-dry device (Bio-Rad). Electrotransference was confirmed by staining with 0.15% Ponceau S and 1% acetic acid [vol/vol]. Membranes were destained and free binding sites were blocked overnight in phosphate-buffered saline blocking buffer (1% bovine serum albumin supplemented with 0.05% [vol/vol] Tween 20, 5% [wt/vol] skim milk, pH 7.6) at 4°C. To determine the best dilution of serum, one sample was tested at four dilutions (1:100, 1:200, 1:500, and 1:1000) against yeast extracts. Afterward, for all sera, membranes were probed individually with primary antibody diluted 1:500 at 25°C for 2 h. Membranes were washed three times with Tris-buffered saline (pH 7.5) containing 0.05% [vol/vol] Tween-20 for 10 min and incubated with horseradish peroxidase-conjugated goat anti-feline IgG (1:1000) for 2 h at room temperature. Membranes were then washed with Tris-buffered saline (pH 7.5) containing 0.05% [vol/vol] Tween-20 and signal was detected with an enhanced chemiluminescence detection kit (GE Healthcare). Blots were imaged in a transilluminator (Uvitec Cambridge). Allience 4.7 software was used to take several images at different time exposures, from 2 s each to a total of 10 images over 2 s.
Proteins were separated via 2D gel electrophoresis as previously described [45, 47]. Briefly, proteins (300 μg) were precipitated using the 2D Clean-up Kit (GE Healthcare) and resuspended in rehydration buffer (7 M urea, 2 M thiourea, 2% CHAPS, 1.2% DeStreak, 2% vol/vol isoelectric focusing buffer pH 4–7, and trace amounts of bromophenol blue) to a final volume of 250 μL. Immobilized pH gradient strips (pH 4–7, 13 cm; GE Healthcare) were rehydrated at 30 V for 12 h. Isoelectric focusing was performed at 20°C using a Multiphor III system (GE Healthcare) as follows: 200 V for 2 h, 500 V for 2 h, 1000 V for 5 h, and a gradient applied from 1000 to 5000 V for 2 h. Finally, the voltage was set to 5000 V for 60,000 Vhr. After 1D isoelectric focusing, the IPG strips were reduced for 15 min with 1.5% dithiothreitol and alkylated for 15 min with 2.5% iodocetamide in equilibration buffer (6 M urea, 50 mM Tris-HCl pH 6.8, 30% glycerol, and 2% sodium dodecyl sulfate). Second-dimension separation was carried out by placing equilibrated IPG strips onto 10 % polyacrylamide gels, sealing them with 0.5 % [wt/vol] low-melting-point agarose, and separating the proteins at 10°C using a Hoefer SE 600 unit (15 mA/gel for 30 min and then 23 mA/gel until the dye front reached the bottom of the gel). Proteins were developed with silver staining [51] or were directly transferred for immunoblotting.
For 2D immunoblotting, proteins were transferred onto 0.45-μm polyvinylidenedifluoride membranes at 25 V for 1 h with transfer buffer [52] using the Trans-Blot SD semi-dry system. The success of electrotransference was evaluated by staining with 0.15% Ponceau S and 1% acetic acid 1% [vol/vol]. Membranes were destained and free binding sites were blocked overnight in phosphate-buffered saline blocking buffer (1% bovine serum albumin supplemented with 0.05% [vol/vol] Tween 20, 5% [wt/vol] skim milk, pH 7.6) at 4°C.
Membranes obtained from 2D gels were probed with 1:500 primary antibody (gold standard pooled feline sera; n = 10) under the conditions used for 1D immunoblotting. Immunoreactive antigens were detected using an enhanced chemiluminescence detection kit (GE Healthcare). 2D immunoblots were imaged using the method used for 1D immunoblots.
Diagnostic values included sensitivity, specificity, positive predictive value, and negative predictive value. Receiver operating characteristic (ROC) curves were prepared and analyzed to determine the sensitivity and specificity of each antigen preparation for ELISA. The area under the curve (AUC) for ROC analysis was calculated to evaluate the diagnostic value of ELISA. We assumed that a test lacked diagnostic power when the ROC curve was linear with an AUC of 0.5 (the ROC curve coincided with the diagonal). A powerful test was assumed to yield an AUC of ~1.0, indicating the absence of both false-positives and false-negatives (the ROC curve reached the upper left corner of the plot). To measure the degree of concordance of the results from preparations from strains CBS 132990, CBS 132021, CBS 132974, and CBS 132984, we calculated the kappa statistic and its 95% confidence interval (CI). Kappa values were interpreted as follows: 0.00–0.20, poor agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderate agreement; 0.61–0.80, good agreement; 0.81–1.00, very good agreement [53]. P-values ≤0.05 were considered statistically significant. All calculations were performed with MedCalc Statistical Software version 14.8.1 (MedCalc Software bvba; http://www.medcalc.org). Findings are reported in line with the STARD checklist for studies of diagnostic accuracy (S1 Checklist).
We previously reported a high prevalence of S. brasiliensis in feline sporotrichosis outbreaks [8, 9, 20]. Based on this information, the main goal of the present investigation was to evaluate the presence and diversity of serum-derived antibodies against S. brasiliensis antigens in naturally infected cats. Further, we previously proposed the existence of a convergent humoral response in human sporotrichosis against antigens from S. brasiliensis, S. schenckii, and S. globosa [45]. To establish whether S. brasiliensis and S. schenckii express different antigens, we assessed whole cellular protein extracts from two strains of S. brasiliensis plus two strains of S. schenckii s. str. that were previously characterized by molecular [8, 9, 25, 54] and serological [3, 44–47] methods. Remarkably, and in support of our hypothesis that immunological distance increases with phylogenetic distance, sera from these cats reacted similarly, with no significant differences in titer between ELISA plates coated with proteins from S. brasiliensis or S. schenckii (Fig 1).
ELISA detection of the four antigen preparations exhibited similar results, medians, and ranges for cats infected with sporotrichosis (n = 49): S. brasiliensis CBS 132990, median 1.313 OD, 95% CI 1.262–1.489 OD; S. brasiliensis CBS 132021, median 1.632 OD, 95% CI 1.462–1.714 OD; S. schenckii CBS 132974, median 1.296 OD, 95% CI 1.157–1.442 OD; and S. schenckii CBS 132984, median 1.028 OD, 95% CI 1.027–1.294 OD (S1 Table). When using the assay to diagnosis cats with sporotrichosis, the area under the ROC curve was 1.0 (95% CI 0.94–1.000; P<0.0001; Fig 2), indicating excellent performance. The control group of 19 non-Sporothrix infected animals was associated with lower medians and smaller ranges: S. brasiliensis CBS 132990, median 0.2640 OD, 95% CI 0.2592–0.3098 OD; S. brasiliensis CBS 132021, median 0.2590 OD, 95% CI 0.2517–0.2942 OD; S. schenckii CBS 132974, median 0.2730 OD, 95% CI 0.2512–0.3136 OD; and S. schenckii CBS 132984, median 0.2670 OD, 95% CI 0.2567–0.2907 OD (S1 Table). Differences between the absorbance values for the infected and non-infected groups were statistically significant (P<0.0001). Sera from cats with other infections were non-reactive. Similar cutoff values yielded 100% specificity and sensitivity: S. brasiliensis CBS 132990, 0.377 OD; S. brasiliensis CBS 132021, 0.363 OD; S. schenckii CBS 132974, 0.407 OD; and S. schenckii CBS 132984, 0.346 OD (S1 Fig). ELISA results showed very good agreement for the antigens assayed (kappa = 1.0). To diagnosis feline sporotrichosis via ELISA, we recommend the use of antigen preparations of S. brasiliensis, since this is the most prevalent species in feline sporotrichosis outbreaks.
Antigen diversity was assayed with 1D immunoblots using the four antigen preparations of S. brasiliensis and S. schenckii tested via ELISA. Proteins extracts were evaluated according to the amount of protein extracted, the diversity of bands, the integrity of the samples, and the reproducibility of extraction. Approximately 2 μg of Sporothrix yeast whole-cell extracts were resolved by SDS-PAGE; silver staining revealed numerous proteins ranging from 10 kDa to 160 kDa in size, with different intensities. The Tris-Ca2+ extraction protocol [45, 47] was suitable for the study of Sporothrix antigenic molecules during feline sporotrichosis, yielding samples with high amounts of protein and no degradation.
As expected, antibodies from cats with sporotrichosis reacted with a wide variety of S. brasiliensis and S. schenckii proteins 20kDa to >160kDa in size (Fig 3). Cat-to-cat variation resulted in characteristic banding patterns for each animal (Fig 3); supporting the hypothesis that in a genetically diverse population, the antibody repertoire is expected to vary among individual cats. On the other hand, we detected minor or no differences in IgG-reacting banding patterns between antigen preparations (Fig 3), consistent with the close genetic distance between S. brasiliensis and S. schenckii [8, 9]. Despite this variation, all cats produced antibodies against a 60-kDa molecule in the S. brasiliensis proteome and a 70-kDa molecule in the S. schenckii proteome.
The major antigenic S. brasiliensis molecules (CBS 132990 and CBS 132021) recognized by feline IgG consisted of the following sizes: 60 kDa (100% and 100%, respectively), 90 kDa (92% and 92%, respectively), 100 kDa (86% and 86%, respectively), 38 kDa (60% and 56%, respectively), 40 kDa (56% and 58%, respectively), 45 kDa (44% and 42%, respectively), 30 kDa (36% and 26%, respectively), 52 kDa (30% and 32%, respectively), and 110 kDa (28% and 30%, respectively) (Fig 4A and 4C). Minor molecules recognized by feline IgG had sizes of 80 kDa, 25 kDa, 28 kDa, 120 kDa, 160 kDa, 35 kDa, 20 kDa, 55 kDa, 85 kDa, and 23 kDa (Fig 4A and 4C). The major antigenic S. schenckii molecules (CBS 132974 and CBS 132984) recognized by feline IgG had sizes of: 70 kDa (100% and 100%, respectively), 90 kDa (86% and 88%, respectively), 100 kDa (76% and 82%, respectively), 38 kDa (74% and 62%, respectively), 40 kDa (64% and 56%, respectively), 52 kDa (58% and 56%, respectively), 30 kDa (50% and 34%, respectively), 55 kDa (48% and 48%, respectively), and 45 kDa (40% and 30%, respectively) (Fig 4B and 4D). The minor S. schenckii molecules recognized by feline IgG had sizes of 25 kDa, 80 kDa, 28 kDa, 110 kDa, 120 kDa, 23 kDa, 35 kDa, 160 kDa, 85 kDa, and 20 kDa (Fig 4B and 4D). Sera from uninfected cats did not react with S. brasiliensis or S. schenckii antigens. Sera from cats with other infections were also non-reactive in the immunoblot assay. The frequencies at which Sporothrix molecules were recognized in the antigen preparations are presented in S2 Table. There was no association between the number of bands recognized by each serum and the distribution and number of skin lesions on cats with sporotrichosis.
We previously reported that the 60-kDa and 70-kDa proteins in S. brasiliensis and S. schenckii, respectively, are related to virulence profiles and are the main antigenic molecules during murine [3] and human [45] sporotrichosis. We also determined that this protein undergoes post-translational modification and is present as isoforms and glycoforms in the S. brasiliensis and S. schenckii proteomes [45]. We therefore investigated whether antibodies present in cat sera recognize all six isoforms in the S. brasiliensis proteome, as previously shown with human antibodies [45]. S. brasiliensis proteins were therefore resolved via 2D electrophoresis and immunoblotted with pooled sera from cats with sporotrichosis (n = 10) and optimal antibody titers according to ELISA. Serum-derived antibodies in naturally infected cats mainly recognized all six isoforms of gp60 (Fig 5). The present results confirm that S. brasiliensis 3-carboxymuconate cyclase is a highly polymorphic protein [45] with sizes of 55–62 kDa and with isoelectric points of 4.45–4.80. 2D immunoblotting revealed less diversity and more weakly reacting spots than 1D immunoblotting, perhaps due to protein loss during sample preparation for 2D electrophoresis (compared to the crude extracts used in 1D immunoblotting and ELISA) and serum dilution during pooling.
In the past, sporotrichosis was reported in horses more frequently than in other animals [55, 56]. Although the disease has been reported in several animal species, cats are currently the most frequently affected domestic animal. Outdoor cats are exposed to the fungus through contact with natural environmental sources or other sick cats. Cats have gained importance in the zoonotic transmission of Sporothrix to humans [8, 28–30, 57, 58]. Presently, the all-time high number of feline sporotrichosis cases reported in Brazil has reached epidemic proportions [8, 20, 29, 59]. Unlike humans, cats are highly susceptible to this fungus due to the low frequency of granulomas and the richness of fungal elements observed during skin histopathology [18]. Moreover, unlike human-disseminated sporotrichosis, which classically affects immunocompromised individuals [24, 26], systemic disease in cats occurs frequently and is not associated with immunodeficiency caused by feline immunodeficiency virus and/or feline leukemia virus [18]. In this scenario, absence of an efficient host immune response is a key factor in disease progression. The low frequency of granulomas and uncontrolled fungal growth suggest that a lack of adequate cellular immunity underlies disease severity and pathology.
ELISA has been achieved with various antigen preparations, thus enabling serodiagnosis of human [41–43, 60, 61] and feline [44] sporotrichosis. The present investigation suggests that ELISA-based quantitation of anti-S. brasiliensis IgG is remarkably sensitive for the detection of feline sporotrichosis (cutoffs of 0.346–0.407 OD); as expected, this strategy does not differentiate between S. brasiliensis or S. schenckii as the agent of infection. Despite its high prevalence in South and Southeast Brazil, S. brasiliensis infection in cats is not an exclusive host-pathogen association, since the sibling agent S. schenckii s. str. also occurs in cats in Brazil [8, 9, 20] and Malaysia [62], albeit with significantly lower frequency. Our serology-based observations of a convergent antigenic response are similar to previous 2D immunoblotting results for human sporotrichosis [45]. Interestingly, in other thermally dimorphic fungi such as Paracoccidioides brasiliensis and P. lutzii (Onygenales), antigen composition seems to vary considerably between species, an observation that supported development of a differential diagnosis system based on titration of serum-derived antibodies from humans infected with distinct strains of Paracoccidioides [63]. It is likely that similarities in antigen profiling among clinical Sporothrix spp. (S. brasiliensis, S. schenckii, and S. globosa) could be related to a recent speciation event and therefore be less susceptible to variability; however, this hypothesis requires further investigation. In addition, factors related to environment or to host association may impose strong selection pressures on Sporothrix antigen profiles.
To date, no information about the humoral response in feline sporotrichosis has been reported in the literature; it is a completely unknown area in veterinary medicine that merits exploration. Here, sera from cats with sporotrichosis displayed immunoblotting patterns of Sporothrix-specific immunogenic molecules that were markedly different from patterns from sera from uninfected cats. These immunogenic components were 20–160 kDa in molecular weight. The main molecules recognized by cat sera were a 60-kDa protein in the S. brasiliensis proteome and a 70-kDa protein in S. schenckii s. str., followed by molecules weighing 100 kDa, 90 kDa, 40 kDa, and 38 kDa. The variety of antigenic components recognized by each serum may be due to specific antigens secreted by individual fungal strains [3, 64] as well as to different mechanisms of activation of each host’s immune system [45, 65]. However, variation in antibody repertoire seems reasonable in a genetically diverse host population [66].
We interpreted published data lacking taxonomic information or protein identification via matrix-assisted laser desorption/ionization time of flight (MALDI-ToF)/mass spectrometry (MS) and identified an immunodominant fraction that oscillated between 60 kDa and 70 kDa in various studies (Table 2). In this scenario, the humoral immune response may be influenced by the infection strain and the antigen preparation used to detect the humoral response. Regarding human sporotrichosis, Mendoza et al. [67] observed that 147-kDa, 90-kDa, 74-kDa, 55-kDa, and 40-kDa molecules are commonly recognized by sera from patients with sporotrichosis, and Scott & Muchmore [68] showed that 70-kDa, 40-kDa, 36-kDa, and 22-kDa molecules are immunodominant in human sporotrichosis. A 70-kDa molecule was previously highlighted as immunodominant during murine sporotrichosis [69–71], with IgG1 and IgG3 predominant [71]. This 70-kDa molecule was originally described as an adhesin molecule for fibronectin and laminin in S. schenckii s. str. [69, 70] that localized to the fungal cell wall [46]. More recently, we identified this molecule via 2D immunoblotting followed by MALDI-ToF/MS as 3-carboxymuconate cyclase (GenBank: KP233225), the major antigen of human sporotrichosis [45]. Based on 1D electrophoresis [3, 46] and 2D electrophoresis [45], the molecular weight (55–73kDa) and isoelectric point (4.33–4.85) of 3-carboxymuconate cyclase vary intra- and interspecifically [3, 45, 46, 64]. Variation in gp60/gp70 may be related to differential glycosylation patterns and amino-acid substitution along the protein core, since all glycoforms/isoforms display identical MALDI-ToF/MS spectra [45] (Table 2).
Here, the intensity of the recognition of distinct molecules differed among sera from different cats, but serum from an individual cat displayed little variation when probing different antigen preparations. These data suggest that the antibody response differs between cats and that there are few qualitative variations in the expression of cellular antigens by S. brasiliensis and S. schenckii s. str. In this study, the clinical presentation of sporotrichosis in cats corresponded mainly to multiple skin lesions; we observed no association between the distribution and number of skins lesions and the number or type of molecules recognized by antibodies in the sera. Although it remains to be clarified whether the antibodies produced during active infection in feline sporotrichosis are protective, antibodies against 3-carboxymuconate cyclase appear to inhibit fungal adhesion to the host in a dose-dependent manner [70, 73]. In another dimorphic fungus, P. brasiliensis, passive administration of monoclonal antibodies against the immunodominant antigen gp43 or against the recombinant protein before and after intratracheal or intravenous infections reduced fungal burden and decreased pulmonary inflammation in mice [75, 76].
Gaining insight into host-parasite interplay in the immunological context is essential for understanding the emergence of feline sporotrichosis and is critical to serodiagnosis and the development of vaccines. Here, we demonstrated that antigens derived from yeast cell extracts of S. brasiliensis and S. schenckii s. str. yielded excellent results in ELISA and immunoblotting. The variety of molecules recognized by sera may be related to certain characteristics of the isolate, such as virulence, or even related to immune-system activation in each individual host. During infection, Sporothrix antigens elicit an IgG-mediated response; 3-carboxymuconate cyclase (gp60 in S. brasiliensis and gp70 in S. schenckii) is the immunodominant molecule in feline sporotrichosis, similar to murine and human disease. Therefore, this molecule may also be useful as a marker in the diagnosis of feline sporotrichosis and is a promising candidate for the development of therapeutic vaccines to tackle sporotrichosis in highly endemic areas.
|
10.1371/journal.ppat.1004292 | Phase Variation of Poly-N-Acetylglucosamine Expression in Staphylococcus aureus | Polysaccharide intercellular adhesin (PIA), also known as poly-N-acetyl-β-(1–6)-glucosamine (PIA/PNAG) is an important component of Staphylococcus aureus biofilms and also contributes to resistance to phagocytosis. The proteins IcaA, IcaD, IcaB, and IcaC are encoded within the intercellular adhesin (ica) operon and synthesize PIA/PNAG. We discovered a mechanism of phase variation in PIA/PNAG expression that appears to involve slipped-strand mispairing. The process is reversible and RecA-independent, and involves the expansion and contraction of a simple tetranucleotide tandem repeat within icaC. Inactivation of IcaC results in a PIA/PNAG-negative phenotype. A PIA/PNAG-hyperproducing strain gained a fitness advantage in vitro following the icaC mutation and loss of PIA/PNAG production. The mutation was also detected in two clinical isolates, suggesting that under certain conditions, loss of PIA/PNAG production may be advantageous during infection. There was also a survival advantage for an icaC-negative strain harboring intact icaADB genes relative to an isogenic icaADBC deletion mutant. Together, these results suggest that inactivation of icaC is a mode of phase variation for PIA/PNAG expression, that high-level production of PIA/PNAG carries a fitness cost, and that icaADB may contribute to bacterial fitness, by an unknown mechanism, in the absence of an intact icaC gene and PIA/PNAG production.
| Staphylococcal polysaccharide intercellular adhesin (PIA), also known as β-1-6-linked N-acetylglucosamine (PNAG) plays a role in immune evasion and biofilm formation. Evidence suggests that under certain circumstances PIA/PNAG production is beneficial, whereas at times, it may be advantageous for the bacteria to turn production off. In S. epidermidis, PIA/PNAG can be switched off when an insertion sequence recombines into the intercellular adhesin locus (ica). In this study, we have found a short tandem repeat sequence in the ica locus of S. aureus that can undergo expansion and contraction. The addition or subtraction of non-multiples of three of this repeat shifts the reading frame of the icaC gene, resulting in the complete loss of PIA/PNAG production. We hypothesize that certain conditions that make the PIA/PNAG-negative phenotype advantageous during infection, such as the development of an effective immune response to PIA/PNAG on the bacterial surface, would select for repeat mutants. In support of this hypothesis, we found clinical isolates with expansion and deletion of the repeat. These findings reveal a new on-off switch for the expression of PIA/PNAG.
| Phase variation functions as a reversible on/off switch for the expression of a particular gene. The result is commonly an alteration in the expression of some cell surface-expressed antigen. Slipped-strand mispairing is one mechanism that can lead to the production of a phase variant. Slipped-strand mispairing occurs during DNA replication when there is mispairing between mother and daughter DNA strands in regions of DNA that contain simple 1–10 nucleotide repeats [1]. This results in the addition or subtraction of one or more repeats that can bring about a change in transcriptional efficiency or shift the reading frame to alter or halt translation.
Staphylococcus aureus infections are responsible for an enormous loss of life; deaths from methicillin resistant S. aureus (MRSA) alone exceed 18,000 yearly in the United States, making it the leading cause of death by a single infectious agent [2]. Antibiotic resistance is a mounting problem and an effective vaccine is not yet available. Biofilm formation plays an important role, particularly in device-related infections, and it contributes to antibiotic failure and resistance of the bacteria to host immune defenses.
Biofilm formation is the aggregation of bacteria on a solid surface within a self-produced extracellular polymeric matrix. Formation of a biofilm confers several survival advantages to the resident bacteria. The biofilm provides protection from adverse environmental conditions such as heat, shear force, and UV damage; as well as protection from the host immune system and antibiotic challenge [3]. Biofilm bacteria are resistant to antibiotic levels up to 1,000-fold higher than planktonic bacteria that are genetically identical [4], [5]. A major component of the S. aureus biofilm extracellular matrix is the polysaccharide poly-N-acetylglucosamine.
The staphylococcal polysaccharide intracellular adhesin (PIA) is a high molecular weight polymer of β-1-6-linked N-acetyl-glucosamine (PNAG) [6], [7]. In addition to its role in intercellular adhesion and biofilm formation, PIA/PNAG also plays a role in immune evasion [8], [9]. Evidence suggests that antibodies against PIA/PNAG often recognize secreted PIA/PNAG rather than the surface-associated form, resulting in an ineffective immune response [8]. In contrast, an effective immune response against surface-associated PIA/PNAG, which can be directed by a conjugate vaccine, can successfully eradicate an infection [10]. Thus, PIA/PNAG protects the bacteria from immune defenses but under certain circumstances could actually be the target of an effective immune response.
PIA/PNAG is synthesized by the proteins encoded in the icaADBC intercellular adhesin locus [11], [12]. IcaA is a transmembrane glucosyltransferase that, together with IcaD, produces short PIA/PNAG oligomers [13]. IcaC is an integral membrane protein that is necessary for linking the short oligomers into longer polymer chains, and is thought to be involved in translocation of these chains to the cell surface [13]. Once there, IcaB is responsible for partial deacetylation of the PIA/PNAG molecule, which is required for retention at the cell surface [14].
A number of regulators modulate icaADBC transcription, including IcaR and CodY, which are repressors, and SarA and GraRS, which are positive regulators [15], [16], [17], [18]. Other regulatory mechanisms have been implicated as well and are described in a recent review [19]. We found previously that a 5-nucleotide deletion mutation within the icaADBC promoter was sufficient to induce constitutive transcription of the icaADBC genes and high-level PIA/PNAG production, resulting in a mucoid phenotype and strong biofilm production [15]. In the present study, we noted that growth of the mucoid strain in liquid culture resulted in the rapid accumulation of non-mucoid variants. We investigated this phenomenon and found that the most frequent mutation leading to the PIA/PNAG-off phenotype was a change in the number of a specific tandem repeat in icaC. This mutation was reversible and was found in clinical isolates as well. The PIA/PNAG-negative variants had a growth advantage over the PIA/PNAG-overproducing parent and rapidly predominated the cultures. This represents a newly recognized mechanism of PIA/PNAG regulation in S. aureus.
S. aureus strain MN8m is a spontaneous PIA/PNAG-overproducing mutant of strain MN8 [15], [20]. A 5 bp deletion in the icaADBC promoter region of MN8m is responsible for constitutive icaADBC transcription and the constitutive hyper-production of PIA/PNAG that gives the strain its mucoid appearance [15]. It is highly aggregative in liquid culture whereas non-mucoid strains are dispersed, making the culture turbid (Fig. 1A). We found that MN8m cultures frequently exhibited an appearance that was somewhere between that of a turbid, non-mucoid strain and an autoaggregative mucoid strain (Fig. 1A). When these cultures were plated on Congo red agar (CRA), both mucoid colonies, which appear dry with irregular edges, and non-mucoid colonies, which are slick, circular, and occasionally surrounded by a transparent red perimeter, were observed (Fig. 1B).
We isolated 15 of the variant colonies and sequenced the icaADBC locus (Fig. 2). All of the isolates still exhibited the 5 bp deletion that leads to constitutive icaADBC transcription. The sequence of the icaADBC locus of four of the isolates was identical to that of MN8m, suggesting a mutation had occurred elsewhere in the chromosome. One of the isolates had a single point mutation within icaB. As icaB has been shown previously to be dispensable, it is likely that a secondary mutation outside of the icaADBC locus was present in this strain as well. One of the isolates had a nonsense mutation at the 5′-end of the icaC gene. The remaining nine isolates all shared the loss of a “ttta” tetranucleotide repeat within the icaC gene. The insertion led to a shift in the translational reading frame and truncation of the protein; reducing IcaC from 350 amino acids to 303. Because this mutation was the most common amongst the variants, we chose to study it further. We focused on isolate JB12, a tetranucleotide insertion variant.
The icaADBC genes are co-transcribed, so to determine levels of the full-length transcript in JB12, we quantified levels of the 3′-most transcript, icaC, by realtime RT-PCR. As shown in Fig. 3A, icaADBC transcript levels were more than 300-fold greater in MN8m than in the non-mucoid parent strain MN8 and the level remained elevated in the non-mucoid variant. To determine whether or not PIA/PNAG was produced by the JB12 variants, we performed slot-blot analysis using PIA/PNAG-specific rabbit polyclonal antiserum. As depicted in Fig. 3B, no PIA/PNAG was detected on the cell surface or in the spent media of JB12 cultures. Therefore, the mutation in the icaC gene resulted in the complete loss of detectable levels of PIA/PNAG.
To confirm that the nucleotide insertion was responsible for loss of the mucoid phenotype, we complemented the icaC gene in trans. Expression of the intact icaC gene in strain JB12 from the IPTG-inducible plasmid pCL15, lead to restoration of the mucoid colony morphology on CRA plates (Fig.4A), PIA/PNAG synthesis (Fig. 4B), and biofilm formation (Fig. 4C). Introduction of the empty vector into the variant had no effect.
Phase variation is, by definition, a reversible on/off switch. Therefore, if the tetranucleotide repeat insertion was an example of phase variation, then we would expect to isolate variants of JB12 in which the mucoid phenotype was restored. To determine whether or not the phenotype was reversible we plated cultures of JB12 onto CRA. The reversion back to the mucoid phenotype was a rare event and only 1 variant was detected per approximately 45,000 cell divisions. We sequenced 6 mucoid variants from separate JB12 cultures, and all 6 variants sequenced had lost the 4 bp repeat unit that was gained in JB12 meaning that they had reverted back to the MN8m genotype.
To determine whether the mutation in icaC was RecA-dependent, we disrupted the recA gene in MN8m with the bursa aurealis mariner transposon. PIA/PNAG-negative phenotypic revertants were still isolated from the recA mutant and of these revertants, the prevalence of the icaC repeat mutation was equivalent (6 out of 10 revertants). These results indicate that the mutation was RecA-independent and strongly suggest that the mutation occurred through slipped-strand mispairing.
To determine whether or not this tetranucleotide repeat indel occurred outside of the laboratory, we examined 51 fully sequenced genomes available in NCBI and 52 sequenced genomes within the NARSA repository. Of these, 9 clinical isolates (∼9%) contained a variation in the tandem repeat region in icaC described in this study, demonstrating that the phase variants do occur in vivo (Table 1). Sequences from two representative clinical MRSA strains with deletion or expansion of the repeat units are illustrated in Fig. 5A. The isolates with altered repeat number were PIA/PNAG-negative Fig. 5B.
We also analyzed a fully sequenced genome using the online server Burrows-Wheeler Tandem Repeat Searcher (BWtrs) to determine whether tetranucleotide repeat indels occur within other genes [21]. Strain MN8m has not been fully sequenced so we chose a strain with an icaC tetranucleotide expansion, strain Bmb9393. The genome harbored 59 regions with 3 or more direct tandem tetranucleotide repeats. Out of these 59 regions, 13 exhibited indels when the region was compared to 48 completed S. aureus genomes and 438 scaffolds or contigs in the NCBI database. Nine of these indel regions were within intergenic regions, but 4 of them occurred within open reading frames. Two of these indels were in putative phage anti-repressor proteins (SABB_01096 and SABB_02531), one was in a hypothetical phage protein that was not annotated in Bmb9393, and the fourth was in icaC.
The frequency of reversion from non-mucoid (JB12) to mucoid was very low. We hypothesized that the higher frequency with which nonmucoid variants were isolated from MN8m cultures was due to a fitness cost imparted by high-level PIA/PNAG production. To determine if there was a fitness cost associated with constitutive PIA/PNAG synthesis, we inoculated competitive co-cultures with equivalent numbers of MN8m and JB12 and examined shifts in the population over time by assessing colony morphology on CRA. We observed that there does indeed appear to be significant growth advantage in the PIA/PNAG-negative variant JB12, and that by 12 hours, more than 95% of the culture was non-mucoid (Fig. 6). Direct calculation of the fitness cost of PIA/PNAG over-production versus PIA/PNAG loss resulted in fitt (relative bacterial fitness) values of +1.401 at 6 hours, and +1.386 at 12 hours, with a value greater than 1 indicating a significant fitness advantage of the JB12 PIA/PNAG-negative phenotype over the mucoid MN8m.
We calculated the generation time for strains with differing levels of PIA/PNAG production. The generation time for MN8, a low-PIA/PNAG-producing strain, was 48 minutes, while that of MN8m, the overproducing strain, was 67 minutes. Interestingly, while the PIA/PNAG-negative strain JB12 had a generation time of 54 minutes, a strain lacking the entire ica locus, MN8ΔicaADBC::erm had a longer generation time of 59 minutes.
It stood to reason that the difference in generation time was largely responsible for the frequent isolation of non-mucoid variants from MN8m cultures and that the actual frequency of the repeat insertion was low, similar to the rate of repeat deletion in JB12. To minimize the contribution of growth rate, we inoculated liquid medium from single MN8m colonies, plated half of the suspension immediately on CRA to ensure that the starter culture was free from variants, and incubated the culture for only 67 minutes, enough time for a single round of cell division before plating the remainder. We did not detect any variants in 45,000 cell divisions, suggesting that the frequency is less than 1 in 45,000 divisions.
In the absence of a fitness advantage to select for PIA/PNAG-off mutants, we would not expect a high prevalence of mutants in culture. Unlike strain MN8m, strain MN8 produces a more typical, moderate amount of PIA/PNAG. Therefore, to determine whether the icaC mutation was only selected for when the parent strain was a PIA/PNAG-overproducer or whether the mutation would occur within an average PIA/PNAG-producing strain, we used high throughput sequencing to determine whether the mutation occurred in MN8. We amplified a 344-bp region of the icaC gene that included the tandem repeats and sequenced the product. Out of ∼88,000 reads, ∼140 contained a 4-nt expansion or contraction of the repeat region confirming that the mutation occurs in an average PIA/PNAG-producing strain in vitro.
We also investigated whether the clinical MRSA isolates with icaC-off mutations, NRS63 and NRS264, were derived from average PIA/PNAG-producing strains or hyper-producing strains. We performed realtime RT-PCR and found that, unlike strain MN8m, icaC transcript levels were moderate and comparable to strain MN8, not strain MN8m (data not shown). We selected 48 colonies with a darker phenotype on CRA from each strain, performed slot-blot analysis using the PIA/PNAG-specific antiserum, and sequenced the ica loci. Revertants in which the icaC gene reverted to the “on” genotype produced only modest levels of PIA/PNAG (data not shown). Together these data suggest that the “icaC-off” strains NRS63 and NRS264 arose from moderate PIA/PNAG producing strains.
We did not directly measure function to confirm that the IcaA, IcaD, and/or IcaB proteins were functional in JB12; however, complementation of the mucoid phenotype in trans by a copy of the icaC gene alone suggests that only the function of IcaC is altered in JB12 and that IcaA, IcaD, and IcaB are still present and functional. We hypothesized that one or more of these proteins may have alternative roles within the cell that contribute to bacterial fitness and that this was the reason why icaC was the target for phase variation. Under starvation conditions, when the bacteria were switched to minimal media containing no carbon source, JB12 survived significantly longer than MN8Δica::tet (data not shown). To confirm that this difference was not due to a secondary mutation introduced during strain passage, three isogenic strains that differed only at the ica locus were made. We performed allelic exchange in strain MN8Δica::tet to replace the tetracycline resistance cassette and the interrupted ica locus with the entire MN8m (MN8icaC_On) or the entire JB12 (MN8icaC_Off) ica locus on the chromosome. The allelic exchange mutants exhibited survival profiles similar to MN8m and JB12 and both strains survived longer in the minimal media than MN8Δica::tet (Fig. 7). Production of intact, and presumably functional IcaA, IcaD, and IcaB appeared to significantly increase survival of JB12 under these growth-limiting conditions.
The polysaccharide PIA/PNAG plays an important role in virulence both through its contribution to biofilm formation and immune evasion. In fact, the benefits of PIA/PNAG to survival appear to have resulted in its ubiquitous production by a wide variety of pathogens [22]. It is clear however, that it is not always necessary for survival during infection as PIA/PNAG-negative S. aureus and S. epidermidis strains have been isolated [23], [24]. In S. epidermidis, and in a minority of S. aureus strains, PIA/PNAG production can be switched off by the insertion of an IS256 element in the icaC gene, however, prior to this study, phase variation of PIA/PNAG expression in S. aureus isolates that lack this insertion element was not clear [25], [26].
We noted that it was very difficult to maintain pure mucoid cultures of the PIA/PNAG-overproducing strain MN8m. Over time, MN8m cultures appear to contain a mixture of both mucoid and non-mucoid bacteria. In this study, we investigated the molecular basis for this phenomenon. The most common mutation resulting in the PIA/PNAG-negative phenotype was an expansion of a 4-nt repeat within the icaC gene. The rapid increase in the proportion of PIA/PNAG-negative bacteria in these cultures was due to their increased fitness relative to the PIA/PNAG-overproducing parent strain. Of note, we grew our cultures in conical tubes rather than Erlenmeyer flasks, and the depth of the medium likely resulted in microaerobic conditions. The low levels of oxygen during growth may have contributed to PIA/PNAG production and to the relatively low endpoints (OD600nm ∼3.0) in our growth curves [27] and may have also contributed to the growth advantage of the PIA/PNAG-negative mutants. The mutation frequency was very low but the growth advantage allowed non-mucoid variants to take over liquid cultures in time. This resulted in a wide variability in the number of variants present in different cultures depending on the point at which the first variants arose. Therefore, the final number of variants present at the end of the culture period would depend upon the timing of the first mutation event and would be stochastic.
The mucoid strain, MN8m produces approximately 1,000-fold more PIA/PNAG than most clinical isolates. The fitness advantage of the PIA/PNAG-negative variants isolated upon in vitro culture, such as JB12, may be due to the metabolic cost of hyperproduction of PIA/PNAG. During infection, PIA/PNAG plays an important role in immune evasion, and PIA/PNAG-negative variants would likely be more susceptible to neutrophil-mediated killing. Therefore, despite the metabolic cost, selection pressures that favor the PIA/PNAG-negative phase variants are likely less pronounced in vivo. However, skin colonization and ocular infections appear to favor the PIA/PNAG-negative phenotype in S. epidermidis and ica-negative clinical isolates of S. aureus have been detected as well [23], [24], [28]. Furthermore, as PCR amplification of the ica genes is often used to demonstrate the capacity to produce PIA/PNAG [29], [30], our finding that a 4-nt indel mutation can shut off PIA/PNAG production suggests that PIA/PNAG negative clinical isolates may be more prevalent than previously appreciated. When we analyzed 103 S. aureus genomic sequences, we found that 9 (∼9%) contained a slipped strand mutation in the tandem repeat region in icaC. We were able to obtain 2 of these isolates, NRS63 and NRS264, and both were PIA/PNAG-negative. When we isolated “IcaC-on” revertants from NRS63 and NRS264 we found moderate PIA/PNAG production, suggesting that PIA/PNAG hyper-production is not a prerequisite for selection of “IcaC-off” mutants. An immune response against PIA/PNAG could also serve as a selective pressure against PIA/PNAG production. Antibodies against deacetylated PIA/PNAG effectively mediate opsonophagocytosis [31]. Therefore, in the event that the host mounts an effective antibody response, the capacity to switch PIA/PNAG production off could benefit the bacteria in vivo. It could also benefit the bacteria in conditions that favor the planktonic mode of growth.
The 5 bp deletion in the MN8m ica promoter leads to constitutive icaADBC transcription and the accumulation of at least 300-fold more transcript than the clinical isolate parent strain MN8. Such high-level transcript production and protein synthesis would seem to be metabolically costly even in the absence of PIA/PNAG production. We therefore found it somewhat surprising that the most common mutation (9 out of 15 variants) occurred within the last gene in the operon (icaC) and that icaADBC transcript was still being produced at MN8m levels. Furthermore, the insertion sequence IS256 has been shown to turn PIA/PNAG synthesis off through insertional inactivation of icaC [26] suggesting again that mutation of icaC is the preferred “off switch” for PIA/PNAG production. We tested our hypothesis that there might be some advantage to continuing synthesis of IcaA, IcaD, and IcaB even though PIA/PNAG is not produced in the absence of IcaC. Our results indicate that under nutrient-limiting conditions, possession of functional icaADB genes was advantageous for survival. The basis for this survival advantage is unclear at this time. Further work is necessary to conclusively determine whether or not the IcaA, IcaD, and IcaB proteins function in some capacity in addition to PIA/PNAG synthesis.
In conclusion, we found that the RecA-independent expansion or contraction of a 4-nt tandem “ttta” repeat shifts the reading frame of icaC, leading to a premature stop codon, truncating the protein at 303 amino acids; 47 amino acids shorter than full-length protein. Structural prediction indicates that the mutation disrupts a transmembrane domain of IcaC, and we found that the mutation resulted in the complete abrogation of PIA/PNAG production. We found that the mutation frequency was low, but that in vitro, elevated production of PIA/PNAG carried a fitness cost and consequently, PIA/PNAG-negative phase variants quickly increased in number relative to PIA/PNAG over-expressers. Of note, IcaC appears to be the chosen target for phase variation in PIA/PNAG production and loss of this protein appears to confer a survival advantage under nutrient-poor conditions relative to loss of the entire operon suggesting that the other proteins encoded within the ica locus could have some other function. Alternatively, PIA/PNAG precursors could accumulate within the bacterial cells in the absence of IcaC, and affect growth. Studies to determine the effect of the IcaADB proteins on growth are underway.
Staphylococcus aureus strain MN8 is a clinical isolate from a case of toxic shock syndrome [32]. Strain MN8m was a spontaneous mutant isolated from a chemostat culture of strain MN8 [20]. Strain SA113Δica was provided by Dr. Sarah Cramton [12] and the mutation was transduced to strain MN8 by phage 80α to produce MN8Δica::tet. Strain MN8Δica::tet was complemented by allelic exchange with the ica loci from the entire MN8m and JB12 strains to produce strains MN8icaC_On and MN8icaC_Off, respectively. To this end, the ica loci were amplified by primers SA11 and SA12 as previously described [12] cloned into the pMAD vector, and allelic exchange was performed as previously described [33]. Mutants were tetracycline sensitive and allelic exchange was confirmed by sequencing the ica loci. A recA mutant of strain MN8m was produced by transducing the bursa aurealis transposon from strain NE805 (NARSA repository) using phage 80α. NRS264 and NRS63 are sequenced clinical isolates obtained from the NARSA repository. NRS264 was associated with bacteremia and abscess and NRS63 was a bacteremia isolate. All strains were grown aerobically at 37°C on tryptic soy agar (TSA) plates containing the appropriate antibiotic. Liquid cultures were in tryptic soy broth containing 1% glucose (TSBG), incubated in air at 37°C, 200 rpm in 5 mL in 50 mL conical tubes (microaerobic conditions). Congo red agar was composed of brain heart infusion (BHI) agar +3.6% sucrose +0.5% glucose +0.08% congo red.
Plasmid purifications were performed using the QIAprep Spin Miniprep kit (Qiagen, Valencia, CA) according to the manufacturer's instructions. Primers were custom synthesized by Integrated DNA Technologies (Coralville, IA). Restriction enzymes were purchased from New England Biolabs (Beverly, MA). The icaC gene was cloned into the isopropyl-β-d-thiogalactopyranoside (IPTG)-inducible vector pCL15 (kindly provided by Dr Chia Lee, University of Arkansas) [34]. The gene was PCR amplified from strain MN8m genomic DNA with the primer set, icaCSphIFwd (5′-CCGCGCATGCCAAAAATGGCAGAGAGGAAGA-3′) and icaCKpnIRev (5′-CCGCGGTACCCCGCGTGTTTTTAACATAGC-3′). Initial cloning was performed in E. coli using the pCR4TOPO vector (Invitrogen, Grand Island, NY) according to manufacturer's instructions. The genes were digested from the cloning vector with the appropriate restriction enzymes, purified after gel electrophoresis using the QIAquick Gel Extraction kit, and ligated into pCL15 using Ready-To-Go T4 DNA ligase (GE Healthcare, Piscataway, NJ). After passage through E. coli, all plasmid constructs were transformed into the restriction-deficient S. aureus strain RN4220 according to the method of Lee [35]. Constructs were transferred to other strains of S. aureus by transduction with phage 80α.
To generate a bacterial growth curve for use in calculating strain generation time, TSBG was inoculated with individual bacterial colonies and gently sonicated for 30 seconds to break apart clusters. Each culture was diluted to OD600nm of 0.1 and used to inoculate fresh TSBG 1∶100, with separate tubes for each time point. The cultures were incubated at 37°C, 200 rpm. At each time point the cultures were gently sonicated, the OD600 nm measured, and plated on CRA plates to monitor population mucoid/non-mucoid phenotype over time and ensure that variants did not arise to influence observed growth rates. Logarithmic growth was determined to occur between 180 and 360 minutes for each of the strains. For this time period, the A600 measurements were converted into log2 values, and the generation time was calculated as the inverse of the slope of the line of best fit.
For the competitive fitness assay, cultures were grown overnight in TSBG, and gently sonicated for 30 seconds. Each culture was diluted to a concentration of 103 cells and mixed 1∶1 with MN8m + JB12, with separate tubes for each time point. The cultures were incubated at 37°C, 200 rpm. At each time point the cultures were gently sonicated, serially diluted and plated in triplicate on CRA plates for CFU counting. Calculation of the difference in fitness was determined using the function derived from Sander et al. [36] LN(((nmt/mt)/(nmt-1/mt-1))∧(1/gen)) where nmt and mt represent the non-mucoid and mucoid cells, respectively at a given time t. While nmt-1 and mt-1 denote the quantity of non-mucoid and mucoid cells at the preceding timepoint. The quotient of the ratios was standardized with the exponent 1/generation, with the assumption that cell numbers determined at 24 hours represents approximately 17 generations. The relative bacterial fitness for a given time was calculated as fitt = 1+St. The fitness value is equal to 1 if there is no difference in fitness between the competing strains, less than 1 if the non-mucoid phenotype reduces fitness, or greater than 1 if the non-mucoid phenotype increases bacterial fitness.
Microtiter plate assays for biofilm formation were performed essentially as described previously by Christensen et al. [37] with minor modifications. Cultures were grown overnight in 4 ml of TSBG or TSBG +10 µg ml−1 chloramphenicol, diluted 1∶200 in the same media or media with 1mM IPTG added for plasmid induction, and aliquoted into 96-well polystyrene flat-bottom microtiter plates (Greiner Bio-One, Monroe, N. Carolina). After 24 hours at 37°C, the wells were emptied and washed once with phosphate-buffered saline (PBS). The plates were dried at room temperature, stained with 200 µl safranin for 1 minute, washed gently with water, and allowed to dry. The biofilms were assessed qualitatively by visual inspection and images were taken using a digital scanner. The safranin was then resuspended in 200 µl 33% acetic acid and the wells were analyzed by spectrophotometry at OD562 nm using a 96-well plate spectrophotometer.
RNA was isolated from exponentially growing bacteria, following induction with 1 mM IPTG for 2 hours if pCL15 was present, using the Qiagen RNeasy kit (Qiagen, Valencia, Calif.) according to the manufacturer's instructions. Contaminating DNA was digested with Turbo DNAse (Ambion, Austin, Texas), and the mRNA transcript levels were measured by quantitative reverse transcriptase (RT)-PCR. Reverse transcription of 1 µg of RNA was performed using the Tetro cDNA synthesis kit (Bioline, Taunton, Mass.) according to manufacturer's instruction, and 10 pmol icaCRTRev (5′-CGTTCCAATAGTCTCCATTTGC-3′), and 16SRTRev (5′- TATGCATCGTTGCCTTGGTA-3′). Controls for DNA contamination contained no reverse transcriptase. SensiMix SYBR & Fluorescein mix (Bioline, Taunton, Mass.) was used for the quantitative real-time PCR with the primer sets: icaCRTFwd (5′-CGAACAACACAGCGTTTCAC-3′) and icaCRTRev, or 16SRTFwd (5′-GAACCGCATGGTTCAAAAGT-3′) and 16SRTRev.
PIA/PNAG slot blots were performed essentially as described previously by Cramton et al. [27] with minor modifications. Bacteria were grown overnight in TSBG or TSBG +10 µg ml−1 chloramphenicol +1 mM IPTG. For cell surface extracts 109 cells were collected by centrifugation, washed once with PBS, and resuspended in 250 µl of 0.5 M ethylenediaminetetraacetic acid (EDTA). To analyze secreted PIA/PNAG, 250 µl of spent media was retained. All samples were incubated in boiling water for 5 minutes, cooled, and incubated at 65°C for 1 hour with 20 µl proteinase K. Samples were boiled for an additional 5 minutes to inactivate the protease, diluted in PBS, and immobilized on nitrocellulose with a vacuum manifold. Blots were blocked overnight at 4°C in 5% bovine serum albumin, probed with 1∶5,000-diluted rabbit antiserum specific for PIA/PNAG (kindly provided Dr. Gerald B. Pier) [7] for 2 hours at 21°C, washed, and probed with 1∶10,000-diluted goat anti-rabbit immunoglobulin-horseradish peroxidase conjugate for 1 hour at 21°C. Bands were visualized with the ECL Plus western blotting detection system (GE Healthcare).
MN8 was cultured for 4 hr at 37°C. DNA was purified using the DNeasy Blood and Tissue kit (Qiagen) and amplified by PCR using SSSeqFWD (5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGGGAACGTTACCAGCTTTTCATATTC-3′) and SSSeqREV (5′-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGCCACCGCGTGTTTTTAACATAGC-3′). The PCR product was sequenced at the Nucleic Acids Research Facilities at VCU using Illumina MiSeq. Sequencing yielded ∼88,000 paired end reads, which were compared to the MN8 parent sequence using JBrowse to detect indels [38].
Liquid cultures were initially grown shaking aerobically in TSBG for 24 hours. Bacteria were collected by centrifugation, and equal numbers of bacteria for each sample were resuspended in equal volumes of MOPS minimal media (Teknova, Hollister, Ca) with no glucose, or other carbon source. The cultures were incubated at 37°C while shaking. At each time point the cultures were serially diluted and plated in triplicate on TSA plates for CFU enumeration. Each strain was analyzed in technical triplicate and biologic replicates.
|
10.1371/journal.pmed.1002381 | Women’s and men’s reports of past-year prevalence of intimate partner violence and rape and women’s risk factors for intimate partner violence: A multicountry cross-sectional study in Asia and the Pacific | Understanding the past-year prevalence of male-perpetrated intimate partner violence (IPV) and risk factors is essential for building evidence-based prevention and monitoring progress to Sustainable Development Goal (SDG) 5.2, but so far, population-based research on this remains very limited. The objective of this study is to compare the population prevalence rates of past-year male-perpetrated IPV and nonpartner rape from women’s and men’s reports across 4 countries in Asia and the Pacific. A further objective is to describe the risk factors associated with women’s experience of past-year physical or sexual IPV from women’s reports and factors driving women’s past-year experience of partner violence.
This paper presents findings from the United Nations Multi-country Study on Men and Violence in Asia and the Pacific. In the course of this study, in population-based cross-sectional surveys, 5,206 men and 3,106 women aged 18–49 years were interviewed from 4 countries: Cambodia, China, Papua New Guinea (PNG), and Sri Lanka. To measure risk factors, we use logistic regression and structural equation modelling to show pathways and mediators. The analysis was not based on a written plan, and following a reviewer’s comments, some material was moved to supplementary files and the regression was performed without variable elimination. Men reported more lifetime perpetration of IPV (physical or sexual IPV range 32.5%–80%) than women did experience (physical or sexual IPV range 27.5%–67.4%), but women’s reports of past-year experience (physical or sexual IPV range 8.2%–32.1%) were not very clearly different from men’s (physical or sexual IPV range 10.1%–34.0%). Women reported much more emotional/economic abuse (past-year ranges 1.4%–5.7% for men and 4.1%–27.7% for women). Reports of nonpartner rape were similar for men (range 0.8%–1.9% in the past year) and women (range 0.4%–2.3% in past year), except in Bougainville, where they were higher for men (11.7% versus 5.7%). The risk factor modelling shows 4 groups of variables to be important in experience of past-year sexual and/or physical IPV: (1) poverty, (2) all childhood trauma, (3) quarrelling and women’s limited control in relationships, and (4) partner factors (substance abuse, unemployment, and infidelity). The population attributable fraction (PAF) was largest for quarrelling often, but the second greatest PAF was for the group related to exposure to violence in childhood. The relationship control variable group had the third highest PAF, followed by other partner factors. Currently married women were also more at risk. In the structural model, a resilience pathway showed less poverty, higher education, and more gender-equitable ideas were connected and conveyed protection from IPV. These are all amenable risk factors. This research was cross-sectional, so we cannot be sure of the temporal sequence of exposure, but the outcome being a past-year measure to some extent mitigates this problem.
Past-year IPV indicators based on women’s reported experience that were developed to track SDG 5 are probably reasonably reliable but will not always give the same prevalence as may be reported by men. Report validity requires further research. Interviews with men to track past-year nonpartner rape perpetration are feasible and important. The findings suggest a range of factors are associated with past-year physical and/or sexual IPV exposure; of particular interest is the resilience pathway suggested by the structural model, which is highly amenable to intervention and explains why combining economic empowerment of women and gender empowerment/relationship skills training has been successful. This study provides additional rationale for scaling up violence prevention interventions that combine economic and gender empowerment/relationship skills building of women, as well as the value of investing in girls’ education with a view to long-term violence reduction.
| Understanding the past-year prevalence of physical and or sexual intimate partner violence (IPV) and risk factors is essential for building evidence-based prevention.
Previous studies have not compared men’s and women’s past-year prevalence reports and have been limited by a predominant focus on risk factors for lifetime exposure to IPV.
Monitoring SDG 5.2 and building evidence-based prevention require a relative understanding of the measures of past-year prevalence and the drivers of this violence.
We use data from 4 countries of the UN Multi-country Study on Men and Violence in Asia and the Pacific to compare the population prevalence rates of past-year IPV and nonpartner rape from women’s and men’s reports and present an analysis of drivers of women’s experience of past-year physical or sexual IPV.
Women’s reports of past-year male-perpetrated IPV were similar to those from men.
Four groups of variables are important drivers of IPV: poverty, all childhood trauma, quarrelling and women’s limited control in the relationship, and partner factors (substance abuse, unemployment, and infidelity).
Past-year IPV indicators based on women’s reported experience that were developed to track SDG 5 are probably reasonably reliable.
Women appear to gain resilience to violence through combined economic power and understanding gender empowerment/relationship skills, as well as education; this is an important foundation for intervention.
Further research is needed on the validity of men’s and women’s reports of IPV, which could not be determined from these data.
| In 2015, eliminating all forms of violence against women and girls (VAWG) was adopted as a target for the Sustainable Development Goal (SDG) 5 on gender equality and empowerment of women. To achieve this, we must develop and roll out effective measures to prevent male-perpetrated violence and show their effect. The indicators of progress towards this target are not finalized but will be a measure of women’s experience of intimate partner violence (IPV) and of nonpartner sexual violence in the past 12 months. According to most recent estimates, 30% of women aged 15 years and over have experienced male-perpetrated physical and/or sexual IPV, and 7% nonpartner sexual violence, in their lifetime [1,2].
In low- and middle-income countries, the World Health Organization instrument that was developed for its Multi-country Study on Women’s Health and Domestic Violence against Women is generally seen as the gold standard measure for women. Parallel research with men has developed a methodology for measuring perpetration, but the 2 measures of violence in heterosexual relationships have not been compared. Given that widely used indicators will most likely focus on reports of just 1 gender for reasons of resource constraints, it is important that there be an understanding of the comparability of men’s and women’s reports. Without this, we have uncertainty about the validity of women’s reports of experiences of IPV and nonpartner sexual violence. There is particular concern that sexual violence may be under-reported by women because rape is highly stigmatized, which may result in minimization of events, but it is also possible men might under-report perpetration of violence so as not to incriminate themselves [3,4].
Prevention of VAWG needs to be built on evidence of drivers among women currently at risk (as well as those of perpetration). There is a reasonably large amount of literature on risk factors for experience of IPV (for example, summarized in the World Health Organization’s 2010 review [5]), but major limitations include a focus on lifetime exposure (rather than past year) and overadjustment of models for (nonamenable) at-risk groups rather than focusing on risk factors. In the case of the former, this means that the outcome modelled is not exactly the ‘problem’ for which interventions are required (which is current, or future, violence). In the case of the latter, the analyses focus largely on who is at risk rather than understanding factors driving risk. The literature is also mostly focused on a single country and is cross-sectional [6], and given the variability in the variables measured and the modelling approaches used, this often constrains the ability to compare across countries and global regions. Prevention science is better informed by looking at risk factors amenable to intervention and linked to past-year experience of IPV, which are likely to differ from factors associated with lifetime experience of IPV.
The UN Multi-country Study on Men and Violence was designed to address many of the gaps in previous data sources [7]. It has a large multicountry dataset with women’s reports of IPV, collected using gold standard exposure measures, and also includes standard measures of the most important currently recognised drivers of violence as well as some hypothesised ones. We present the population prevalence of women’s experiences of past-year IPV and nonpartner rape and compare it to men’s reported perpetration across 4 countries in Asia and the Pacific, and we present risk factors associated with women’s experience of past-year physical or sexual IPV (risk factors for men’s perpetration in this dataset have been presented previously [8,9]). We present structural models to show pathways and mediators.
Ethical approval was provided by the Medical Research Council of South Africa; the College of Humanities, Beijing Forestry University; National Ethics Committee for Health Research of Cambodia; and the Faculty of Medicine at the University of Colombo, Sri Lanka.
The survey was developed by Partners for Prevention in collaboration with the Medical Research Council of South Africa and the country research teams. Research was conducted in 2011–2012. Of the 6 country surveys, only 4 had male and female interviews: China, Cambodia, Bougainville in Papua New Guinea, and Sri Lanka. The present study is intended to contrast women's reported experience of IPV and nonpartner rape with men's reported perpetration of IPV and nonpartner rape; therefore, our analysis focuses on these 4 surveys. The sample from Cambodia and the sample from Papua New Guinea were representative, respectively, of Cambodia and the island of Bougainville. The Chinese site was a county with a town and rural area, and in Sri Lanka, Colombo and 3 contrasting districts were surveyed. Further details of the research can be found elsewhere [7,10].
In each setting, we selected census enumeration areas, with a probability proportionate to size, and systematically selected households within these areas. In households, we invited a man or woman (depending on the cluster) aged 18–49 years (where necessary, randomly selected) for interview, with a trained sex-matched interviewer. Most interviews were face to face, but for men, answers to most sensitive questions were self-completed on audio-enhanced personal digital assistants (APDAs). In China, a household list of individuals in each cluster by age and sex was available and used for sampling within selected clusters, and the entire questionnaire was self-completed. Full details of the methods, sampling, and response rates are presented elsewhere [10]. We conducted surveys with women on their health and experiences of violence in 4 sites (Cambodia, China, Bougainville, and Sri Lanka). We sampled men and women in separate clusters. We conducted interviews with 3,106 women (between 477–1,103 per country) and 5,206 men (between 849–1,777 per country across the 4 analysed here). The proportion of enumerated and eligible women interviewed per site was between 92.7% (in Cambodia) and 73.9% (in Sri Lanka). For men, it ranged between 97.3% (in Cambodia) and 58.7% (in Sri Lanka; for details [7]). Measures used in the questionnaire are presented in Table 1. We followed ethical and safety guidelines for research on violence against women [11,12]. The interviewees received an information sheet and provided written consent.
The data analysis was largely planned at the point of commencement of the work on the paper. Authors EF and RJ were involved in the research from its inception and had planned the questionnaire so that it would be possible to undertake an analysis of prevalence of violence and risk factors. They ensured as much as possible that the main variables previously described in the literature [5] were included in the dataset. We planned the analysis to test the relationships between the independent variables and the outcomes. This study is reported as per the STROBE guidelines (S1 STROBE Checklist).
We combined the datasets and analysed the data using Stata, version 13. All procedures took into account the multistage structure of the dataset, with stratification by site within a country and enumeration areas as clusters. The sample was self-weighting. Women’s experiences of violence and male partner violence perpetration, as well as the independent variables, were summarized as percentages (or means), with 95% confidence limits calculated using standard methods (Taylor linearization).
We categorised the type of violence exposure according to the most severe type experienced, where greatest severity was considered as exposure to physical and/or sexual IPV, as this is the category that has been the basis of most health consequences research [1] and is consistent with the paper on male risk factors for IPV published from the same dataset [10]. It is currently common practice in the field not to model a combined variable with sexual and physical IPV and economic and emotional abuse, although this has been sometimes done [13]. This is because the field’s understanding of the latter is at a much earlier stage, with limited agreement on how to measure it, how to prevent it, and the implications (of emotional abuse alone) for health and development outcomes. It is important for the field that the issue is not ignored, hence its inclusion here, but we do not feel that the field is quite ready for it to be meaningfully pooled with sexual and physical violence for risk factor modelling and interventions. This approach has been followed by other authors, for example, Mahenge et al. [14].
The multiple emotional/economic abuse category consisted of women who had experienced more than 1 act of economic or emotional abuse but never experienced sexual or physical abuse. All ever-partnered women and men were classified into 5 violence exposure categories: none, emotional/economic without sexual or physical (henceforth referred to as ‘emotional/economic’), sexual without physical and with or without emotional/economic (henceforth referred to as ‘sexual’), physical without sexual and with or without emotional/economic (henceforth referred to as ‘physical’), or sexual and physical with or without emotional/economic (henceforth referred to as ‘physical/sexual’).
We also evaluated the relationship between the outcome (IPV) and nonresponse (missing data) in putative risk factors. No association was found between a woman’s IPV status and her nonresponse to any of the possible risk factors. However, to increase the sample of women with responses to scale measurements (e.g., gender attitudes and relationship control), women with partial responses to scale items were also included. Three methods for imputing for missing data were initially compared. These involved imputing for missing scale items using either (1) a woman’s responses to other items in the scale (individual respondent mean) or (2) the average for each item adjusted for IPV status, or (3) the average of the overall score adjusted for IPV status. There were no significant differences in the 3 methods for both gender attitudes and relationship control scores. We used ‘the average of the overall score (adjusted for IPV status)’ to impute for missing scores.
The exercise of testing variables and building model drew on current theories about risk factors and drivers of violence against women. The selection of variables as putative risk factors was informed by the state of knowledge in the field. Drawing on a life-course modified ecological model of violence risk [15], we conceptualized possible risk factors as (1) structural, (2) those pertaining to the women (including stemming from her childhood), (3) those pertaining to her partner, and (4) those pertaining to their relationship. We further were informed in our thinking by research on masculinities that views a range of male behaviours as indicator variables for hegemonic masculinity [16]. The connections between hegemonic masculinity and violence against women have been extensively theorized. In building the structural equation model, we drew on our extensive knowledge base on gender-based violence. It is well recognized that IPV is strongly associated with poverty and that poverty increases the likelihood of experience of adversity in childhood and influences access to education [9,17–19]. Research with men has shown that childhood trauma exposure influences ideas about gender equity, which is why we hypothesized this direction of effect for women [19]. Further research has shown that women’s ideas about gender influence partner selection, as does exposure to childhood trauma [20].
To show associations between independent variables that were putative risk factors, we first conducted a bivariable analysis with a (by type) lifetime IPV exposure measure and a multinomial regression with no physical, sexual, or severe economic/financial violence as the comparison group. A maximum likelihood multinomial logit model, which adjusted for the survey design, was used to compare factors associated with different types of IPV experienced with the no-violence reference category. We initially fitted bivariable models and then included all factors that were significantly associated with IPV experience in the bivariate models into an overall model, which was adjusted for the country and age group of the woman.
We examined factors associated with past-year experience of IPV considering the same independent variables, but with a past-year exposure to any physical and/or sexual IPV as the outcome, due to sample size considerations, we did not perform multinomial modelling. We sought to model 19 covariates in the logistic regression model, which, according to generally accepted rules of thumb [21], would require a total of 190 events. Because the category ‘physical and sexual IPV’ contained only 124 events, having this as an outcome in a multinomial model could have resulted in overfitting. We therefore decided to fit a regression model specifying the combined outcome of ‘any exposure to physical and/or sexual IPV’. Since this combined outcome contained 2,765 x 16.7% = 461 events, this decision allowed us to proceed with less concern about overfitting.
Multivariable logistic regression was used to determine risk factors associated with past-year physical and/or sexual IPV experience in women, with those not experiencing this as the reference group. To enable the use of a variable on frequency of quarrelling, which was not measured in Cambodia, a dummy level for Cambodia was created for the quarrelling variable for use in the logistic regression model. All variables were included in the multivariate analysis. We focus the discussion on variables with P < or = 0.05 in the model, which is adjusted for country/site and age-group of the woman.
The population attributable fractions (PAFs) for each category of IPV were calculated using the formula PAF = ((RRR − 1) / RRR) * Pe, where RRR is the adjusted relative risk ratio from the adjusted model and Pe is the proportion of women who had experienced that particular IPV type and who had the exposure.
Structural equation modelling (SEM) was conducted using Stata 13.0 to assess the interrelationship between variables associated with physical and/or sexual IPV in the multinomial regression model. The model outcome was a past-year IPV variable that had 4 levels drawn from the physical and sexual IPV questions: no exposure, sexual IPV, physical IPV, and physical or sexual IPV. The correlation between each hypothesized variable and the IPV variable was then tested by building variable pairs. All associations were tested by running a full-information maximum likelihood method to deal with missing values. This method was chosen over multiple imputations because it has been shown to yield superior results in structural equation modelling [22]. As a next stage, a measurement model was fitted with the variables allowed to freely correlate. To assess model fit of the observed data, we used the comparative fit index (CFI) (>0.95); Tucker-Lewis Index (TLI) (>0.9) for acceptable fit and (>0.95) as indicative of good fit [23]; and root mean square error of approximation (RMSEA) (of 0.05 or less) [24,25].
We fitted a path model using full information maximum likelihood (FIML) estimation to model all available data. The final model was built based on theory and statistically meaningful modifications using backwards elimination to exclude endogenous variables that did not mediate any path (with significance set at the P < 0.05 level) from the exogenous variables to IPV in order to ensure model parsimony. Before adjusting standard errors for clustering of participants in countries, model fit was very good (p(χ2) = 0.519, RMSEA < 0.001, CFI = 1.000, and TLI = 1.001). After adjusting for clustering, the coefficient of determination (CD) was 0.215. The model did not include any error covariances.
In total, 3,106 women aged between 18 and 49 years were interviewed in the 4 countries, among whom 2,855 (91.9%) were ever-partnered. Of the ever-partnered women, 90 (3.3%) did not respond to any of the questions related to IPV experience and were thus excluded from analysis. In total, 5,206 men were interviewed in the 4 countries, and 4,360 (83.8%) had ever been partnered. Four thousand and fifteen men completed the IPV questions, and 5,062 completed the non-partner rape questions.
Comparing lifetime reports of women’s experiences and men’s reports of IPV by type from the 4 countries (Table 2) reveals that sexual IPV was quite similarly reported by men and women, except women less often disclosed lifetime sexual IPV in Cambodia (9.1% versus 21%) and China (8.3% versus 19.4%) and men reported less past-year sexual IPV in Sri Lanka. Men reported less lifetime and past-year physical IPV than women in Cambodia, but much more in China. Men reported more lifetime physical IPV than women in Bougainville, but past-year reports were similar. In Sri Lanka, the overall level of violence reported by men and women and the rates for each type were similar. In every country, women reported much more past-year emotional and financial IPV than men.
In Cambodia, 0.4% (95% CI 0.1%–1.73%) of women had experienced nonpartner rape in the past year, and 1.9% (95% CI 1.12%–2.70%) of men disclosed perpetration. In China, 2.3% (95% CI 1.49%–3.43%) of women had experienced nonpartner rape in the past year, and 1.7% (95% CI 0.94%–2.52%) of men disclosed perpetration. In Bougainville, 5.7% (95% CI 4.21%–7.75%) of women had experienced nonpartner rape in the past year, and 11.7% (95% CI 9.02%–14.30%) of men disclosed perpetration. In Sri Lanka, 0.5% (95% CI 0.15%–1.40%) of women had experienced nonpartner rape in the past year, and 0.8% (95% CI 0.22%–1.43%) of men disclosed perpetration.
The prevalence of past-year physical and/or sexual IPV experience increased with age (Table 3). Poverty, indicated by present food insecurity and problems finding money for an emergency, was associated with a greater risk of IPV, as was the women being the main breadwinner. Families in which the wife provided most of the money for the home were twice as likely to have food insecurity (P < 0.001) as those in which the husband provided, another provided, or both the husband and wife shared equally in providing.
All 3 forms of childhood abuse (sexual, physical, and emotional) and witnessing abuse of mother were more common among women with past-year physical or sexual IPV experience. Women whose partners earned more than them had a lower past-year IPV prevalence than those earning the same as their partners or women who earned more. Partner characteristics associated with women’s past-year IPV experience were the male partner’s regular alcohol use, ever or past-year drug use, lack of fidelity, and unemployment. Women who were highly controlled by their partner were more likely to have experienced past-year IPV, as were those who quarrelled more often and those holding less gender-inequitable views.
S1 Table shows the prevalence of women’s social characteristics, victimisation history, partner characteristics, and gender attitudes and relationship factors by lifetime IPV exposure category for the combined dataset (all 4 countries), with the unadjusted associations and the adjusted associations shown in S2 Table. These tables show very similar patterns of associated factors as was seen in the past-year physical or sexual IPV exposure analysis.
Table 3 shows the logistic regression models of factors associated with past-year IPV. In the past 12 months, 461/ 2,765 (16.7%) women had experienced sexual or physical (or both forms of) IPV. The risk factors shown are experiencing more poverty; having experienced abuse in childhood (sexual, physical, or emotional); having a partner who drinks alcohol, uses drugs, may be unfaithful, is unemployed, or is highly controlling; and having more frequent quarrelling in the relationship. The PAF was the largest for quarrelling often, but the second greatest PAF was for the group related to exposure to violence in childhood, followed by the PAF for the group related to the woman being controlled by her partner. The partner characteristics (substance abuse, unemployment, and infidelity) had the next highest PAFs. In the backwards/forwards elimination model, currently married women were at much higher risk.
Results for the structural equation model are presented in Fig 1 and Table 4 and follow recommended guidelines outlined by Mueller and Hancock [26]. The paths between socioeconomic status and IPV were mediated by childhood trauma exposure (i.e., poorer women had a higher trauma exposure) and increased IPV risk or by women’s educational attainment (i.e., wealthier women had been in school for longer) and having more equitable gender attitudes, which conveyed IPV protection, unless associated with more quarrelling. Childhood trauma was linked to IPV through 4 pathways. One was direct, such that childhood trauma increased the risk of IPV. One was mediated by partner alcohol use and frequency of quarrelling, such that childhood trauma reduced the chance of having a low-alcohol-using partner and thus lower quarrelling. One path was mediated by (more inequitable) attitudes to gender equity. The fourth path was mediated by partner fidelity such that risk was associated with greater confidence in him being faithful. Witnessing abuse of the woman’s mother was more common in women exposed to trauma in childhood and was included to improve model fit but did not mediate a pathway. A figure with all significant and nonsignificant paths and standard errors is presented in S1 Fig.
Between a quarter and two-thirds of women in the 4 countries studied had experienced IPV, and 1.7% and 15.9% had experienced nonpartner rape. There was very great diversity in the prevalence of IPV between countries, as previously reported in Asia and the Pacific [1]. Reports by men and women show much similarity, but overall, women’s reported prevalence of lifetime physical and sexual violence experience was lower than men’s reports of perpetration, notably in sexual violence reporting. Men’s reporting of past-year nonpartner rape was much higher than women’s in Bougainville. A different pattern was seen in past-year reports that were not clearly patterned with respect to those of men, except in the area of emotional/financial abuse, for which in all countries women reported much more.
We would not necessarily expect men’s and women’s reports of nonpartner sexual violence to concur, and some women are at much higher risk than others in the population and may experience multiple rapes [27]. Although we did not have couples’ reports on partner violence, we do expect the acts/experiences of violence of men and women to be similar at a population level for past-year violence, as 75% of men had had only 1 sexual partner in the last year, and most women were married (77.7%) or cohabiting (2.9%). It is possible that women tended to minimise or forget some lifetime experiences of partner violence, but it may also be the case that higher levels of reports by men are explained by men using violence on some types of female partners more often than on their wives. Given the differences in men’s and women’s lifetime reports, we must conclude that the current global lifetime prevalence rates that are based on women’s reported experiences may underestimate the lifetime perpetration of IPV and nonpartner rape by men.
We saw 4 important groups of risk factors for IPV experience. First, our results confirm that past-year IPV victimisation is more common in a context of poverty [6]. Secondly, exposure to physical, sexual, and/or emotional childhood trauma was very strongly associated with experience of all forms of IPV (past year or lifetime). This advances current research that has focused on sexual violence or on witnessing maternal abuse [5,6]. In the structural model, childhood trauma had a direct pathway to IPV experience, and it mediated several indirect paths. This helps explain why childhood trauma exposure is such an important risk factor (as shown by the PAF). The analysis of factors associated with IPV perpetration by men has also shown the importance of all forms of childhood trauma [10]. We observed also that childhood trauma exposure was associated with a more conservative position towards gender equity. It is possible that this is easier for women to adopt if they have lower self-esteem and more insecurity after trauma, as it generally is socially rewarded and normative.
Witnessing abuse of one’s mother has been found to be associated with both experience of and perpetration of IPV in many studies [17,28–35]. We confirmed this, but in the structural model, it was not as important as childhood trauma. Since previous research has often focused on witnessing abuse rather than more thoroughly measuring childhood trauma, it is possible that assumptions that there is a direct intergenerational learning process normalising IPV victimisation among women and girls are overemphasising this 1 traumatic experience, and witnessing abuse may be better interpreted as an indicator of exposure to wider childhood experiences of emotional and other trauma, all of which elevate IPV risk. The latter explanation fits better with the knowledge that witnessing abuse of one’s mother is traumatic and repulsive, which has long been an observation that fits uncomfortably with a direct learning explanation.
The third variable group consists of partner characteristics: his drinking, past-year drug use, controlling behaviour, unemployment, and fidelity. Generally, these are previously well-established risk factors, although research with men has not confirmed associations with drug use in Asia and the Pacific, except in relation to perpetration of multiple perpetrator rape [8,10]. Alcohol abuse combines a direct impact on behaviour, financial tensions, and gender-inequitable masculinity; the fidelity measure reflects the male sexual entitlement dimension of the latter [6,8,36,37]. Highly controlling behaviour is an abusive practice that is closely related to the use of physical and sexual violence [38] and is viewed by some authors as part of the concept of emotional abuse. In the structural model, male partner alcohol consumption and infidelity both mediated pathways between childhood trauma and IPV experience—in the former case, mediated by frequency of quarrelling. These partner variables may highlight the potential for enhanced prevention intervention impact if men and women are both involved in interventions to reduce violence [39]. Partner unemployment was significant on 1 of the models and would generally be interpreted as contributing to poverty in the relationship, with associated tensions, but it may also impact on self-perceived manliness, and violence may be used as a response to this [39].
The frequency of quarrelling was very strongly associated with IPV, as it was in the models of men’s perpetration in the 4 countries [10]. Although quarrelling is linked to men’s and women’s ideas about gender equity, intervention research shows that it can be reduced within relationships by training in communication skills, and this can reduce partner violence [40].
One of the most important findings of the structural model was a pathway that can be interpreted as indicating variables that build women’s resilience to violence. This linked higher wealth, higher educational attainment, and having more gender-equitable attitudes. This is very important because all of these factors are amenable to intervention, and it highlights the role of poverty reduction and interventions to enhance girls’ schooling, which may be supported for many reasons related to development and the general upliftment of women, in IPV prevention. In this study, the Gender Equitable Men (GEM) scale was used to measure women’s gender attitudes. This is a broad measure that includes attitudes towards the use of violence against women. The latter alone have been shown to be very strongly associated with risk of violence [41,42]; however, we found strong correlations with IPV in a version of the scale without the question about attitudes towards violence.
Economic empowerment has been shown to be a fruitful area of intervention with women [43], but more consistently so when combined with a gender empowerment intervention [44]. Our analysis suggests that interventions with adult women would do better to include a focus on gender empowerment and relationship dynamics in order to ensure that empowerment alone does not result in greater quarrelling and violence. Our structural model provides some indication of why interventions that impact on several variables in the resilience pathway for women (economic status and gender attitudes/relationship skills) may be much better than single-component interventions.
Reducing childhood trauma exposure is ultimately critical to reducing women’s experience of violence and is strongly related to poverty. Whilst there is much work on early interventions in childhood to reduce the experience of trauma and IPV in the next generation, it is possible that poverty reduction will have the greatest impact.
The study findings reflect the sampled sites; generalizability beyond this is unclear, and the combined dataset analysed here does not reflect the whole region. Since the research was cross-sectional, temporality may be questioned, but since this was recent violence, this is not likely to be a great problem. All the prevalence estimates for violence were compared with estimates weighted for the number of eligible men and women per household. The latter were not significantly different in any site, and thus, we have used unweighted estimates. The main analysis was on past-year IPV exposure, and because this is less common than lifetime exposure, the power of the analysis was inevitably impacted. However, the focus has strengthened the interpretability of the results for programming, as it is the goal of IPV prevention to reduce exposure of women at risk in the future and recent abuse is the best measure of this. A study limitation is that we do not have a comparison of men’s and women's reports from the same relationship. In accordance with WHO ethics and safety guidelines, we did not interview men and women in the same location, much less in couples. The motivation is to avoid the (to our knowledge small) possibility of retaliatory violence associated with partners learning of the interview content. This risk is not justified in cross-sectional research but prevents comparison of couples’ reports.
Our findings suggest that newly emphasised past-year IPV indicators that were developed to track SDG 5 would be reasonably reliable if based on women’s interviews. Interviews with men to track past-year nonpartner rape perpetration are important. We have shown an important IPV resilience pathway. This helps us to understand why interventions that combine women’s economic empowerment and building gender-equitable attitudes (and communication skills), such as Pronyk and colleagues’ Image [43], may be more effective than those with a single-component focus. This is a very important advance in understanding as these are imminently amenable risk factors through work with populations of adult women. However, integrated approaches that reach women and men with a comprehensive set of interventions to address different risk factors would almost certainly bring the most benefit.
|
10.1371/journal.pcbi.1006304 | Age-dependent Pavlovian biases influence motor decision-making | Motor decision-making is an essential component of everyday life which requires weighing potential rewards and punishments against the probability of successfully executing an action. To achieve this, humans rely on two key mechanisms; a flexible, instrumental, value-dependent process and a hardwired, Pavlovian, value-independent process. In economic decision-making, age-related decline in risk taking is explained by reduced Pavlovian biases that promote action toward reward. Although healthy ageing has also been associated with decreased risk-taking in motor decision-making, it is currently unknown whether this is a result of changes in Pavlovian biases, instrumental processes or a combination of both. Using a newly established approach-avoidance computational model together with a novel app-based motor decision-making task, we measured sensitivity to reward and punishment when participants (n = 26,532) made a ‘go/no-go’ motor gamble based on their perceived ability to execute a complex action. We show that motor decision-making can be better explained by a model with both instrumental and Pavlovian parameters, and reveal age-related changes across punishment- and reward-based instrumental and Pavlovian processes. However, the most striking effect of ageing was a decrease in Pavlovian attraction towards rewards, which was associated with a reduction in optimality of choice behaviour. In a subset of participants who also played an independent economic decision-making task (n = 17,220), we found similar decision-making tendencies for motor and economic domains across a majority of age groups. Pavlovian biases, therefore, play an important role in not only explaining motor decision-making behaviour but also the changes which occur through normal ageing. This provides a deeper understanding of the mechanisms which shape motor decision-making across the lifespan.
| Decisions in everyday life often require weighing the probability of successfully executing an action (e.g., successfully crossing a street) against potential rewards and punishments. Although older individuals take fewer risks during such motor decision-making scenarios, the underlying mechanism remains unclear. Similar age-related changes in economic decision-making are explained by a decrease in Pavlovian attraction toward reward. However, despite the role of Pavlovian biases in linking action with reward and avoidance with punishment, their impact on motor decision-making is unclear. To address this, we developed a novel app-based motor decision-making task (n = 26,532). We found that motor decision-making was subject to Pavlovian influences. Although we found age-related changes for both punishment and reward-based decision-making processes, the most striking effect of ageing was a decrease in the facilitatory effect of Pavlovian attraction on action in pursuit of reward. Using data from an independent economic decision task in the same individuals (n = 17,220), we demonstrate similar decision-making tendencies for motor and economic domains across a majority of age groups. Hence, Pavlovian biases play an essential role in not only explaining motor decision-making behaviour but also the changes which occur through normal ageing.
| Optimal decision-making requires choices that maximise reward and minimise punishment. Two key mechanisms play an important role in shaping the level of sub-optimality observed; a flexible, instrumental, value-dependent process, and a hard-wired, Pavlovian, value-independent process [1–3]. Choice behaviour in economic decision-making tasks has been widely studied and is often described using parametric decision models based on prospect theory that operationalise instrumental (value-dependent) concepts such as risk preference and loss aversion [4–7]. However, recent studies showed that changes in Pavlovian biases, which promote action towards reward and inaction in the face of punishment irrespective of option value [2, 8, 9], were able to account for aberrant choice behaviour during economic decision-making tasks. For example, the diminished economic risk-taking observed in older adults can be better explained by a reduction in dopamine-dependent Pavlovian attraction to potential reward [8, 10]. Importantly, value-dependent parameters based on prospect theory provide a poorer explanation of these changes in choice behaviour, suggesting that Pavlovian processes play a key role in the age-related changes observed during economic decision-making.
Motor decision-making, a unique type of decision-making, requires weighing potential rewards and punishments against the probability of successfully executing an action [11–14]. Motor decision-making has primarily been explained in the context of instrumental-based processes [14–19]. Within this framework, older adults display reduced risk-seeking behaviour [15]. However, given recent findings in economic decision-making [10], we asked whether Pavlovian biases might provide a more parsimonious explanation of age-related changes in motor decision-making. Although there is strong evidence that Pavlovian biases shape motor performance [20–23], and that healthy ageing leads to a reduction in Pavlovian biases on motor performance [24, 25], it is currently unknown whether Pavlovian biases influence motor decision-making. Sampling a large population through an app-based motor decision-making game, we provide a novel demonstration that Pavlovian biases have a substantial impact on motor decisions, and are able to account for age-related changes in choice behaviour during motor decision-making.
We developed a novel app-based motor decision-making task that examined participant sensitivity to reward (gaining points) and punishment (losing points) when making a ‘go/no-go’ decision based on their perceived ability to successfully execute a motor action (Fig 1A and 1B). Using an app-based platform (‘How do you deal with pressure?’ The Great Brain Experiment: http://www.thegreatbrainexperiment.com/) [9, 26, 27], we obtained data from a large cohort (n = 26,532; 15,911 males; S1 Fig) in which six age groups were considered: 18-24yrs: n = 5889; 25-29yrs: n = 4705; 30-39yrs: n = 7333; 40-49yrs: n = 4834; 50-59yrs: n = 2452; and 60+yrs: n = 1319 (Fig 1C) [9, 26, 27].
The game required participants to sequentially tap 5 targets distributed along a pre-defined path that could vary in both curvature and direction (Fig 1A and 1B; see Methods). If a participant successfully tapped all 5 targets within 1.2 seconds, then the action was considered a success. There were 7 different target sizes, with the task becoming progressively more difficult as target size decreased (Fig 1A and 1B; see Methods). At the beginning of each trial, participants saw the required action (e.g., the target size and trajectory) and were asked whether they wanted to take the motor gamble. There were two types of trials: reward and punishment. For reward trials, participants had to decide whether to skip the trial and stick with a small reward (10 points) or gamble on successfully executing the tapping action (Fig 1B). If successful they received a greater reward (20, 60 or 100 points) or 0 points if they failed. For punishment trials, participants had to decide whether to skip the trial and stick with a small punishment (-10 points), or gamble on successfully executing the tapping action (Fig 1A). If successful, they avoided the small punishment (lose 0 points) but failure resulted in a greater punishment (-20, -60 or -100 points). Participants began with 250 points and the overall goal was to accumulate as many points as possible. All trial-by-trial data (including tasks parameters, behavioural results, modelling results and accompanying code) are available on our open-access data depository (https://osf.io/fu9be/).
We found that older adults won fewer total points than younger adults (Fig 1D; r = -0.056, 95%CI = [-0.069, -0.044], p<0.001; all age-related r values represent a partial correlation between the measurement of interest and age, whilst controlling for the effects of gender and education; the values in square brackets represent bootstrapped 95% CI; p values were computed by permutation tests; see Methods). The final points accumulated during this task were dependent on (1) the decisions made (to gamble or not) and (2) the motor performance (success rate of executing the tapping action). Therefore, prior to examining participant gambling choices it was crucial to determine whether motor performance differed across age groups.
Although success on the motor task was similar across age groups (Fig 1E; r = 0.009, 95%CI = [-0.003, 0.021], p = 0.142), older adults used devices with larger screen sizes than younger age groups (Fig 1F; r = 0.233, 95%CI = [0.221,0.245], p<0.001, S1 Fig). As target size was scaled to device screen size (see Methods), we assessed how the relationship between age, target size and screen size affected motor performance. We found that decreased success rate was linked to a combination of smaller target sizes, smaller screen sizes and older age (Fig 1G, multiple linear regression final model: success rate = 0.78–0.60 * targetsize + 0.10 * screensize −0.25 * age * targetsize + 0.24 * targetsize * screensize + 0.20 *age * targetsize * screensize; F(7,179043) = 6940, p<0.001, R2 = 0.213; all factors were normalized between 0 to 1). In addition, by examining the position of each tap relative to the centre of the 5 targets during successful trials, we found that all age groups showed a similar level of precision across levels and screen sizes (S2O–S2U Fig). However, there were clear differences in movement time (S2H–S2N Fig). Specifically, even during the easier levels, older adults moved significantly slower than younger adults (levels 1–3: r = 0.158, 95%CI = [0.154,0.162], p<0.001; partial correlation between movement time and age, whilst controlling for the effects of screen size). Therefore, this suggests that when the task became more difficult and required greater precision, older adults tended to fail more often (S2F and S2G Fig) because they were unable to reduce their movement time (given the time limit of 1.2 seconds). We next assessed choice behaviour in the context of how these factors influenced motor performance on a trial-by-trial basis.
As mentioned above, participants were asked to make decisions between a gamble option and a certain option. Each option can be characterised by its potential outcomes, weighted by the probability of each outcome (i.e. Expected Value [28]). For the gamble option, the expected value is given by: EVgamble = PsuccessVsuccess+(1-Psuccess)Vfailed, where Psuccess is the probability of successfully executing the tapping action; Vsuccess is the points received if successful; Vfailed is the points received on failure. The expected value of the certain option (EVcertain) is Vcertain as the probability of receiving this value is 1. We calculated Psuccess by estimating the probability of motor success based on a participant’s age, screen size of the device used and target-size level. Specifically, the probability of success for a participant within a certain age group, using a certain screen size and facing a certain target size on each trial was estimated using the average success rate across all the participants with the same age, same screen size, and facing the same target size (Fig 1G; see Methods). By comparing choice behaviour, given the difference in expected value between these two options (EVgamble-EVcertain), we were then able to examine the influence of ageing on motor decisions while controlling for differences in motor performance due to age, screen size and target size. However, this formulation relied on an assumption that participants had a good estimate of their probability of success. To test whether this was true, we recruited an additional 120 participants (20 in each age group) who were asked to estimate their probability of success (from 0% to 100% in steps of 10%; see Methods) after being shown the target size and trajectory. After this estimate, they were then asked to perform the tapping action (whilst ignoring the decision-making part of the game). Each participant’s estimation performance was evaluated as the average error across 42 trials (Fig 2A). The error on each trial was calculated as: estimate % - 100% if successful, 0% if failed. Estimate performance did not differ across age groups (Fig 2A; one-way ANOVA: F(5,114) = 0.61, p = 0.695), and was not significantly different from zero (based on 6 t-tests, all p>0.05/6 = 0.0083). Therefore, similar to previous work [12, 14, 15], we found participants were able to estimate their probability of motor success reasonably well across all age groups (Fig 2B–2G). However, for all groups, the estimation performance was worst at extreme probabilities (Fig 2B–2G). This is in line with previous literature [29] and further explored with the use of weighted probabilities in the model (S1 Table).
We found a significant decrease in the proportion of trials in which participants chose to gamble across the lifespan in reward trials (Fig 3A; r = -0.186, 95%CI = [-0.198, -0.175], p<0.001), and to a lesser extent in punishment trials (Fig 3B; r = -0.053, 95%CI = [-0.065, 0.041], p<0.001). To understand these results, we examined age-related changes in choice behaviour given the difference between the options (EVgamble-EVcertain). Interestingly, in reward trials (warm-coloured-dotted lines in Fig 3C), there was a gradual and monotonic decrease in gamble rate across the lifespan which appeared to be independent of the EVgamble-EVcertain value. In contrast, for punishment trials (cool-coloured-triangle lines in Fig 3C), older adults displayed a higher gamble rate during high risk gambles (e.g., EVgamble-EVcertain = -90, where green lines (older adults) are above the blue lines (younger adults)), but conversely a relatively lower gamble rate during low risk gambles (e.g., EVgamble-EVcertain = 0 where green lines are below the blue lines).
Given these results, do older adults make worse motor decisions? An ideal (optimal) decision-maker chooses the option that has the higher expected value, and we therefore compared participant choice behaviour with the optimal behaviour. Specifically, we calculated whether the best decision on each trial was to gamble (if EVgamble-EVcertain>0, coded as 1) or decline (if EVgamble-EVcertain<0, coded as 0). We then subtracted this value from the observed choice of the participant (also coded gamble = 1, decline = 0). If the average absolute difference between these values across trials was 0, then a participant was deemed an optimal decision-maker. In reward trials, there was progressive deviation from optimality across the lifespan (Fig 3D; r = 0.258, 95%CI = [0.245,0.270], p<0.001). In contrast, for punishment trials, all age groups showed a similar level of sub-optimality (Fig 3E; r = 0.001; 95%CI = [-0.011, 0.013], p = 0.832). Therefore, the most pronounced effect of ageing on motor decision-making was a value-independent decrease in gamble rate during reward trials which led to a significant decrease in optimality.
While these data portray many similarities with decision-making under risk [4, 5], there are also clear differences. For example, decision-making models based on prospect theory are not able to explain the gradual, monotonic and value independent decrease in gamble rate across the life span observed during the reward trials [8] (Fig 3C). We predicted that such dichotomies represented the contribution of value-independent Pavlovian approach-avoidance biases to motor decision-making behaviour. To test this prediction, we modelled the choice behaviour using an established decision-making model based on prospect theory, and a newly introduced model that included Pavlovian approach-avoidance parameters [8, 10] (see Methods). For the prospect theory parametric decision-making model, the utility of the gamble option is given by Ugamble = Psuccess x (Vgamble)alpha; and the utility of the certain option is Ucertain = (Vcertain)alpha, where the risk preference parameter (alpha: α) represents the diminishing sensitivity to change in value with an increase in absolute value (S3 Fig). The gamble probability is then a stochastic function of value that is described by the softmax function: Pgamble = (1+exp(-μ(Ugamble- Ucertain)))-1. The logit parameter μ is the sensitivity of the choice probability to the value difference. In addition to these parameters (α,μ), the Pavlovian approach-avoidance model included a value-independent parameter (δ). Specifically, gamble probability is further decided by the parameter δ: Pgamble = (1+exp(-μ(Ugamble- Ucertain)))-1+δ. Positive or negative values of the Pavlovian parameter correspond, respectively, to an increased or decreased probability of gambling without regard to option values (see Methods; Eq 3). According to Akaike’s information criterion (AIC) [30], Bayesian information criterion (BIC) [31] model comparison (Fig 4; S1 Table; see Methods) and model/parameter recovery analysis (S4 Fig; see Methods), we found that an approach-avoidance decision model with 4 parameters (a joint risk preference parameter: α, the inverse temperature parameter: μ, value-independent parameters exclusively for reward δ+ and punishment δ− trials) fitted the motor gamble (choice) data better than other decision models (Fig 5; see Methods). To confirm this result, we performed a repeated-measures ANOVA across all participants using the log-evidence (AIC/BIC) as a summary statistic for each model [32]. The ANOVA showed a significant difference between models (BIC: F(23,529) = 5085.9, p<0.001; AIC: F(23,529) = 4011.6, p<0.001). Posthoc paired t-tests (with Bonferroni correction for multiple comparisons) showed that AIC/BIC values for the winning model ([α, μ, δ+, δ-]; S1 Table) were significantly lower (lower value = preferred model) than for all other models (p<0.001; Fig 4).
Based on the preferred approach-avoidance decision model ([α, μ, δ+,δ-]; S1 Table), we observed age-related changes across the reward and punishment domains for both risk preference and Pavlovian parameters. However, the most striking effect was a decrease in the facilitatory effect of Pavlovian attraction on action in pursuit of reward (δ+). Specifically, we found that healthy ageing did not affect the stochasticity parameter, μ (r = -0.005, 95%CI = [-0.017, 0.0065], p = 0.390), but was associated with a decrease in the risk preference parameter, α (Fig 6A and 6D; r = -0.115, 95%CI = [-0.126,-0.103], p<0.001). The winning model included a α parameter, but this represented different value-dependent biases in reward and punishment (α<1 indicated risk aversion in reward domain and α<1 represented risk-seeking in punishment domain; α = 1 represented risk-neutral; see Methods). Therefore, the reduction in the alpha parameter value across age groups reflected reduced sensitivity to the value change, resulting in greater risk-aversion in reward and risk-seeking in punishment. However, the enhanced risk-seeking in the punishment domain was offset by the fact that ageing was also linked with greater Pavlovian avoidance, causing a reduction in gamble rate (Fig 6B and 6D; δ−; r = -0.076, 95%CI = [-0.089,-0.064], p<0.001), an effect not previously observed in economic decision-making [10]. Nevertheless, the largest impact of ageing was a substantial decrease in Pavlovian attraction (Fig 6B and 6D; δ+; r = -0.138, 95%CI = [-0.150,-0.126], p<0.001). We found a similar decline for both sexes (S5 Fig; male: n = 15911, r = -0.134, 95%CI = [-0.154,-0.116], p<0.001; female: n = 10621, r = -0.140, 95%CI = [-0.156,-0.124], p<0.001), and across all education levels (S6 Fig; school: n = 9171, r = -0.146, 95%CI = [-0.171,-0.122], p<0.001; university: n = 11281, r = -0.129, 95%CI = [-0.148,-0.111], p<0.001; advanced: n = 6080, r = -0.125, 95%CI = [-0.151,-0.100], p<0.001). Importantly, we did not observe this age-related effect for the temperature parameter (μ; Fig 6D), indicating the changes observed in the risk and Pavlovian parameters were not simply a result of large participant numbers.
Finally, through this app-based platform a subset of participants (n = 17,220) also performed an economic decision-making gambling task in which a similar approach-avoidance model was used to explain choice behaviour [10] (S7 Fig). The data and model have previously been published [10]. Briefly, this work compared a single approach-avoidance ([α+, α-, μ, δ+, δ-]) and prospect theory model ([α+, α-, μ]), with the approach-avoidance model fitting the data better [10]. In addition, analysis of laboratory data using a similar economic decision-making task confirmed that the approach-avoidance model outperformed reasonable alternative models including simpler models based on prospect theory [8]. As the economic decision-making app data revealed a similar age-related decline in the Pavlovian approach parameter [10] (S7 Fig), we compared the subset of participant’s, who performed both tasks, parameter values across the two winning approach-avoidance models. A small yet significant positive relationship was found between the tasks for risk preference (Fig 7A and 7B; r(α+) = 0.05, 95%CI = [0.04,0.07], p<0.001; r(α-) = 0.08, 95%CI = [0.06,0.09], p<0.001) and Pavlovian parameters (Fig 7C and 7D; r(δ+) = 0.07, 96%CI = [0.06,0.09], p<0.001; r(δ−) = 0.14, 95%CI = [0.12,0.15], p<0.001). However, we did not observe this correlation for the temperature parameter (Fig 7E; r(μ) = 0.004, 95%CI = [-0.01,0.02], p = 0.587). This relationship was relatively consistent within the first 5 age groups (S8 Fig, S9 Fig). However, although the oldest age groups (60+) showed a similar trend, the positive correlation was not significant (S8 Fig, S9 Fig). Within this age group, it is possible that we did not have enough power (participant numbers) to reliably detect a significant correlation given our observed effect sizes. Specifically, the 60+ age group (n = 663) achieved 0.3–0.69 power (desired level is usually >0.80) to detect a significant correlation across the 4 parameters (S2 Table).
Making decisions under uncertainty is crucial in everyday life, whether it is managing retirement funds, choosing a career, or deciding between pulling out or not on to a busy road whilst driving. The latter example describes motor decision-making, a unique kind of decision that requires weighing potential rewards and punishments against the probability of successfully executing an action [11–14], and often with immediate outcomes. Although healthy ageing has been associated with reduced risk taking during motor decision-making [15], the underlying mechanism was not clear. We contributed answers to this question using a novel motor gambling task that exploited an app-based platform. This enabled us to collect a large cohort of data. Unlike previous work on motor decision-making [11–14], we considered choice behaviour in relation to both value-dependent instrumental and value-independent Pavlovian processes [2, 8, 9]. We found age-related changes across the reward and punishment domain for both value-dependent and independent parameters. However, the most striking effect of ageing was a decrease in the facilitatory effect of Pavlovian attraction on action in pursuit of reward. Through this app-based platform, we also compared a subset of participant choice behaviour on the motor decision-making task and a separate economic decision-making task [10]. We found similar decision-making tendencies across motor and economic domains.
Our large cohort and use of a newly established approach-avoidance computational model [10, 34] enabled us to detect subtle age-related changes in choice behaviour and surprising interactions between value-independent and value-dependent processes. For instance, the risk aversion parameter (α: instrumental value-dependent process) was on average less than 1 across all age groups, indicating risk aversion in reward, and risk seeking in punishment. The use of a joint risk preference parameter (alpha) across reward and punishment signified a similar degree of value-dependent bias across gain and loss trials (risk aversion in reward and risk-seeking in punishment). This is in contrast to the economic decision-making data (10) in which separate alpha values were used, signifying subtle differences across the tasks. Importantly, this parameter progressively decreased with age, suggesting that older adults showed increasing value-dependent biases, i.e., more risk aversion in reward and more risk seeking in punishment. This is in line with previous economic decision-making work which revealed older adults weigh certainty (achieving the small reward or avoiding the small punishment) more heavily than younger adults [35]. Interestingly, the greater risk-seeking in the punishment domain was offset by the fact that ageing also led to greater Pavlovian avoidance, an effect not observed in economic decision-making [10]. It is the interaction between value-dependent and independent parameters that help explain not only the complex changes observed with ageing during punishment trials but also the lack of difference across age groups for punishment-based optimality. Crucially, previous work in motor decision-making using only instrumental-based processes would not have detected such complex behavioural interactions [13–15]. The underlying mechanism behind age-dependent increases in Pavlovian avoidance is unknown. It has been suggested that the neurobiology behind Pavlovian avoidance may involve opponency between the dopaminergic and serotonergic systems [36]. Despite there being evidence of age-related decline in serotonin receptor availability [37], it remains an open question as to the link between serotonin and Pavlovian avoidance during either motor or economic decision-making.
The strongest effect of ageing was a decrease in Pavlovian attraction to reward. As all age groups displayed risk aversion during reward trials, this decrease in Pavlovian attraction led to reduced optimality in older adults. These results are strikingly similar to the ones observed in economic decision-making [10], suggesting Pavlovian attraction plays a pivotal role in explaining age-related changes to reward across both motor and economic decision-making. During economic decision-making, boosting dopamine with L-DOPA increases the influence of Pavlovian attraction on choice behaviour [8]. In addition, healthy ageing is associated with a gradual decline in dopamine availability [38, 39] and neural responses to reward [40]. Therefore, it is possible that the decrease in Pavlovian attraction during motor decision-making in older adults is a result of an age-dependent decrease in dopamine availability.
More broadly, the current work shows the importance of both instrumental value-dependent and Pavlovian value-independent processes in motor decision-making. However, task design may play an important role in determining the size of Pavlovian influences. Here we used a ‘go/no-go’ decision-making task as previous literature has shown the ‘go/no-go’ component induces strong Pavlovian biases [2, 22, 41]. It remains to be seen whether computational models including Pavlovian biases provide a better description of choice behaviour during other motor decision-making tasks which do not involve a ‘go/no-go’ component [14–19].
We found that older participants gambled less frequently in both gain and loss trials. Although we modelled this change with Pavlovian approach-avoidance, this decrease in overall gamble rate could also be explained by other value-independent factors that make the gamble option less ‘attractive’ for the older population such as an increased sensitivity to the cost of physical effort [42], age-related changes in motor performance or an age-related underestimation of movement success. However, one critical feature of Pavlovian approach-avoidance is that it has opposite effects on gain and loss trials, i.e., ‘approach’ in gain trials and ‘avoidance’ in loss trials. This is the main reason that the approach-avoidance model was able to explain data significantly better than the prospect theory model. In particular, across age groups, participants gambled more often in the gain trials relative to loss trials when the EVgamble—EVcertain was close to zero. Such a result cannot be explained without an opposing effect on gain and loss trials, which shifts gamble rate higher for gain trials and lower for loss trials. To our knowledge, there is no evidence to suggest that an age-related change in effort cost, motor performance or estimation of movement success would have such opposing effects on gain and loss trials.
Finally, participants showed similar decision-making tendencies for both instrumental (value-dependent) and Pavlovian (value-independent) parameters across motor and economic domains for all age groups except 60+. At present, we are unsure whether the non-significant correlation in the 60+ age group is a true reflection of motor and cognitive decisions becoming dissociated with age, or simply a lack of power within this age group to observe small effect sizes. Despite this, the general association between cognitive and motor decision-making extends previous work that revealed a similar relationship with parameters derived from parametric decision models based on prospect theory [14, 16] and reinforces the view that the mechanisms which control cognitive (economic) and motor decision-making are somewhat integrated [43]. However, the correlation between the tasks was small, around r = 0.1, suggesting that while participants showed similar behavioural trends across the two tasks, their performance in motor and economic domains was also distinct [16]. Interestingly, the approach-avoidance model not only fitted choice data substantially better for the motor decision-making task, relative to the economic task, but the effect size relating to age was also nearly double in size for all parameters [10]. This indicates that while there are clear similarities between cognitive and motor decision-making, computational models including Pavlovian biases appear to be particularly important for explaining motor decision-making. Despite their role in linking action and reward and inaction with punishment, descriptions of Pavlovian biases are surprisingly absent in current computational models of motor function.
In conclusion, Pavlovian biases play an important role in not only explaining motor decision-making behaviour but also the changes which occur through normal ageing. This provides a greater understanding of the processes which shape motor decision-making across the lifespan, and may afford essential information for developing population wide translational interventions such as promoting activity in older adults.
We tested 26,532 participants (15,911 males, aged 18–70+) who completed the task between November 20, 2013 and August 15, 2015. Data were only included if users fully completed the game and it was their first attempt. All participants gave informed consent and the Research Ethics Committee of University College London approved the study. We also recruited an additional 120 participants (49 males, aged 18–70+) who were asked to estimate their success rate (motor performance). These participants were mainly recruited by email advertisement sent to staff and students at the University of Birmingham.
Using an app-based platform (The Great Brain Experiment: www.thegreatbrainexperiment.com) we developed a motor decision-making task (‘How do I deal with pressure?’) which is freely available for Apple iOS and Google Android systems. The game runs in a 640x960 (3:4 ratio) pixel area, which is then scaled to fit the screen whist maintaining this ratio. The game required participants to ‘throw’ a ball at a coconut in an attempt to knock it off its perch. This was achieved by tapping 5 sequential targets along a pre-defined path. The path was characterised by an angle parameter that represented a section of a sine curve, in degrees. The curves were drawn from the bottom (the starting point) to top of the game window (Fig 1A and 1B). For example, if the angle parameter was 360, then one complete cycle of the sine curve was used to draw the curve. During the task, the angle was randomly chosen between 0 and 360. The 5 targets were evenly spaced along the curves. If the participant tapped all 5 targets sequentially (from bottom to top) within 1.2 seconds, then the action was considered a success and the coconut was hit. If the participant failed to tap all 5 targets accurately or within the allotted time then the action was considered a failure and the ball sailed past the coconut. In addition, the action was deemed a failure if participants did not start the tapping action within 7 seconds after they chosen to do the tapping. There were 7 different target sizes across trials with the tapping action becoming more difficult as the target size was reduced. However, as mentioned above, the game interface was scaled to screen size. Therefore, motor performance (success rate) was examined relative to the interaction between target size and screen size (Fig 1G). All trial-by-trial data (including tasks parameters, behavioural results, modelling results and accompanying code) are available on our open-access data depository (https://osf.io/fu9be/).
At the beginning of each trial, participants were shown the action required (i.e. the position and size of the 5 targets) and were asked to make a motor gamble. There were two types of trials: reward and punishment (Fig 1A and 1B). For reward trials, participants had to decide whether to skip the trial and stick with a small reward (10 points) or gamble on successfully executing the ‘throw’. If successful they received a greater reward (20, 60 or 100 points) but 0 points if they failed. For punishment trials, participants had to decide whether to skip the trial and stick with a small punishment (-10 points) or gamble on successfully executing the ‘throw’. If successful they lost nothing (0 points) but failure resulted in a greater punishment (-20, -60 or -100 points). Hence, there were 6 value combinations. Each combination was repeated for each of the 7 different target sizes (6 values x 7 target sizes = 42 trials). Although there were 7 blocks of the game this did not directly relate to the 7 target sizes. In order to maintain a level of unpredictability, the first 3 blocks included random presentation of the 3 largest (easiest) target sizes, the next 3 blocks included target sizes 4–6 and the final block included the smallest (most difficult) target size. Participants began with 250 points and the overall goal was to accumulate as many points as possible.
For the control study (Fig 2), which examined participant ability to estimate their probability of success, individuals were asked to estimate their probability of motor success (0% to 100% in steps of 10%) after being shown the target size and trajectory. After this estimate, they were then asked to perform the tapping action, whilst ignoring the decision-making part of the game. This experiment was not performed in the laboratory but across campus at the University of Birmingham with the first 60 participants (10 per group) using their own mobile device and the second 60 participants (10 per group) using a device provided with a screen size of 5.1 inches. Screen sizes of the devices used for first set of 60 participants had a similar profile as in the main experiment (Fig 1 and S1 Fig). Importantly, we did not find a significant difference in estimate performance with (M = -0.023, SD = 0.126) or without (M = -0.011, SD = 0.139) screen size control (t(118) = -0.515, p = 0.607; S10 Fig).
Matlab (Mathworks, USA, 2016a) was used for all data analysis. We report partial correlation coefficients (r) for the relationships between the measures of interest (e.g., final points achieved, screen size of the devices, sub-optimality, model parameters) and age, whilst controlling for the effects of gender and education. Specifically, the partial correlation between X (e.g., screen size) and Age given the controlling variables Z = [gender, education] was given by the correlation between the residuals eX and eAge resulting from the linear regression of X with Z and of Age with Z, respectively. These were implemented using the Matlab function ‘partialcorr’. Spearman rank correlation was used as the participants were asked to identify themselves into one of the age groups provided, therefore the variable Age was ordinal. Bootstrapped 95% confidence intervals (CI) were computed based on 10,000 resamples with replacement with the bias-corrected and accelerated (BCA) bootstrap method. This was implemented using the matlab function ‘bootci’. All p values were computed based on permutation tests using 10,000 random shuffles of age labels to determine null distributions [10]. Bootstrapped and resampling techniques were used because (1) previous studies have shown that the coverage of bootstrapped Cis is as good or better than coverage of analytic CIs, especially when using Spearman’s correlation with ordinal data [44] and (2) Bootstrapping and resampling techniques, based on recent work [45], have proven to be a robust alternative to analytic statistical techniques.
On each trial participants faced a gamble that contained a certain option involving a payoff of certain points (+10 in reward trials and -10 in punishment trials), and a gambling option in which the outcome depended on a probability of successfully executing the tapping action. This probability was estimated given a participant’s age, screen size of the device used and target-size level (Fig 1G). Specifically, the probability of success for a participant within a certain age group, using a certain screen size and facing a certain target size on each trial was estimated using the average success rate across all the participants with the same age, same screen size, and facing the same target size. Given the small amount of trials we had for each participant at each target size to estimate their probability of success, we believed this group average approach was the most valid estimate of success probability. However, we also conducted the analysis when success probability was estimated based on each individual’s own data (i.e. the probability of success for a participant facing a certain target size was estimated using their own success rate over the same target size). Importantly, our findings still hold (S11 Fig). We modelled participant motor gamble choices using an established decision-making model based on prospect theory [5] and a newly introduced model which included an extra Pavlovian approach-avoidance [8, 10, 34] component. In the following, we first describe the prospect theory models, followed by the approach-avoidance models.
Parametric decision-making model based on prospect theory: There are three key components in prospect theory models. The first component is the value function. According to prospect theory, the subjective desirability of outcomes is modelled as transformations of objective task quantities. The subjective desirability of the outcomes, v(O) was modelled by a value function (2-part power function) of the form:
v(O)={OαifO≥0−λ∙(−O)αifO<0
Eq 1
where, the risk preference parameter (α) represents the diminishing sensitivity to changes in values as the absolute value increases (if α < 1). The risk preference parameter (α < 1) is equivalent to risk aversion in the reward domain and risk seeking in the punishment domain, as demonstrated by the following examples. Imagine a gamble between a probabilistic reward: 50% of £20; 50% of £0 and a sure reward of £10. The objective expected value of the gamble is £10, similar to the certain option. Hence a risk neutral person would be indifferent between these two options. If α = 0.8, the gamble would have a subjective value of 5.49, and the certain option would have a subjective value of 6.31, which results in participants being more likely to choose the certain option (i.e., risk aversion). Now imagine a gamble between a probabilistic punishment 50% of -£20; 50% of £0 and a sure punishment of -£10. The objective expected value of gamble is -£10, similar to the certain punishment option. If α = 0.8, the gamble would have a subjective value of -5.49, and the certain option would have a subjective value of -6.31, which results in participants being more likely to choose the gamble option (i.e., risk seeking). The loss aversion coefficient (λ) represents the weighting of losses relative to gains, which was set to 1 as we did not have gambles with both positive and negative outcomes. The second component of a prospect theory model is the probability weighting function. Most prospect theory models assume that probabilities are weighted non-linearly. However, we found that the probability weighting parameter (γ) did not significantly improve the model fit (S1 Table, we used a 1-parameter probability weighting function [46]: w(p) = exp(−(−ln(p)γ))). Hence, probabilities and utilities were combined linearly in the form: U(p,O) = p * v(O). The third component of a prospect theory model is the choice function. The probability of choosing to gamble is given by the logit or soft-max function:
F(p,O1,O2,Oc)=(1+exp[−μ(U(p,O1,O2)−U(Oc))])−1
Eq 2
where O1 and O2 are the outcomes in the gamble option [p→O1; (1-p)→O2], and Oc is the outcome of the certain option. The logit parameter μ is the sensitivity of the choice probability to the utility difference. In summary, the prospect theory models included the following free parameters: risk preference parameter (α) and stochasticity of decision-making according to the inverse temperature parameter (μ).
Parametric approach-avoidance decision model: approach-avoidance models were based on the Prospect Theory models, but with an additional component that allows for value-independent influences to choose or not choose gambles. Specifically, Pavlovian parameter (δ) were added to the probability of choosing to gamble (Eq 2) as follows:
F(p,O1,O2,Oc)=(1+exp[−μ(U(p,O1,O2)−U(Oc))])−1+δ
F(p,O1,O2,Oc)=max(0,min(F(p,O1,O2,Oc),1)
Eq 3
Positive or negative values of the parameter (δ) correspond respectively to an increased or decreased probability of gambling without regard to the value of gamble. Other parts of the models were identical to the Prospect Theory models. In summary, the approach-avoidance model included the following free parameters: risk preference parameter (α), stochasticity of decision-making according to the inverse temperature parameter (μ) and Pavlovian parameter (δ).
For each model, a set of parameters for that model and a sequence of outcomes observed by an individual participant can be used to calculate the probability of gambling on each trial (Eq 2, Eq 3). The full joint probability of gambling, given a set of parameters for a participant, is provided by the product of the probability for each response actually made. Therefore, the likelihood of a set of parameters given the data is defined by the probability of the data given the parameters. For each participant, we took the sum of the log likelihood of the parameters (Eq 4). For each participant and model, we estimated the parameters which maximized this likelihood by using the search function fmincon in Matlab (minimizing the negative of the log likelihood).
L(Θ|y,p,O1,O2,Oc)=∑i=1Nyilog(F(p(i),O1(i),O2(i),Oc(i),Θ))+(1−yi)log(1−F(p(i),O1(i),O2(i),Oc(i),Θ))
Eq 4
Where, i indexes the trial number; N is the number of trials; yi indicates participant choice on trial i, (gamble = 1, skip = 0); Θ indicates the parameter vector to be estimated; (p,O1,O2,Oc) represent the gamble options on each trial. Parameters were constrained to the following ranges: α: [0,1]; μ: (0,10]; δ: [−1,1]. The fitting was repeated at 200 random seed locations to avoid local minima.
For each key parameter of prospect theory and approach-avoidance models, we explored the possibility of using separate and joint parameters for reward and punishment domains as well as a weighted or linear probability function. Therefore, we fitted each participant’s choice data with 24 models (S1 Table).
We used Akaike’s information criterion (AIC) [30] and Bayesian information criterion (BIC) [31] to compare model fits. Both of these represent a trade-off between the goodness of fit and complexity of the model and thus can guide optimal model selection. AIC and BIC were given by AIC = −2logL + 2k and BIC = −2logL + klogN respectively, where L is the likelihood (Eq 4), k is the number of parameter, N is the number of data points. Lower AIC and BIC values imply a better fit to the data. AIC and BIC were calculated for each participant and each model. AIC and BIC for each model (Fig 4; S1 Table) was its sum over all participants. Pseudo r2 was calculated with the null model in which α, μ and δ were restricted to 0 (pseudor2=1−ln(L^(model))ln(L^(nullmodel)), where L^ = Estimated likelihood).
To justify our model selection, we performed model/parameter recovery analysis [47] and model falsification [33]. To reiterate, the PT model had two key parameters (α,μ), and AA model had three key parameters (α,μ,δ). If the fitted parameters were reliable, we should be able to take simulated data with known parameters, and estimate those parameters. For both models, we chose parameters to represent “typical participants” and generated simulated responses to participants’ observed outcomes. We used the same process as for the original participant responses to estimate parameters for these simulated responses. S3 Fig shows that the fitted parameters for the PT and AA models (based on 50 simulations for each parameter set) are clustered around the parameters used for data generation, suggesting that the parameters were reliable. As shown in S1 Table, for each key parameter, we also explored the possibility of using separate and joint parameters for gain and loss trials as well as a weighted or linear probability function. First, we found that the models with weighted probability did not fit the data better than their linear probability counterpart (based on pseudo r2) even with one extra free parameter. Hence, there is no strong evidence in our data set for this free parameter. Second, we compared models with and without separate gain and loss parameters. Fig 4 suggests we have the strongest evidence for using separate δ for gains and loss trials, but weaker evidence for separate α and μ. We also ran a likelihood ratio test for each individual, which compares the goodness of fit of the null model: joint parameter and the alternative model: separate parameters. The test decides whether or not to reject the null model. We found that that using separate δ explained an additional 9391 participants (p<0.05), an extra 2596 participants were explained with separate α, and an additional 1375 participants were explained with separate μ. Finally, Fig 5 justifies our preference for the AA model (α, μ, δ+, δ-) relative to the PT model as it clear that the PT model was unable to generate the behavioural pattern observed (model falsification; [33]). Therefore, all model selection analysis supports our conclusion that the AA model ID = 10 (α, μ, δ+, δ-) is the most likely model.
|
10.1371/journal.pgen.1002129 | Increased Susceptibility to Cortical Spreading Depression in the Mouse Model of Familial Hemiplegic Migraine Type 2 | Familial hemiplegic migraine type 2 (FHM2) is an autosomal dominant form of migraine with aura that is caused by mutations of the α2-subunit of the Na,K-ATPase, an isoform almost exclusively expressed in astrocytes in the adult brain. We generated the first FHM2 knock-in mouse model carrying the human W887R mutation in the Atp1a2 orthologous gene. Homozygous Atp1a2R887/R887 mutants died just after birth, while heterozygous Atp1a2+/R887 mice showed no apparent clinical phenotype. The mutant α2 Na,K-ATPase protein was barely detectable in the brain of homozygous mutants and strongly reduced in the brain of heterozygous mutants, likely as a consequence of endoplasmic reticulum retention and subsequent proteasomal degradation, as we demonstrate in transfected cells. In vivo analysis of cortical spreading depression (CSD), the phenomenon underlying migraine aura, revealed a decreased induction threshold and an increased velocity of propagation in the heterozygous FHM2 mouse. Since several lines of evidence involve a specific role of the glial α2 Na,K pump in active reuptake of glutamate from the synaptic cleft, we hypothesize that CSD facilitation in the FHM2 mouse model is sustained by inefficient glutamate clearance by astrocytes and consequent increased cortical excitatory neurotransmission. The demonstration that FHM2 and FHM1 mutations share the ability to facilitate induction and propagation of CSD in mouse models further support the role of CSD as a key migraine trigger.
| We previously reported that mutations of the α2 subunit of the Na,K-ATPase cause familial hemiplegic migraine type 2 (FHM2), a dominant form of migraine with aura. This paper describes the first animal model of FHM2 and represents the further proceeding in this disease investigation. Homozygous knock-in mutant mice die just after birth, while heterozygous mice show no apparent clinical phenotype. However, in vivo analysis revealed a marked facilitation of cortical spreading depression (CSD), the phenomenon underlying migraine aura. Given the evidence for specific functional coupling between the glial α2 Na,K pump and glutamate transporters, we hypothesize that CSD facilitation in the FHM2 mouse model is sustained by inefficient glutamate clearance by astrocytes and consequent increased cortical excitatory neurotransmission. We finally propose this FHM2 mouse as a valuable in vivo model to investigate migraine mechanisms and, possibly, treatments.
| Migraine is a clinically heterogeneous disorder affecting more than 10% of the general population. It generally occurs with unilateral and pulsating severe headache often accompanied by nausea, photophobia and phonophobia. In approximately one third of migraineurs, the headache attack is preceded by aura, a transient neurological symptom that are most frequently visual but may involve other senses [1]. The migraine attack is triggered by a brain dysfunction that leads to activation and sensitization of the trigeminovascular system, particularly trigeminal nociceptive afferents innervating the meninges and lastly to headache [2], [3], [4]. Neuroimaging examination suggests that migraine aura is associated to cortical spreading depression (CSD), a short-lasting, intense wave of neuronal and glial cell depolarization. CSD spreads slowly over the cortex at a rate of approximately 2–5 mm/min and is followed by long lasting depression of neuronal activity [5], [6], [7], [8]. Experimental evidence on patients and animal models supports CSD as both underlying migraine aura [1], [7], [8], [9] and a key triggering event for trigeminal activation [10], [11], [12], although the role of CSD in migraine headache is still debated. As an indirect confirmation, several migraine prophylactic agents cause an increase of CSD initiation threshold [13].
Common migraine has a strong multifactorial genetic component, which is higher in migraine with aura (MA) than in migraine without aura (MO) [14], [15]. As for many other multifactorial diseases whose complexity hampers the investigation of the pathogenetic mechanisms, rare monogenic forms that phenocopy most or all the clinical features of the common disease are of great help for describing the complicated events leading to migraine. Familial hemiplegic migraine (FHM) is a rare autosomal dominant subtype of MA, whose aura symptoms include hemiparesis. Aura symptoms and headache duration are usually longer in FHM than MA, but all other headache properties are similar. FHM is genetically heterogeneous and is associated to mutations in three different genes. Mutations in CACNA1A [16], ATP1A2 [17] and SCN1A [18] genes are responsible for Familial hemiplegic migraine type 1 (FHM1), type 2 (FHM2), and type 3 (FHM3), respectively. The CACNA1A and SCN1A genes both encode neuronal voltage-gated ion channels, whereas the ATP1A2 gene encodes the α2 subunit of the Na,K-ATPase, hence suggesting a key role of cation trafficking in the pathophysiology of FHM.
Until now, more than 50 FHM2 mutations have been identified and most of these are missense mutations. A small fraction of mutations is represented by microdeletions [19] and a single mutation affecting the stop codon, which causes an extension of the ATP1A2 protein by 27 aminoacid residues [20]. Most of the ATP1A2 mutations are associated with pure FHM without additional clinical symptoms [17], [19], [20], [21], [22]. However, a number of FHM2 mutations have been associated to complications like cerebellar ataxia [23], childhood convulsions [24], epilepsy [25] and mental retardation [26]. Interestingly, ATP1A2 mutations associated with non-hemiplegic migraine phenotypes, such as basilar migraine and even common migraine have been reported [27], [28].
The Na,K ATPase is a P-type ion pump that utilizes the free energy of ATP hydrolysis to exchange Na+ for K+ and maintains gross cellular homeostasis. The functional pump is a heterodimer, consisting of one α catalytic subunit and one β subunit that is required for protein folding, assembling, membrane-addressing, and modulates substrate affinity [29]. The α subunit exposes both the amino- and carboxy- termini in the cytoplasm and crosses the plasma membrane with ten transmembrane segments (M1–M10) [30]. Four isoforms of α Na,K-ATPase (α1, α2, α3 and α4) are present in mammals [29], [31]. While no pathogenic mutations are known for the ubiquitous α1- and the testis α4-subunits, mutation in both α2 and α3 isoforms cause neurological diseases when mutated, FHM2 and rapid-onset dystonia parkinsonism, respectively [32]. While in the adult brain the α1 isoform is nonspecifically present in both neurons and glial cells and α3 is neuron-specific, the α2 isoform is essentially expressed in astrocytes [33].
Investigation of the functional consequences of FHM2 mutations in heterologous expression systems revealed that these mutations produce partial or complete loss of function of the α2 Na,K pump [34], [35], [36]. Here, we report the generation of the first mouse model of FHM type 2, a knock-in mutant harboring the W887R ATP1A2 mutation.
The W887R mutation localizes to the extracellular loop between M7 and M8, which includes the β subunit binding site [37] and was shown to produce the almost complete loss of pump activity [17], [38]. Homozygous Atp1a2R887/R887 mutants die just after birth, while heterozygous Atp1a2+/R887 mice are fertile and show no apparent clinical phenotype. However, heterozygous FHM2 mouse displays altered CSD properties, such as decreased threshold and increased velocity of propagation. We hypothesize that inefficient astrocyte-mediated clearance of glutamate from the synaptic cleft is a key event for the enhanced susceptibility to CSD in the FHM2 mouse.
With the aim of investigating the molecular pathogenesis of FHM type 2, we generated a knock-in mouse model by inserting an FHM2 mutation, the transition T2763C that causes the aminoacid replacement W887R in the Atp1a2 murine gene (construct details in M&M). The amino acid sequence conservation between human and mouse α2 Na,K-ATPase proteins is very high and, in particular, in the extracellular domain between transmembrane domains M7–M8, where W887R is located [17]. This mutation was one of the first two mutations reported to be associated to typical cases of the disease. Embryonic stem cells harboring the R887 and the neo cassette were injected in C57Bl/6J blastocysts and then transferred to pseudopregnant CD1 females. We obtained three chimeric mice, one of which transmitted the Atp1a2+/R887-neo allele through germline (Figure 1A). Heterozygous Atp1a2+/R887-neo mice were genotyped by Southern blot analysis (Figure 1C), are fertile and display no apparent phenotype. To remove the neo cassette that hampers the natural expression of the mutant allele, we crossed the Atp1a2+/R887-neo mice with transgenic mice expressing the Flippase recombination enzyme (FLPe) under the control of the human ACTB promoter (TgN(ACTFLPe)9205Dym; The Jackson Laboratory). Hence, we obtained the heterozygous Atp1a2+/R887 knock-in mice (Figure 1B), which are fertile as well and show no visible clinical phenotype. Contrary to heterozygous mice, homozygous Atp1a2R887/R887 mutants do not survive beyond the first day post partum, thus resembling the neonatal lethal phenotype of the Atp1a2 null mutant [39], which succumbs for dysfunctional neuronal activity and respiratory distress. Therefore, we addressed our investigation onto the heterozygous knock-in mouse, which shares the Atp1a2 gene asset with FHM2 patient. The general behavior of heterozygous Atp1a2+/R887 mice was tested by a modified SHIRPA protocol [40] that provides comparable quantitative data on animal motor, sensory, autonomic and neuropsychiatric functions. The scored parameters are summarized in Table 1. No major differences in the sensory-motor functions were observed between heterozygous Atp1a2+/R887 (n = 8) and wild-type (n = 6) mice, except for a higher fear and anxiety of Atp1a2+/R887 at the specific tests of transfer arousal and fear (p<0.05; Table 2).
Mutant and wild type Atp1a2 gene expression was evaluated at E19.5 in Atp1a2R887/R887 for lethality constrain and at adult age in Atp1a2+/R887 mutants. Semi quantitative reverse transcription PCR (RT-PCR) analysis showed that both wild type and mutant Atp1a2R887 alleles express equal amount of transcripts (Figure 2A, left panel). Each RT-PCR experiment was normalized on intra-sample actin transcript level. The nucleotide replacement resulting in the W887R mutation creates a new MspI restriction site that we used to confirm the R887 mutation in the Atp1a2R887 transcript and to quantify the mutant transcript in heterozygous mice as intra-sample control (Figure 2A, right panel). The expression of α2 Na,K-ATPase protein was assessed in embryonic brain. Immunoblot of total lysate and microsomal fractions revealed a markedly reduced amount of α2 Na,K-ATPase in the mutants, which displayed approximately half the level of wild type in the Atp1a2+/R887 mice. In the Atp1a2R887/R887 mutant, the R887 α2 Na,K-ATPase is barely observable (Figure 2B). In order to investigate whether the reduced amount of α2 isoform induces a compensatory increase of expression of the paralogous α1 and α3 isoforms in the adult, when the principal phenotype, CSD, is assessable, we analyzed brain tissues with specific α isoform antibodies. No differences in the level of α1 and α3 isoforms were observed in whole brain of Atp1a2+/R887 mice (Figure 2C) compared to wild type ones. On the contrary, the Atp1a2+/R887 model displays an α2 expression level reduced to approximately 50% in the cortex, 35% in cerebellum and 40% in total brain (Figure 2D).
The loss of α2 protein in the Atp1a2R887/R887 prompted us to investigate the fate of wild-type and mutant α2 proteins by cell transfections. HeLa cells were co-transfected with pA2-R887 or pA2-wt constructs, which express, respectively, mutant and wild type full length c-myc-tagged ATP1A2 cDNAs, each together with pB2 expressing the β2 subunit (as described in [17]). Immunoblot revealed a decreased amount of R887 mutant (Figure 3A), thus confirming the in vivo results on Atp1a2R887/R887 and Atp1a2+/R887 mutant mice. More important, immunofluorescence staining demonstrated a different subcellular localization. Wild type α2 Na,K ATPase showed a typical plasmamembrane and slightly endoplasmic reticulum staining (Figure 3B, upper panels). Differently, most of R887 mutant protein appeared as punctuated pattern localized in the perinuclear region, which overlapped with the endoplasmic reticulum marker calnexin (Figure 3B; a colocalization quantification appears on right panels). Misfolding of the mutant α2 Na,K ATPase induced by the R887 mutation causes very likely the endoplasmic reticulum retention and the inefficient and delayed secretion process. In fact, by inhibiting the proteasome activity with MG132, wild-type and more consistently mutant α2 subunits accumulated in transfected cells (Figure 3C).
Migraine is a complex phenotype that hampers the objective and quantitative evaluation in animal models. In order to assess the effect of the R887 ATP1A2 mutation on an important component of the migraine attack, cortical spreading depression (CSD), we analyzed this neuronal phenomenon in adult Atp1a2+/R887 mice and in their wild type littermates. CSD was induced by electrical stimulation of the visual cortex using a bipolar electrode and recorded at two sites of somatosensory and motor cortex (Figure 4A). Incremental current stimuli were delivered up to CSD induction and the charge delivered at CSD activation was considered as threshold.
Atp1a2+/R887mutants were more susceptible to undergo CSD. Indeed the threshold for induction of CSD in mutant animals was significantly lower than wild type animals (Atp1a2+/R887, 13.00±1.7 µC, n = 20; wild type, 19.9±1.9 µC, n = 18; t-test p<0.01) (Figure 4B, left graph). Moreover, CSD propagation rate was altered in the mutant, which showed higher CSD velocity rising from 3.85±0.35 mm/min (wild type, n = 18) to 5.41±0.41 mm/min (n = 20; t-test p<0.01) in Atp1a2+/R887 mice (Figure 4B, middle graph). No significant difference was observed in CSD duration (Atp1a2+/R887 40.1±3.19 sec, n = 20; wild type 41.1±3.5 sec, n = 18; t-test p = 0.83) (Figure 4B, right graph). After the first CSD, the trace was monitored for further 90 minutes to reveal repetitive CSDs, a parameter correlated with the phenotype severity of FHM models [41]. Heterozygous R887 mutation did not modify the proportion of mice showing repetitive CSD (Atp1a2+/R887 4 out of 20 mice, wild type 4 out of 18 mice). We conclude that the R887 α2 Na,K-ATPase facilitates CSD induction and propagation, but it neither affects its duration nor promotes the induction of repetitive CSDs. Within groups, no difference in CSD threshold and propagation rate was observed between male and female (see Methods).
Genetic mouse models are essential tools to dissect complex pathogenic mechanisms leading to human diseases. Here, we report data on the generation of the first knock-in mouse carrying the W887R ATP1A2 mutation causing FHM2. The R887 allele has been associated to a typical form of FHM with hemiparesis and epileptic episodes [17]. Since the effect of this mutation is an almost complete loss of function [17], [38], we expected a neonatal lethality of homozygous mutant mice. In fact, Atp1a2R887/R887 mutants die few minutes after birth, closely resembling the knock-out models [39], [42], which fail to develop a regular respiratory rhythm [43], [44]. Heterozygous Atp1a2+/R887 mice are viable and fertile. A general behavior characterization by a modified SHIRPA protocol [40] shows a higher susceptibility to fear and anxiety in Atp1a2+/R887 mice. This result resembles the previous reports by Ikeda et al. [42] and Moseley et al [45] showing similar phenotype in the null allele heterozygous Atp1a2+/− mice by more specific tests for anxiety and conditioned fear.
As we demonstrate, the mutant gene is correctly transcribed and translated. However, the mutant R887 protein is ineffectively exported from the endoplasmic reticulum-Golgi system. Mutant R887 isoform is mostly degraded by the proteasomal system as demonstrated by the remarkable accumulation of mutant protein under proteasomal inhibition. This is particularly evident in vivo, where the mutant protein is barely detectable in the homozygous mutant brain. This finding is apparently in contrast with our previous result [17] that showed the mutant R887 subunit localized all over the cytoplasm in COS7 transfected cells and seemingly to plasmamembrane as well. By the recent confocal analysis and employing a transfection system that does not saturate, like the COS7 cells, the cytoplasm of exogenous protein, the mutant protein is shown as mostly endoplasmic reticulum-retained.
It is worth noting that Koenderink and coworkers proposed a plasmamembrane localization of the R887 protein by centrifugal fractionation in Xenopus oocytes, probably due to the different cellular system and conditions (room temperature) and the indirect method of localization [38]. Infact, this test at room temperature may favor the mislocalization of the α2 ATPase mutant protein, as reported in [36]. Considering the autosomal dominant inheritance of FHM, we have addressed our attention to the phenotype analysis of the heterozygous knock-in mouse. CSD represents an excellent phenotype to be analyzed in animal models of migraine as CSD underlies migraine aura in patients [1], [7], [8], [9] and can activate the meningeal trigeminal nociceptors in animals [12]. Atp1a2+/R887 mutant mice, our FHM2 model, are more susceptible to CSD as shown by the decreased threshold of induction and the increased velocity of propagation of CSD induced by electrical stimulation of the cortex in vivo. Duration of CSD in Atp1a2+/R887 mice is unchanged. The facilitation of CSD in our FHM2 mouse model is thus very similar to that previously described in Cacna1a knock-in mice representing the FHM1 models [41], [46]. In fact, both homozygous and heterozygous S218L and homozygous R192Q FHM1 models showed a lower threshold for CSD induction and a higher velocity of CSD propagation, whereas CSD duration was not significantly prolonged. Interestingly, the extent of CSD facilitation correlated with the severity of the clinical phenotype of the two FHM1 mutations in humans [41], [46], [47]. The demonstration that FHM2 and FHM1 mutations share the ability to facilitate induction and propagation of CSD in mouse models further support the role of CSD as a key migraine trigger.
The facilitation of CSD in Atp1a2+/R887 mice could be due to impaired clearance of K+ and/or glutamate by astrocytes during cortical neuronal activity consequent to loss-of-function of the α2 Na,K ATPase pump, as previously suggested [34], [48]. Pharmacological evidence shows that α3 and/or α2 Na,K pumps participate in the clearance of K+ from the extracellular space during intense neuronal activity, although the relative importance of α3 and α2 Na,K pumps is unclear[49], [50]. Most models of CSD include local increase of extracellular [K+] above a critical value as a triggering event in the initiation of CSD, hence predicting that a reduced K+ clearance would result in a lower threshold for CSD induction [51]. Indeed, in hippocampal slices the inhibition of α2 and α3 Na,K pumps by local administration of ouabain (at a concentration which only partially affects the low affinity α1 Na,K pump) reduced the threshold for CSD induction by local pulses of high [K+] [52]. This reduced CSD threshold was accompanied by a large increase in CSD duration (and decrease in post-CSD undershoots of membrane potential and external [K+]), pointing to the involvement of α3 and/or α2 Na,K pump activity in CSD termination. We speculate that our findings of a lower threshold for CSD induction but unaltered CSD duration in Atp1a2+/R887 mice suggest a relatively minor role of the glial α2 Na,K pump in K+ clearance. This is in agreement with the evidence that the α3 isoform contributes most of the Na,K ATPase activity in mouse brain homogenates [53] and, therefore, we suggest that the reduced CSD threshold in FHM2 knockin mice is not primarily due to impaired K+ clearance by astrocytes.
Several lines of evidence indicate a specific role of the α2 Na,K pump in glutamate clearance during synaptic transmission. The α2 Na,K pump is specifically stimulated by glutamate in cultured astrocytes [54]. In the adult somatosensory cortex the α2 Na,K pump shows a specific localization in astrocyte processes surrounding glutamatergic synaptic junctions, which coincides with that of the glial glutamate transporters GLAST and GLT1 [55], [56]. Also, a physical association and functional coupling between the α2 Na,K pump and glutamate transporters has been demonstrated [56].
We therefore hypothesize that CSD facilitation in the FHM2 mouse model is sustained by inefficient glutamate clearance by astrocytes and consequent enhanced cortical excitatory neurotransmission, particularly the NMDA receptor-mediated transmission during high-frequency action potential trains [57]. This glutamatergic hypothesis finds suggestive echoes in the recent report by Anttila et al. [58], where MTDH, a modulator of glutamate transporters has been associated to the common form of migraine with aura. In addition, a mutation of the glial excitatory aminoacid transporter type 1 (EAAT1) leads to neuronal hyperexcitability and subsequent seizures, hemiplegia, and episodic ataxia by impaired glutamate uptake [59]. While this scenario remains to be confirmed in the FHM2 mouse model, FHM1 models displayed an enhanced glutamatergic synaptic transmission due to increased Ca2+ influx through the mutant presynaptic CaV2.1 channels and increased probability of glutamate release at cortical pyramidal cell synapses [60]. A causative link between gain of function of glutamatergic transmission at recurrent cortical pyramidal cell synapses and facilitation of experimental CSD was demonstrated in the FHM1 mouse model [60]. Both FHM1 and FHM2 mice point to a model of CSD initiation, where the activation of NMDA receptors by glutamate released from recurrent cortical pyramidal cell synapses plays a key role in the positive feedback cycle that provokes CSD [4].
Furthermore, the absence of the α2 Na,K pump from the glial processes surrounding GABAergic terminals [55] suggests that FHM2 mutations fail to affect inhibitory neurotransmission, similarly to the FHM1 model, which showed unaltered inhibitory neurotransmission at synapses between fast-spiking interneurons and pyramidal cells [60].
We therefore propose that episodic disruptions of the excitation-inhibition balance and hyperactivity of cortical circuits due to excessive recurrent excitation underlie the vulnerability to “spontaneous” CSD ignition in both the rare forms of FHM1 and FHM2 and, probably, at least a fraction of common migraine cases.
Commercially available rabbit polyclonal antibody directed against α2 Na,K-ATPase isoform (cat. AB9094. Millipore, Billerica, MA, USA); mouse monoclonal antibodies for Na,K-ATPase alpha 1 isoform (α6F; Developmental Studies Hybridoma Bank, Iowa City, IA, USA), for Na,K-ATPase alpha 3 isoform (cat. MA3-915, Affinity Bio Reagents Suite 600 Golden, CO, USA), anti-bovine α-tubulin, mouse monoclonal antibody (cat. A11126, Molecular probes, Inc. 29851 Willow Creek Road, Eugene, OR, USA); GAPDH (6C5) mouse monoclonal antibody (sc-32233, Santa Cruz Biotechnology Inc., California, USA); Ubiquitin (P4D1), mouse monoclonal antibody (sc-8017, Santa Cruz Biotechnology Inc. CA, USA). ECL anti-mouse and anti-rabbit IgG and horseradish peroxidase (HRP)-linked species-specific whole antibodies were purchased from GE Healthcare. Polyclonal rabbit anti-goat IgG/HRP was obtained from Dako (Glostrup, Denmark). For immunofluorescence experiments, the following antibodies were used: monoclonal anti c-Myc (9E10) and rabbit anti- calnexin (Sigma-Aldrich, Milan, Italy). Secondary antibodies were conjugated with Alexa 488 and Alexa 596 (Invitrogen, Carlsbad, CA, USA).
Procedures involving animals and their care were conducted in conformity with guidelines of the Institutional Animal Care and Use Committee at San Raffaele Hospital (Milan, Italy) in compliance with national (D.L. No. 116, G.U. Suppl. 40. Feb 18, 1992, Circolare No. 8 G.U., 14 Lug. 1994) and international (EEC Council Directive 86/609, OJ L 358, 1 DEC.12, 1987; National Institutes of Health Guide for the Care and Use of Laboratory Animals, U.S. National Research Council, 1996) laws and policies. Animals were housed in Specific Pathogen Free (SPF) conditions, maintained on a 12-h light/dark cycle, with free access to food and water. Atp1a2+/R887-neo knock-in mice were generated using homologous recombination in embryonic stem (ES) cells to modify the Atp1a2 gene such that the exon 19 contained the human FHM-2 W887R mutation. In the targeting vector, the original TGG triplet codon (POSITION 2763, CODON 921) was changed into CGG by mutagenesis, creating the W887R mutation. Downstream of exon 19, a PGK-driven neo cassette flanked by LoxP sites was present. ES cells were electroporated, and clones were selected for homologous recombination by Southern blot analysis. The presence of the W887R mutation was tested by PCR using primers 5′-GGCTTCTTTACCTACTTTGTGATA-3′ and 5′-ATGCCCTGCTGGAACACTGAGTTG-3′ and subsequent sequencing analysis of exon 19. Targeted ES cells were injected into C57Bl/6J blastocysts and these transferred into pseudopregnant CD1 females to create chimeric animals. Chimeras were backcrossed with wild-type C57Bl/6J to generate F1 progeny and the agouti offsprings were genotyped for transmission of the mutant allele, generating transgenic line Atp1a2+/R887-neo knock-in mice. Heterozygous Atp1a2+/R887-neo knock-in mice were bred with transgenic mice expressing FLPe recombinase under the control of the human ACTB promoter (TgN(ACTFLPe)9205Dym; The Jackson Laboratory) to remove the neo cassette. Expression of FLPe recombinase as early as embryonic day 10.5 causes the Flippase recognition target (FRT) sites recombination and the removal of the neo cassette. Germ line transmission was obtained and transgenic line Atp1a2+/R887-neo was established. Mice were further bred with C57Bl/6J for seven generations, at this stage the background would nevertheless be >90% congenic. Heterozygous Atp1a2+/R88 and Atp1a2+/+ littermates were used for further analysis.
Sensory-motor function of mutant mice compared with controls was assessed by a modified version of the SHIRPA protocol primary screening [40]. Briefly, undisturbed behavior of each animal was first observed in its own home cage: body position, spontaneous activity and respiration rate were recorded, assigning a score to each behavior. In addition, manifestations of tremors, bizarre behaviors, stereotypes or convulsions were checked at this stage of the protocol. Thereafter mice were transferred individually to a new arena and were tested for transfer arousal, palpebral closing, piloerection, gait, pelvic and tail elevation, touch escape and positional passivity. There followed a sequence of manipulations using tail suspension and a grid across the width of the arena; animals were scored for trunk curl, limb grasping and grip strength. To complete the assessment, the animals were restrained in a supine position to record autonomic behaviors (heart rate, skin color, limb and abdominal tone, lacrimation, salivation) prior to measurement of the righting reflex after flip of the animal. Vocalizations and irritability (during supine restrain) were also recorded. Fear was assessed based on reaction to transfer to a new environment. A score was assigned to each behavioral test as described in Table 1.
Total RNA was extracted from embryonic mice (E19.5) (n = 9, 3 embryos for each genotype) neuronal (brain) tissues by Trizol method (Invitrogen, Carlsbad, CA, USA). RNA was reverse transcribed using random hexamers SuperScript® First-Strand Synthesis System (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions. Atp1a2 cDNA was amplified using forward primer on exon 19 (5′-GGCTTCTTTACCTACTTTGTGATA-3′) and reverse primer on exon 20 (5′-ATGCCCTGCTGGAACACTGAGTTG-3′) with Hot Master Taq DNA polymerase (Eppendorf, Hamburg, Germany) at 94°C for 2 min, 35 cycles at 94°C for 30 s, 58°C for 30 s, 65°C for 30 s, and 65°C for 5 min. This strategy allows amplification of both endogenous wild-type and mutant allele (PCR product: 254 bp). The relative Atp1a2 amount was normalized to the β-actin expression levels (610 bp PCR product). Since the R887 missense mutation introduces a new restriction site for MspI enzyme, the PCR product was subsequently digested with MspI (New England Biolabs, Ipswich, MA, USA) to discriminate the endogenous gene (uncut, band size: 254 bp band) and the mutant (cut, bands size: 178 bp+76 bp). PCR products were run on a 2% agarose gel in TAE buffer.
To prevent proteolysis during the procedure, all steps were carried out on ice, and all buffers contained protease inhibitor cocktail (Roche, Mannheim, Germany) and phenylmethanesulfonyl fluoride (1 mM). Embryonic brains of the various genotypes (n = 12, 4 for each genotype) were processed simultaneously. For the extraction of membrane proteins, whole brain was homogenized with a glass-Teflon homogenizer in Sucrose solution (0.32 M Sucrose, 5 mM Hepes pH 7.4, 2 mM EDTA). After a short centrifugation (5000 rpm, 20′4°C) the supernatant was centrifuged for 1 hr at 42,000 rpm 1 h 4°C (Beckman, ultraTL100, rotor TL100.3) and the pellet resuspended in Sucrose buffer. Protein concentration was measured using the Bio-Rad Protein Assay according to the manufacturer's instructions. The preparation of cells and tissues (total brain, cortex and cerebellum) total lysates were performed adding RIPA buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.5% sodium deoxycholate, 0.1% SDS, 2 mM EDTA, and 1% Igepal CA-630) to the collected samples and left 30′ on ice than the lysates were centrifuged 13000 g 20′4°C. The protein content of the supernatant was measured using the BCA protein assay with bovine serum albumin as standard. We resuspended equal amounts of proteins (15 µg each sample in 20 µl) in SDS-PAGE buffer (100 mM Tris-Glycine pH 6.8, 0.56 M mercaptoethanol, 2% SDS, 15% glycerol, and 0.1% BFB), and separated them for 2 h at 100volt in 8% SDS-polyacrilamide gels. Proteins were electrophoretically transferred to hybond ECL nitrocellulose membranes (GE Healthcare, Munich, Germany) and blots were blocked overnight with 5% non-fat milk 0.1% Tween-20 in PBS. The blocked blots were incubated for 2 h with subunit specific antibodies, washed three times for 10 minutes each with 0.1% Tween-20 in PBS then incubated with the appropriate peroxidase conjugated secondary antibodies. After another series of washes (three times for 10 minutes each) peroxidase was detected using a chemiluminescent substrate (GE Healthcare, Munich, Germany).
Plasmid constructs were the same as in [17]) by metafectene (Biontex, Martinsried/Planegg, Germany) according to the manufacturer's instructions. We selected the ratio range of Metafectene (µl) to plasmids DNA (µg) of 5∶1. 48 h after transfection, we fixed HeLa cells in 4% paraformaldehyde (PFA) for 30 min at RT and blocked and permeabilized with 10% donkey serum 0.2% Triton-X100 in phosphate-buffered saline solution (PBS) for 30 min at RT. Permeabilized cells were then incubated with primary antibodies for 2 hr at RT, than washed (three times) in PBS, incubated with appropriate secondary antibodies and washed three times with PBS solution. We placed cells in fluorescent mounting medium (Dako Cytomation, Glostrup, Denmark) over microscope slides and confocal microscopy was performed on the Perkin Elmer UltraVIEW.
Immunofluorescence colocalization was visualized by confocal microscopy and analyzed by Wright Cell Imaging Facility (WCIF) colocalization plug-in of Image J software (http://www.uhnresearch.ca/facilities/wcif/imagej/colour_analysis.htm). The following parameters were measured: Pearson's correlation coefficient (Rr; 1, perfect correlation, to −1, perfect exclusion); Mander's overlap coefficient (R; 1, highest, to 0, random correlation); Ch1∶Ch2, the red∶green pixel ratio.
Proteasome inhibitor MG132 (carbobenzoxy-L-leucyl-L-leucyl-L-leucinal) was obtained from Sigma-Aldrich, Milan, Italy (cat. C2211). MG132 were dissolved in DMSO and applied to cells at the concentration of 10 µM, after 48 hours of transfection for the time periods indicated in the text and. An equivalent volume of DMSO was added to control cells. Anti ubiquitin antibody was used to reveal the increase of ubiquitinated proteins after proteasome inhibition.
CSD was recorded as described in Van den Maagdenberg, et al. [46]. Briefly, mice (20–30 g) were anaesthetized with urethane (20% in saline; 6 ml/kg i.p.). Animals, mounted on a stereotaxic apparatus were continuously monitored for adequate level of anesthesia, temperature, heart rate and nociceptive reflexes. Blood oxygen saturation and flux as well as heart and breathing rates were monitored non-invasively using an oximeter (Starr, Life Science Corp.). Oxygen was supplied to maintain blood oxygenation above 93% for the entire duration of the experiment. Heart rate was between 400–600 beats/min, and breathing rate approximately 200 breaths/min. Animals not meeting these criteria were excluded from our sample. To record CSD three holes were drilled in the skull over the left hemisphere. The first corresponded to the occipital cortex and was used for access of the electrical stimulation electrode (0 mm A-P, 2 mm M-L from lambda). The second hole, at the parietal cortex (1 mm M-L, 1 mm caudal to bregma) and the third hole, at the frontal cortex (1 mm M-L, 1 mm rostral to bregma), were used for placement of the CSD recording electrodes. The steady (DC) potential was recorded with glass micropipettes filled with NaCl (3 M, tip resistance 1–2 MΩ) inserted 200 µm below the dural surface. An Ag/AgCl reference electrode was placed subcutaneous above the nose. Cortical stimulation was conducted using a copper bipolar electrode (0.2 mm tip diameter, 0.3 mm intertip distance) placed on the cortex surface after removing the dura. Single pulses of increasing intensity (20, 30, 40, 50, 60, 80, 100, 120, 140, 160, 180, 200, 230, 260, 290, 320, 350, 380, 430, 480, 530, 600, 700, 800, 900, 1000 µA) were applied for 100 ms at 3-min intervals by using a stimulus isolator/constant current unit (WPI, USA) until a CSD event was observed [13]. DC cortical potential was amplified (10×) and low-pass filtered at 200 Hz (Cyberamp, Axon Instruments, Union City, CA). Signals were continuously digitized and recorded using Labview data acquisition and analysis system. The minimal stimulus intensity at which a CSD event was elicited was taken as the CSD threshold. In all mice, when CSD was elicited, recordings were continued for 90 min to detect multiple CSDs. To estimate CSD propagation velocity, the distance between the two recording electrodes was divided by the time elapsed between the CSD onsets at the first and second recording sites. The percentage of mice with multiple CSD events was determined only from the mice that could be recorded for one full 90 min following the first detected event. CSD duration was measured at half-maximal amplitude [13]. Because no difference in CSD threshold and propagation rate was observed between male (N = 11 wild type and N = 11 mutants) and female (N = 7 wild type and N = 9 mutants) within each genotype (wild type: threshold male 20.7±2.1 µC, female 18.7±3.5 Mann-Whitney test p = 0.61; propagation rate male 3.9±0.42 mm/min, female 3.8±0.66 mm/min Mann-Whitney test p = 0.86; mutants: threshold male 13.4±3.0 µC, female 12.6±1.3 t-test p = 0.82; propagation rate male 5.2±0.32 mm/min, female 5.6±0.84 mm/min Mann-Whitney test p = 0.94) the results from males and females were pooled.
For SHIRPA protocol primary screening, comparisons were performed with the Mann-Whitney nonparametric test. The Student's t-test with one-tail distribution was used for significance calculation in densitometric analysis.
Statistical analysis for CSD recordings was performed using Sigma Stat 3.1 (Systat Software, Chicago IL USA). Multiple groups were compared by ANOVA followed by post-hoc comparisons applying Bonferroni correction or Holm-Sidak test. When two groups were compared, t-test was applied. Normality and homoschedasticity of the data was checked. Data not normally distributed were compared using nonparametric Kruskal-Wallis ANOVA or Mann-Whitney rank sum test. Significance level was equal to 0.05. Data are reported as average ± SEM.
|
10.1371/journal.pcbi.1000550 | Structure of Protein Interaction Networks and Their Implications on Drug Design | Protein-protein interaction networks (PINs) are rich sources of information that enable the network properties of biological systems to be understood. A study of the topological and statistical properties of budding yeast and human PINs revealed that they are scale-rich and configured as highly optimized tolerance (HOT) networks that are similar to the router-level topology of the Internet. This is different from claims that such networks are scale-free and configured through simple preferential-attachment processes. Further analysis revealed that there are extensive interconnections among middle-degree nodes that form the backbone of the networks. Degree distributions of essential genes, synthetic lethal genes, synthetic sick genes, and human drug-target genes indicate that there are advantageous drug targets among nodes with middle- to low-degree nodes. Such network properties provide the rationale for combinatorial drugs that target less prominent nodes to increase synergetic efficacy and create fewer side effects.
| Genome-wide data on interactions between proteins are now available, and networks of protein interactions are the keys to understanding diseases and finding accurate drug targets. This study revealed that the architectural properties of the backbones of protein interaction networks (PINs) were similar to those of the Internet router-level topology by using statistical analyses of genome-wide budding yeast and human PINs. This type of network is known as a highly optimized tolerance (HOT) network that is robust against failures in its components and that ensures high levels of communication. Moreover, we also found that a large number of the most successful drug-target proteins are on the backbone of the human PIN. We made a list of proteins on the backbone of the human PIN, which may help drug companies to search more efficiently for new drug targets.
| There is a growing awareness that networks of protein interactions and gene regulations are the keys to understanding diseases and finding accurate drug targets [1]. With the increasing availability of genome-wide data including those on protein interactions and gene expressions, numbers of studies have been done on the structure and statistics of protein interactions and how diseased genes and drug targets are distributed over the network [2],[3]. Understanding the topological and statistical properties of interaction networks and their relationships with lethal genes as well as currently identified drug targets should provide us with insights into robust and fragile properties of networks and possible drug targets for the future. We studied budding-yeast and human protein-protein interaction networks (PINs) to identify the architectural properties of network structures.
PINs have often been argued to be “scale-free” [4],[5], which mostly means they have power-law frequency-degree distributions. However, this definition diverges from the original meaning of being scale-free in terms of the self-similarity of geometric properties of subject systems and there have been reports that claim such distributions are “more normal than normal”; thus, they are not considered to be particularly exotic by themselves [6]. In addition, there are different network topologies with different robustness and performance properties that maintain power-law distributions [7]. Therefore, it is very important to identify the architectural features of the network bearing the specific utilization of analysis results in mind. Our goal in this study was to identify the network topology of PINs and their relationship with lethal genes and possible drug targets so that the statistical likelihood of novel drug targets could be inferred.
A particularly interesting issue in the field of systems engineering, physics, and systems biology is the trade-off between the properties of robustness, fragility, and efficiency. Highly optimized tolerance (HOT) theory is a conceptual framework that can be used to explain this issue. Although a system conforming to HOT theory is optimized for specific perturbations and has highly efficient properties, such a system is extremely fragile against unexpected perturbations [8],[9]. Doyle et al. [8] demonstrated that the Abline Internet2 router-level topology network conformed to HOT theory. Nodes in the Abline network with extremely high-degree nodes connect to a large number of low-degree nodes, while links between these high-degree nodes are suppressed and thus they do not form a core backbone for the whole network. A network having similar structures to the Abline network is defined as a HOTnet [8]. It would be very interesting to clarify whether PINs are HOTnets or not.
The two questions addressed in this paper are: (1) what is the global architecture of PINs? Do they follow the possible architectural features of scale-free networks created by preferential attachments or conform to HOT theory, and (2) are there specific statistical features for proteins that are likely to be drug targets? To answer these questions, budding yeast and human PINs were used to analyze their structural properties using a series of analysis methods.
Scale-free Network vs. Highly Optimized Tolerance Network: A series of analyses was carried out using budding yeast and human PIN data to identify the topological features of PINs.
In this study, we defined low-degree nodes as nodes with degrees of less than 5 because Han et al. [10] and Partil and Nakamura [11] defined hubs as nodes with degrees of more than 6. We then developed a method called moving stratification by degrees (MSD) to extract sub-networks consisting of hubs with specific degree distributions where indices such as average cluster coefficients would be computed (see Materials and Methods for details). The analyses revealed that the average cluster coefficient was very high for sub-networks consisting of hubs with degrees from 6 to 38, while it was very low for hubs with degrees of more than 39 in the yeast PIN (see Figure S1 and Table S1). Notably, for hubs with degrees of less than 38, the difference in cluster coefficients was generally significant between the yeast PIN and random network, while there were no significant differences in cluster coefficients for hubs with degrees of more than 39 (see Figure S1). Therefore, we defined middle-degree nodes as those with degrees from 6 to 38 and those with degrees of more than 39 as high. In the same manner, we defined middle- (from 6 to 30) and high-degree (more than 31) nodes in the human PIN (see Figure S2 and Table S2). Note that, when we used more stringent thresholds for middle- (from 10 to 50) and high-degree (more than 51) nodes, the results did not change essentially, i.e., the average cluster coefficient for middle-degree nodes was much higher than that for high-degree nodes (see Tables S3 and S4).
The analyses revealed three findings: (1) the network structure for middle-degree nodes (from 6 to 38 for yeast and from 6 to 30 for human PINs), and high-degree nodes (more than 39 for yeast and more than 31 for human PINs) has different structures, (2) middle-degree nodes are tightly connected and form a structure often called a “stratus”, and (3) high-degree nodes do not connect, but connect with low-degree nodes, and form an “altocumulus” structure (Figures 1 and 2). Notably, we used more stringent thresholds for middle- (degrees from 10 to 50) and high-degree nodes (degrees more than 51), and found that changing the thresholds did not essentially affect the results (see Figure S3 and S4). These results suggests that PINs have an architecture where highly interconnected middle-degree nodes form a core backbone for the whole network and large numbers of low-degree nodes connect to high-degree nodes (see Figure 2). This architecture is a type of network that is suggested as a HOTnet, i.e., a network with HOT properties, also seen in the Internet router-level topology [8]. To further confirm this observation, we calculated a graph-theoretic quantity, s(g), that defines the likelihood high-degree nodes will be connected to one another (see Materials and Methods for details). S(g), a value normalized against smax, indicates that networks with tightly interconnected high-degree nodes tend to be closer to 1.0, whereas networks with only sparsely interconnected high-degree nodes tend to be closer to 0.0 (see Materials and Methods for details). Doyle et al. reported randomly generated preferential-attachment-type scale-free networks had relatively high values such as 0.61, whereas a HOTnet exemplified by a network abstracted from an actual Abilene Internet2 router topology network had a value as low as 0.34 [8]. We found that the value of S(g) for the yeast PIN was 0.25 and that of the human PIN was 0.38. Thus, we could conclude that PINs are HOTnets.
PINs are networks with a modular structure [12]–[14]. Here, modularity is defined as characteristics where there are fewer links between nodes with similar degrees. This only means there are limited links between high-degree nodes (hubs), whereas there are links between hubs and low-degree nodes. This is a feature that was also confirmed in this study (see Figure 2). Modularity in PINs implies that networks have two features [13]: First, functional units may be composed of many low-degree nodes that are directly connected to a hub node. Second, confusion between modules is avoided by avoiding direct connection between hubs. While there are arguments against this claim that hubs are tightly connected because they need to influence one another to achieve an integrated function for the whole system [15], analysis results indicate that such integration is most likely to take place via middle-degree nodes instead of high-degree nodes (see Figure 2).
The distribution of essential genes, synthetic genes, and other genes are shown in Figure 3. It is interesting to note that both essential genes and synthetic lethal genes have similar distributions. The average degree of essential proteins is 4.95 and that of synthetic lethal proteins is 4.40. However, the Wilcoxon rank sum test demonstrated that there is no statistical significance between them (P = 0.334). In either case, essential and synthetic lethal proteins are concentrated on middle-degree nodes and high-degree nodes. However, the average degree among synthetic sick genes is 4.07 and this is significantly lower than that among synthetic lethal genes (P = 0.0015). This means genes that have less severe impact are distributed toward regions with a lower-degree distribution.
Scale-richness: The power law distribution often characterized for scale-free networks only means that local frequency-degree distributions are independent of location along the degree axis, rather than self-similarity of network structures. However, Tanaka demonstrated that bacterial metabolic networks are scale rich in the sense there are different categories of metabolites and enzymes depending on the degree of nodes [16]. A group of nodes with high degree tends to be composed of currency molecules such as ATP and a group of nodes with low degree mostly consists of enzymes involved in specific cellular functions. In this study, we investigated if the frequency-degree distribution of proteins for each functional category exhibited the scale-rich characteristics reported by Tanaka. Figures 4 and S5 correspond to frequency-degree plots for proteins in different functional categories in the yeast PIN and the human PIN. The functional categories were assigned based on the GO slim ontology. As shown in the figures, the degree distribution patterns differ among functional categories. Moreover, proteins with different GO slim annotations have different average degrees (See Tables S5 and S6). Note that many functional categories have significantly higher (or lower) average degrees than the whole PINs (See Tables S5 and S6). These results suggest that the yeast and human PINs are scale-rich.
Drug Targets: Drug-target molecules are distributed over low- to middle-level degree nodes with higher probability on middle-degree nodes. Consistent with reports already published, the average degree among drug-target nodes (4.74) is higher than the average degree among all nodes (4.06).
The distribution of known drug targets is shown in Figure 5 and this is predominantly distributed to middle-degree nodes and mostly on backbone of the network. There are almost no drug targets for high-degree nodes. The distribution of drug targets for cancer and non-cancerous diseases are in sharp contrast. While the average degree of target nodes for cancer drugs was 7.82, the targets for non-cancerous diseases scored only 4.24 (P = 0.01). Moreover, we found that the proportion of drug targets among low-degree proteins were similar to random expectation. Figure 6 shows distribution of drug targets marked on degree-rank plot. The drug target molecule that has highest degree is Src with 41 which is the target for drugs such as Dasatinib. Target molecules for anti-cancer drugs are shifted toward high degree nodes compare against average and non-anti-cancer drugs.
A series of analyses revealed that both the budding yeast and human PINs are scale-rich and have HOT networks. There are extensive interconnections among middle-degree nodes that form the backbone of the network (see Figure 2). Most drug-target genes concentrate on middle-degree nodes and parts of low-degree nodes, but not on high-degree nodes. Interestingly, Feldman et al. (2008) [17] reported that genes harboring inherited disease mutations also concentrated on middle-degree nodes. Because of the potential lethality observed in budding yeast (Figure 3A) and reported high lethality in mouse knockout [2], high-degree nodes are unlikely to be preferred drug targets or genes with disease mutations. Since oncogenes tend to be high-degree nodes, they are less likely to be drug targets, or one has to accept major potential side effects. The fact that the degree distribution of cancer-drug targets is higher than that of non-cancer-drug targets is consistent with the report by Yao and Rzhetsky [18]. Since high-degree nodes are predominantly connected with low-degree nodes (Figures 1, 2, S3, and S4), the elimination of high-degree nodes is likely to affect large numbers of low-degree nodes. This may result in unacceptable side effects since a group of genes that bear certain functions may be made collectively dysfunctional. Detailed case studies are warranted to test and verify this possible interpretation. However, the average degree distribution of synthetic sick genes (4.07) is less than that of essential genes (4.95) and synthetic lethal genes (4.40). This implies that a drug design strategy to generate synergetic effects by targeting less important targets can be a reasonable option because each compound in such drugs can select targets that have less impact on the overall system alone.
We found that middle-level degree nodes are the optimal targets for therapeutic drugs. A similar observation was reported by Yao and Rzhetsky [18], although they measured the mean degree among drug targets. In this study, we investigated the degree distribution of drug targets in greater detail, because we measured a fraction of drug targets to all nodes with degree k as well as mapping drug targets on the network structure. It was clearly identified most of drug targets for drugs that are currently on the market are concentrated on middle degree nodes that are back bone of the network and low-degree nodes that tends to have specific function specific effects. One of novel findings here is that the distribution of drug targets for low-degree nodes is similar to random expectation, indicating that there are a certain number of low-degree drug targets. From these results, we can expect that the most advantageous targets for combinatorial drugs could be among low-degree nodes because these could have less severe impact on the overall system of the human body. This is consistent with the idea of “long-tail drugs”[19].
Are there any relationships between structures in molecular networks (i.e., scale-richness in PINs) and the properties of their underlying genome? Rzhetsky and Gomez [20] proposed a stochastic model describing the evolutionary growth of molecular networks. Their model predicts that, in a molecular network, the shape of the degree distribution will be similar to the shape of the distribution of domains in the genome. Actually, they showed that, in the case of the entire yeast PIN, both the degree distribution and the distribution of the domain followed a power law. Therefore, it might be interesting to see whether, for each functional category, the shape of the degree distribution was similar to that of the domain distribution, when the entire architecture of domains in genomes becomes available.
In this study, we assumed that the PINs represented all functions of genes. However, the PINs are just composed of binary protein-protein binding and proteins have other types of functions, such as catalyzing reactions with non-protein substrates. Therefore, PINs reflect a subset of the entire cellular function. This indicates that, if the complete picture for cellular protein functions could be considered, our conclusions from the PINs may diverge from what we presented here. Moreover, at present, the yeast and human PINs represent incomplete pictures of the actual entire PINs of these organisms. When data on all the actual entire PINs become available, we intend to examine all the actual entire PINs to see whether similar observations to those in this study can be made or not.
It is interesting to note that both PINs and the Internet topology are HOTnets. Many of the observed properties in Internet router topology may be applied to PINs as well. Such properties include robustness against node failure and optimized performance [21]. It has been reported that analysis using several possible router topologies found that a HOTnet configuration was most efficient, providing more maximum overall bandwidth to users than that with other network-configuration approaches such as random and preferential attachment [21]. The implication is that biological PINs have evolved to become efficient and error tolerant. The series of analyses presented in this report indicate that there are changes whereby we can rationally design drugs by taking into account network properties, and additional insights from engineering and physics may further extend our opportunities for exploring network-based biology.
|
10.1371/journal.pgen.1003605 | Genetic and Anatomical Basis of the Barrier Separating Wakefulness and Anesthetic-Induced Unresponsiveness | A robust, bistable switch regulates the fluctuations between wakefulness and natural sleep as well as those between wakefulness and anesthetic-induced unresponsiveness. We previously provided experimental evidence for the existence of a behavioral barrier to transitions between these states of arousal, which we call neural inertia. Here we show that neural inertia is controlled by processes that contribute to sleep homeostasis and requires four genes involved in electrical excitability: Sh, sss, na and unc79. Although loss of function mutations in these genes can increase or decrease sensitivity to anesthesia induction, surprisingly, they all collapse neural inertia. These effects are genetically selective: neural inertia is not perturbed by loss-of-function mutations in all genes required for the sleep/wake cycle. These effects are also anatomically selective: sss acts in different neurons to influence arousal-promoting and arousal-suppressing processes underlying neural inertia. Supporting the idea that anesthesia and sleep share some, but not all, genetic and anatomical arousal-regulating pathways, we demonstrate that increasing homeostatic sleep drive widens the neural inertial barrier. We propose that processes selectively contributing to sleep homeostasis and neural inertia may be impaired in pathophysiological conditions such as coma and persistent vegetative states.
| An annual 234 million surgical procedures are performed worldwide, making general anesthetics among the most common drugs administered to humans. Remarkably, however, we still do not understand the mechanisms by which general anesthetics render patients unconscious or the processes that re-establish consciousness upon emergence from anesthesia. We previously showed that the brain resists transitions between the wakeful and anesthesia states by generating a barrier to such transitions in both directions. We also showed that the existence of this barrier is conserved from invertebrates to mammals. In our present work, we use the genetic tractability and the simplified nervous system of the fruit fly Drosophila melanogaster to show that four genes are required to maintain this barrier. We also show that, as in mammals, there is overlap between pathways regulating natural sleep and general anesthesia. We propose that some of these shared pathways are impaired in conditions such as coma and persistent vegetative states, in which the barrier to transitioning to the waking state appears to be insurmountable.
| Inherent in the design of robust and bistable switches is hysteresis, which prevents small or random fluctuations from triggering a state change in the system [1]. Arousal states display bistable behavior and are regulated by a biologic switch that possesses hysteretic properties [2]–[5]. Inhaled general anesthetics offer the opportunity to study the molecular and neuroanatomical pathways essential for the aroused, conscious state as well as the orderly transition to and from the unconscious state [6], [7]. General anesthetics are known to exert their hypnotic properties in part by interacting with endogenous systems that regulate arousal state [8]–[10]. Functionally these interactions include modulation of ion channels to suppress neuronal excitability [11]. Behaviorally the effects of these interactions are described by various endpoints that correspond to different depths of general anesthesia including (in order) amnesia, hypnosis, and ultimately immobility [12]. Although historically most studies of anesthetics have been performed on mammals, similar endpoints have been described for invertebrates. Furthermore, in vertebrates and invertebrates similar concentrations of anesthetics induce those endpoints [13]. Phylogenetically and functionally related classes of genes also alter anesthetic sensitivity across multiple phyla [7], [14]–[16]. Collectively these data suggest that mechanisms of arousal control have been conserved throughout evolution, even if gross brain anatomy has diverged.
We previously established in both mice and fruit flies that different concentrations of anesthetics are required for induction of and emergence from general anesthesia, and that this hysteresis cannot be explained solely by pharmacokinetics [7]. Hysteretic dissociation of anesthetic induction from emergence is consistent with the existence of a barrier termed “neural inertia” that separates and stabilizes behavioral states. The inertial barrier leads to maintenance of wakefulness or anesthesia, and presumably exists to oppose rapid and potentially catastrophic transitions between these states. The effective size of the inertial barrier can be estimated by measuring the area between the induction and emergence curves. Switching between wakeful and anesthetized states would thus be difficult with high neural inertia but would occur easily with low neural inertia. Here we sought insight into the mechanisms underlying this behavioral state barrier by studying its genetic and anatomical bases as well as its relation to other arousal-regulating processes such as circadian clock function and sleep.
Previous studies have demonstrated that the concentration-response curve for induction of anesthesia can be manipulated genetically, particularly by mutations that alter excitability [7], [17]. In the present study we demonstrate that the inertial barrier can be collapsed by loss-of-function mutations in genes that have opposing effects on induction of isoflurane anesthesia. These genes encode the hyperpolarizing Shaker potassium channel (Sh) and its positive modulator SLEEPLESS (SSS), the loss of which causes resistance to anesthesia induction, as well as the depolarizing cation channel, narrow abdomen (NA) and its positive modulator UNC79, the loss of which increases sensitivity to anesthesia induction. The requirement of all four genes for maintenance of neural inertia by isoflurane is consistent with a model in which these genes contribute to mutual inhibition by arousal-promoting and arousal-suppressing loci to create a bistable system in which either the waking or anesthetized state predominates, similar to the “flip-flop” switch that has been proposed to stabilize waking and sleep in mammals [2]. Indeed, we find that the sss gene acts in different sets of neurons to influence induction of and emergence from anesthesia. We also find that arousal per se does not control neural inertia since the inertial barrier is unaffected by certain hyperaroused mutants. Instead, as in previous studies with other anesthetics [18]–[20] we report that emergence from anesthesia becomes more difficult in sleep-deprived animals. Consequently, the neural inertial barrier to reversing the anesthetized state is broadened with sleep deprivation. Collectively our data suggest that some molecular and anatomical arousal pathways that underlie sleep homeostasis also contribute to neural inertia.
We undertook the present study to determine whether distinct mechanisms control induction of and emergence from anesthesia. To establish baseline levels of hysteresis for wildtype animals we first established dose-response curves for induction and emergence using isoflurane. As in mammals [7] the two curves are distinct in flies (Figure 1a), suggesting that induction and emergence are not caused by identical processes operating in reverse. However, unlike mammals some flies do not resume movement during the stepwise, downward anesthetic titration. These animals are not dead, but rather exhibit a slower pattern of emergence not amenable to plotting on this time scale (Figure 1b, c, f). The failure of a Drosophila population to fully emerge when anesthetic levels are reduced below the limit of detection is a property subject to genetic regulation and consequently contributes to the measurement of neural inertia [7].
Next we examined induction and emergence curves for animals bearing lesions in genes that have previously been implicated in anesthetic sensitivity. In agreement with published studies [16], [21], [22] we found that disruption of na dramatically increased sensitivity to induction of the anesthesia state by isoflurane, as did disruption of unc79, a gene that is believed to act in the same pathway (Figure 1b). Since wildtype NA is thought to underlie a leak sodium current that promotes excitability [23], we asked whether the correlation between change in excitability and anesthesia induction would apply to other genes that regulate excitability. We began by examining the contribution of Shaker (Sh) potassium channels, which decrease excitability, and confirmed our recent finding that a loss of function mutation in Sh decreases sensitivity to induction (Figure 1c).
The phenotypes of animals bearing mutations in na/unc79 and Sh suggest that excitability is positively correlated with resistance to induction of isoflurane anesthesia. The Sh mutation increases excitability and also increases resistance to induction of anesthesia by isoflurane. We hypothesized that a similar positive correlation would exist between excitability and ease of emergence from isoflurane anesthesia. Indeed, Sh mutants readily emerged from anesthesia. In fact, in these flies emergence is impacted much more than induction and occurs at relatively high concentrations of isoflurane, thereby leading to a collapse of neural inertia (Figures 1c,e). The same reduction in neural inertia can be observed for animals with disrupted expression of the sleepless (sss) gene, which positively regulates Sh K channels [24], [25]. Like Sh mutants, sss mutants show resistance to anesthesia induction (Figure 1f). And as with Sh mutants, the emergence curve for strong sss mutants is compressed against the induction curve, leading to a collapse of neural inertia (Figures 1e,f). The ability of sss mutants to reduce the neural inertial barrier is correlated with the strength of the underlying mutation. sssP1 mutants, with no detectable SSS protein, have a more extreme phenotype than hypomorphic sssP2 mutants in which SSS expression is reduced by ∼30% (Figure 1e, Figure S1a and [25]).
However, a surprising result arises from analysis of na/unc79 mutants. Although these mutants have decreased excitability and therefore would be predicted to resist emergence from anesthesia, they exit the anesthetized state at doses of isoflurane similar to or greater than those required for induction. Thus, na/unc79 mutations reduce the barrier to changing behavioral states in both directions (Figures 1b,d). That is, they promote transitions from the aroused to the anesthetized state and also from anesthesia back to the aroused state. Consistent with this observation, na mutants have highly fragmented bouts of waking and sleep (Figure S2a).
sss is known to regulate Shaker K channels [24], [25], so we combined sss and Sh mutants to determine if the two genes act in the same pathway to affect neural inertia. Consistent with this interpretation, the EC50 for induction in Sh;sss double mutants was similar to or only slightly higher than that in Sh or sss single mutants (Figure S1b–d; Table S1). We also found that Sh loss of function heterozygotes have reduced neural inertia, whereas sssP1 heterozygotes do not, indicating that anesthetic sensitivity is more responsive to reductions in Sh than in sss (Figure S1e).
Having determined that anesthesia induction and emergence are controlled by different genes, we next asked whether different types of anesthetics act on the same or different arousal-regulating pathways. To address this question, we measured dose-response curves for induction and emergence in the presence of halothane, another common volatile anesthetic, using both wildtype and sssP1 mutants. As with isoflurane, halothane exposure revealed a neural inertial barrier between the awake and anesthetized states in control animals. In contrast to what was observed with isoflurane, however, the halothane induction curve was unaffected and the emergence curve was slightly left-shifted in sssP1 mutants, leading to expanded neural inertia (Figure 1g). The failure of isoflurane and halothane to elicit qualitatively similar shifts in induction and emergence in sss mutants is consistent with published reports suggesting different anesthetics act on different molecular or neuroanatomical pathways [26], [27].
The neural pathways underlying the actions of volatile anesthetics are not well understood in mammals, and in invertebrates even less is known. Progress has been stymied in part by an inability to identify and study the roles of the different circuits that control arousal, each of which may be affected to different degrees by a given anesthetic. Our ability to collapse neural inertia with mutations that have opposing effects on isoflurane induction suggests that induction can be genetically dissociated from processes that stabilize the anesthetized state and prevent emergence from it (Figures 1b–g). Genetic dissociation of neural inertia and anesthesia induction raises the possibility that these phenomena may also be anatomically separable.
Because sleep phenotypes of sss mutants are effectively rescued by localized expression of a sss transgene, we used this approach to determine if the induction and neural inertia phenotypes of sssP1 mutants arise from distinct anatomic loci. We coupled various promoters driving the GAL4 transcription factor to a transgene encoding wildtype sss in a homozygous sssP1 mutant background, then determined correlations between expression patterns and rescue of the two sssP1 phenotypes: (a) right-shifting of induction and (b) a more dramatic right-shifting of emergence with consequent collapse of neural inertia. As expected, the native sss promoter rescued these phenotypes robustly (Figure 2a,b). SSS expression is high in the head and particularly in the brain compared to the body [25], so we asked whether sss expression in the nervous system is sufficient to regulate transitions between the anesthesia and waking states. Importantly, the pan-neuronal driver elav-GAL4 rescued induction, emergence, and neural inertia whereas the glial driver repo-GAL4 had no effect on these phenotypes (Figures S3a–c). These results are consistent with the idea that a barrier between the waking and anesthetized states is generated by neurons in the brain.
Another driver, vglut-GAL4, which expresses in glutamatergic neurons, phenocopied the rescue of the induction phenotype observed with sss-GAL4 in a sssP1 mutant background (Figure 2c; Table S1). Restoring wildtype SSS protein to glutamatergic neurons also significantly altered the EC50 for emergence (Table S1), shifting the emergence dose-response curve roughly 20%, in parallel with the induction rescue. However, unlike the sss promoter, the vglut promoter could not rescue the collapse of neural inertia in sssP1 mutants (Figure 2d). Importantly, this result illustrates that glutamatergic expression of sss is insufficient to restore the barrier between the waking and anesthetized states. Unlike vglut, another promoter, D42, failed to rescue the induction phenotype of sssP1 mutants. However, restoration of sss expression in D42-expressing neurons of sssP1 mutants rescued the concentration-response curve for emergence, leading to wildtype levels of neural inertia (Figures 2e,f). Together, the results of rescuing the sssP1 anesthesia phenotypes with vglut-GAL4 and D42 suggest that different sets of neurons are involved in entry into, as well as exit from and stabilization of, the anesthetized state.
Promoters with broad expression patterns such as cha and C309 rescued both induction as well as emergence to varying degrees. For emergence, significant partial or full rescue was observed with cha-GAL4, MZ1366, Mai301, Sep54, 30y and C309. However, neural inertia was only rescued by a subset of these promoters, namely Mai301, Sep54 and 30y. Importantly, induction was not rescued by any of these drivers. Moreover, the majority of drivers failed to alter any phenotype (Figures S3a–c). These data suggest that large but divergent populations of neurons separately control induction and emergence and consequently the stability of the anesthesia state, although we cannot exclude the possibility that small subsets of cells labeled by the positive drivers are responsible for the rescue.
Anesthesia and sleep may both involve suppression of arousal [9], [10], an idea that is supported by the effects of mutations in Sh and sss on these behavioral states [7], [24], [25], [28]. We next addressed whether anesthesia and sleep are regulated by similar biological processes. Sleep drive has been modeled as the combined output of the circadian clock and a homeostatic process of unknown composition [29]. To test whether the same processes modulate the arousal circuitry affected by isoflurane we first attempted to measure concentration-response relationships at different times of day. Measuring the transition from the awake to the anesthetized state in our assay requires that animals be active prior to exposure to drug. This waking activity could not be achieved during long time periods including ZT3-9 and ZT14-22 since at these times animals have a high probability of being immobile due to their natural propensity to sleep. Thus, we addressed circadian regulation by assaying effects of circadian clock mutants. We restricted all measurements described herein to ∼2 hrs starting just after ZT10, near one of the two daily peak activity times. During this period we addressed the circadian contribution to anesthetic sensitivity using a mutant in which the output signal from the clock is abolished, pdf01, and two core clock mutants, cyc01 and Clkjrk. We found that the induction and emergence profiles, and hence neural inertia, were unaffected in all three mutants (Figures 3a–c; Figure S4a), indicating that the circadian clock is not required for isoflurane-dependent anesthesia.
In addition to abolishing circadian clock cycling, cyc01 and Clkjrk mutations cause reductions in sleep [30], [31], much like Sh and sss loss of function mutations [25], [28]. Sh and sss mutants, however, display both sleep and isoflurane anesthesia phenotypes, whereas cyc and Clk mutants do not exhibit the latter. We wondered how common it is to find mutations like cyc01 and Clkjrk that lead to dissociation of the anesthesia and sleep phenotypes. It has been suggested that general anesthetics co-opt arousal pathways that have evolved to regulate the sleep/wake cycle [9], [10]. We thus hypothesized that anesthesia involves an overlapping set, or even a subset, of arousal pathways normally utilized to regulate sleep. If this were the case then non-circadian mutants might also be identifiable that reduce sleep without affecting the anesthetized state. To test this hypothesis, we examined the effects of DATfmn mutants, which have impaired dopamine transporter function, on the concentration-response relationships of induction of and emergence from isoflurane-dependent anesthesia. Like cyc01 and Clkjrk mutants, DATfmn mutants show normal anesthetic sensitivity but abnormally low sleep (Figures 3c,d; Figure S4b, and [30]–[32]). Thus, not all arousal pathways are shared between sleep and anesthesia.
cyc01, Clkjrk, Datfmn Shmns and sssP1 reduce daily sleep, and we show here that a mutation in na causes an increase in sleep as well as fragmentation of sleep and wake bouts (Figure S2a,b). Thus, all these mutations alter levels of daily sleep, but only sss mutants are known to reduce sleep homeostasis, the process that promotes sleep in response to prolonged wakefulness. To address directly whether the homeostatic component of sleep contributes to the response to anesthesia, we tested whether sleep deprivation could alter sensitivity to isoflurane. In wildtype animals, 6–24 hrs of sleep deprivation elicits robust homeostatic recovery sleep [25], [33], a reflection of increased sleep drive and depressed arousal. We exposed experimental animals to mild mechanical agitation for 24 hrs, up to and including times at which animals were treated with isoflurane. Control animals were similarly agitated only during isoflurane treatment and for 15 minutes beforehand. We have previously observed that such agitation is sufficient to awaken sleeping flies but not those that are anesthetized. Consistent with the hypothesis that the anesthesia state may use pathways underlying sleep homeostasis, we found that increasing homeostatic sleep drive led to a small but significant shift in the EC50 for emergence. Although no change was observable in the EC50 for induction of the anesthesia state relative to controls, the net effect was a significant increase in neural inertia for sleep-deprived animals (Figures 3e,f; Table S1).
We previously demonstrated an evolutionarily conserved property of the brain, resistance to changes in arousal state, which we have termed neural inertia [7]. One hallmark of this observed phenomenon, hysteresis of anesthetic action, has been described in mathematical simulations of cortical activity in response to anesthetics as well [5], [34]. In these models and in various biological systems, bistability and ultimately feedback are required for hysteresis. By bistability we mean that a system can exist in either of two stable states. In our case these are the anesthetized and waking states. Other examples of bistability abound in nature, such as metabolic adaptations [1], [35], [36] and cell fate decisions [1], [37]. In these situations, changes in concentration of a biochemical signal lead to positive or negative feedback, resulting in a subsequent change in sensitivity to the initial signal. Consequently, exit from the particular state must proceed along a different concentration-response curve than led to entry into the state.
Another way to think about bistability is in terms of state diagrams. In the simplest example, an inducer (a drug in our case) provides the binding energy to initiate the transition from the awake state to a state of anesthesia. Once the transition is complete and the state change has occurred, a feedback mechanism is initiated that increases the sensitivity of the system to the drug, thus requiring an even greater opposing shift in concentration of drug to reverse the process. Feedback can come at the single cell level, as we have outlined above, but it can also derive from recruitment of other cell types into a unified circuit. A relevant example of this phenomenon can be found in the mutual excitation of thalamic and cortical neurons required for waking. Excitation of thalamic nuclei by arousal systems leads to a switch from the burst firing state characteristic of sleeping or anesthesia to the tonic firing state characteristic of waking [38], [39]. The result is recruitment of cortical neurons into a positive feedback loop that maintains excitation of both sets of neurons, thus stabilizing the waking state.
It has been hypothesized that anesthetics recruit sleep circuitry, perhaps by suppressing arousal systems [9], [10]. But what is the nature of this circuitry? One possibility is that anesthetics could act on a bidirectional neuronal pathway that regulates both induction and emergence. In this scenario, initial anesthetic exposure would alter activity in the pathway such that upon emergence, the population would behave differently and thus produce hysteresis. Alternatively, anesthetics could affect two separate (or partially non-overlapping) pathways: one whose function is disrupted to permit induction and a second whose function must recover to permit emergence. We cannot say for certain where general anesthetics such as isoflurane or halothane act in the fly brain. However, we find that different drivers can separately rescue the shifts in induction and emergence caused by the sssP1 mutation. Thus, our results support a role for distinct anatomical circuits in control of bistability of the waking and unconscious states. Notably, neural inertia is distinct from sensitivity to induction of the anesthesia state since we can collapse hysteresis both with mutations that profoundly inhibit and those that facilitate induction of anesthesia. Most strikingly, na/unc79 mutations facilitate induction of anesthesia, which might be predicted based upon their decreased neural excitability. But they also promote emergence from anesthesia, indicating that they more generally destabilize behavioral states. na mutants also show frequent transitions between sleep and waking (i.e. fragmentation of sleep and wake bouts) and provide perhaps the best genetic evidence for the existence of molecules that stabilize behavioral states.
Collectively our findings suggest the existence of certain features of a minimal neural circuit that underlies neural inertia. First, components must exist to stabilize the waking vs the anesthesia state. This requirement is illustrated in the following example. In the absence of bistability, a simple kinetic model describes the transitions between two states, one unbound and the other bound to drug (Figure 4a). The resulting dose-response curves for the forward and reverse reactions are coincident (Figure 4b). In a bistable situation such as waking and anesthesia, we propose that upon entry into either state, distinct feedback mechanisms are activated to shift drug sensitivity toward stabilization of the state (Figure 4c). As a result the dose-response curves for induction and emergence show hysteresis (Figure 4d). At a circuit level, feedback could take the form of mutual inhibition or positive reinforcement by neurons that facilitate each state (Figure 4e).
Next, we can assign additional components based on measured effects of mutations on induction and emergence. Since loss of excitatory NA facilitates both entry into anesthesia (induction) and exit from this state (emergence), we suggest that na/unc79 is expressed in both arousal-promoting and arousal-inhibiting cells (Figure 4e). If Sh/sss were expressed in the same neurons, mutations in these genes should have opposing effects to those in na/unc79. However, while mutations in Sh/sss retard entry into anesthesia, they do not retard exit from this state. Thus, we place Sh/sss in arousal-promoting but not arousal-inhibiting cells (Figure 4e).
Lastly, there appear to be at least 2 subpopulations of neurons that have distinct effects on induction and emergence when sss is present. Thus, we divide the arousal-promoting portion of our circuit into two parts that reinforce each other's activity as well as suppress the arousal-inhibiting side of the circuit (Figure 4e).
Now we can assess how well our simple 3-cell model explains our data (Figure 4e). During isoflurane anesthesia, activity in the wake-suppressing side of the circuit (blue, A) dominates. Once activated, A cells impede emergence by inhibiting the wake-promoting system (red, W). As a result, exiting the anesthetized state requires that anesthetic be lowered substantially below the level required to enter this state. This effect is responsible for the leftward shift of the emergence curve relative to the induction curve (contrast Figure 4b with Figure 4d).
During waking the situation reverses. Activity within W cells dominates and is stabilized by mutual reinforcing connections (red vertical arrows). This positive feedback increases the amount of anesthetic required to overcome the waking state and induce anesthesia. This effect leads to a rightward shift of the induction curve relative to the emergence curve in Figures 4b,d. Additional stability in the waking state is provided by inhibition of the A cells.
This model also explains the effects of our mutants. We propose that loss of na in cell 1 leads to reduced activity in the W circuit, thus left-shifting the induction curve. We also propose that loss of na in cell 3 leads to reduced activity in the A circuit, thus right-shifting the emergence curve. The net effect is collapse of hysteresis. For sss mutants we propose that activity is increased in cells 1–2 of the W circuit, which results in two changes. The first is a right-shift of the induction curve. The second is inhibition of the A circuit even during anesthesia, which destabilizes this state and right-shifts the emergence curve. Again, the net effect is collapse of hysteresis.
Our model also explains how restoration of sss expression in distinct cells can rescue the induction, emergence and neural inertia phenotypes of sss mutants. We propose that sss in cell 1 reduces suppression of the A side of the circuit during waking, thus restoring the position of the right-shifted induction curve. In contrast, sss in cell 2 reduces suppression of the A side of the circuit during anesthesia, thus restoring the position of the right-shifted emergence curve.
We have also addressed a long-standing hypothesis about the means by which anesthetics are thought to modulate arousal - that is, by co-opting existing sleep-regulatory mechanisms [9], [10]. We have demonstrated that of 8 genes we tested that have been reported to contribute to control of baseline (daily) sleep in flies, only a subset affect induction and stability of isoflurane-dependent anesthesia. Among the genes that have no effect are 3 that are essential to timekeeping by the central circadian clock, suggesting that circadian control of arousal is not required for normal isoflurane sensitivity. Similarly, reduced dopamine transporter function does not affect induction of or emergence from isoflurane-dependent anesthesia, despite leading to a profound reduction in sleep.
If these distinct arousal pathways do not contribute to circuits underlying anesthesia, then which ones do? A recent study suggests that dopaminergic inputs to the fan-shaped body contribute to sensitivity to isoflurane anesthesia, but this study did not distinguish between effects on induction and emergence [40]. Notably we find that D42-driven expression of sss, which rescues altered emergence and neural inertia but not induction in sssP1 mutants, does not appear to express in the fan-shaped body [24], so it is likely that other neurons contribute to the circuitry underlying isoflurane anesthesia as well. D42 is a promoter that is known to express in mixed populations of central neurons as well as some neurons of the peripheral nervous system [24]. D42 was derived from an enhancer trap screen, rather than a cloned gene regulatory element, and the site of insertion of its Gal4-containing P-element is unknown. Thus, the fly gene that it is associated with and any corresponding mammalian gene, including the neurons that express the latter, are also unknown. Due its broad expression pattern, it is difficult to say which neurons are mediating the effects of the D42 driver. However, one possibility is the mushroom bodies, where D42 is known to express [24] and which we have previously shown to participate in sleep regulation [41].
Like our own work, several studies also indicate that mechanisms underlying sleep homeostasis may contribute to the anesthetized state [18]–[20] (though unlike ours, these studies suggest that sleep deprivation impacts both induction and emergence). Consistent with this hypothesis, we find that elevated homeostatic pressure to sleep suppresses arousal and increases neural inertia. This hypothesis is also supported by our finding that sssP1 mutants, which show reduced sleep homeostasis, exhibit reduced neural inertia. This effect is likely to be confined to specific brain circuitry since the promoters that rescue collapsed neural inertia represent a subset of the promoters that rescue sleep loss in sss mutants [24]. However, our hypothesis does not explain why mutants such as cyc01, Clkjrk and DATfmn have normal neural inertia. These mutants sleep substantially less than controls [30], [31], [32] and thus might be expected to have accumulated homeostatic drive to sleep. We hypothesize that these two effects - reduced sleep and increased sleep drive - counteract each other in terms of neural circuit activity, thus leading to no net effect on isoflurane sensitivity. In contrast, in the absence of intact sleep homeostatic mechanisms, such as we find in sss mutants [25], the resulting imbalance in neural circuit activity unmasks changes to the induction and emergence processes. To extend this hypothesis further, mutations that alter induction, emergence or neural inertia may lead to the identification of genes that contribute to sleep homeostasis.
Interestingly, the relationship between sleep homeostasis and neural inertia cannot necessarily be generalized to all anesthetics. Indeed, our data show that although isoflurane-dependent neural inertia is collapsed in sss mutants, neural inertia resulting from halothane-induced anesthesia is not. Taken alongside our rescue of anesthesia induction and neural inertia in sss mutants using different promoters, these data strongly suggest that different anesthetics utilize different arousal pathways to render animals unresponsive. That is, whereas anesthesia has often been treated as a whole-brain phenomenon, our data support actions for different anesthetics in specific circuits that govern arousal. Interestingly, of the mutations that been shown to affect general anesthesia, those with the biggest impact in flies (our data) and mammals [42] cause impairment of ion channel function. Whether these effects are due to loss of drug binding sites in the proteins affected by these mutations, or whether the resulting changes in membrane potential alter anesthetic efficacy [43] remains to be determined. Pharmacokinetics do not appear to be a factor, however, since at the EC50 for emergence in both flies and mammals, isoflurane concentrations are similar in controls and mutants that have altered neural inertia [7]. In any case, specific molecular and neuroanatomical changes clearly alter the state of anesthesia, thus supporting the idea that general anesthetics act on selective targets [11].
In summary, we have provided further evidence that neural inertia represents a barrier to changes in arousal state. We have also shown that this barrier can be genetically and anatomically dissected, and that it is distinguishable from the processes that control induction of anesthesia, at least when this state is studied with isoflurane. While these conclusions are based on studies of Drosophila, it is worth noting that we previously demonstrated genetic control over neural inertia in mammals as well, including mice deficient in noradrenaline production [7]. The commonality of neural inertia in such disparate organisms argues for conserved basic circuit design underlying control of arousal throughout evolution. It should be noted that although we have emphasized the possibility that circuit-based feedback mechanisms underlie bistability in our system, it is also possible that post-translational modifications contribute to this property.
In either case, the clinical importance of our findings is particularly notable for two reasons. First, our results confirm that the sensitivity to induction of anesthesia cannot be used to reliably predict how easily a patient will exit from the anesthesia state. Second, feedback and bistability may be impaired in coma or persistent vegetative states such that the neural inertial barrier separating waking from unconscious states is widened beyond the range of reversibility by normal physiological processes. The conservation of mechanisms underlying waking and anesthesia among distantly related phyla suggest that extension of our current work in Drosophila will continue to shed light on the genetic and anatomical processes underlying behavioral state stability, an issue of fundamental importance to both neuroscience and clinical medicine.
All mutant and transgenic flies were outcrossed 4–7 times into an isogenic w1118 (iso31) background. Unless otherwise stated, controls for mutant animals were outcrossed siblings. GAL4 lines were generated or obtained as previously described [24], except for Gr21a and nos, which were obtained from the Bloomington Stock Center (Bloomington, IN). The Shmns and ShDf lines were obtained from D. Bushey, C. Cirelli and B. Ganetzky (University of Wisconsin), and DATfmn flies were obtained from K. Kume (Kumamoto University). nae04385 and unc79f03453 were obtained from Bloomington, and unc79c04794 was obtained from Exelixis (Harvard). sssP1, sssP2, and UAS-sss were described previously [24], [25].
3–4 male and 5–8 female flies were combined on standard molasses-yeast-cornmeal food and allowed to mate at 21–23°C for 7–10 days. Adults were then discarded, and newly eclosing flies were collected over a 4 day period. 1–5 day-old females were loaded into 65×5 mm cylindrical tubes containing 5% sucrose and 2% agarose and entrained to a 12-hr∶12-hr light∶dark cycle for at least 2 d before being assayed for anesthetic sensitivity or sleep at 25°C using the Drosophila Activity Monitoring System (Trikinetics, Waltham, MA).
Anesthetics dissolved in air were delivered to flies in parallel, and final concentrations and flow rates were measured as previously described [7]. With flow rates set at 15 ml/min/tube, we calculate that gas concentrations inside our .75 ml tubes will reach equilibrium within 18 seconds. For anesthesia measurements, individual flies were exposed to increasing and then decreasing dosages of isoflurane using a previously described protocol [7]. The anesthetic endpoint that was used was immobility, with induction being defined as the lowest concentration at which movement ceased for five or more minutes, whereas emergence was defined as the highest concentration at which movement resumed.
Locomotor counts over 5 min periods for each individual fly were converted to a value of 1, signifying activity, or 0, indicating no movement. Flies that did not move for 15 minutes prior to the start of anesthesia or during the first 5 minutes at the lowest anesthetic dose were excluded from subsequent analysis. Flies that did not recover activity during the 24 hours following anesthesia were also excluded from analysis (<2% for the genetic background for all our experiments, w1118 iso31). Behavior was analyzed using custom software written in MATLAB (MathWorks, Natick, MA) where sleep was identified as periods of inactivity lasting at least 5 min [44]. Concentration-response curves were fit to the Hill equation using Prism 4 (GraphPad, La Jolla, CA), in which the top constant, degree of cooperativity (Hill coefficient) and EC50 were allowed to vary and only the bottom constant was constrained to zero.
Anesthetic experiments were conducted during the evening locomotor activity peak (ZT10:20 to ZT12:40). During this period, flies show consolidated activity and wakefulness. Responses to anesthetics are thus unlikely to be confounded by inactivity due to normal sleep. To calculate neural inertia, the area between the induction and emergence concentration-response curves was integrated over the range of the induction curve's EC1 to the emergence curve's EC99, as previously described [7]. Neural inertia for each set of induction and emergence curves is expressed as the mean ± standard error.
To elicit sleep homeostasis, mechanical stimulation was applied to iso31 animals for 1 second every min for 24 hrs, ending at the last dose of applied isoflurane, using DAMS monitors mounted to a platform vortexer. Control iso31 animals received identical mechanical stimulation throughout dosing of anesthetic, but were not sleep-deprived prior to this time. Specifically, controls were placed on a vortexer with experimental animals beginning 15 minutes before the first dose of isoflurane and mechanically perturbed for 1 second every minute until the final dose of isoflurane at ZT12:40. Pilot studies were used to find the appropriate strength of mechanical stimulation to awaken sleeping but not anesthetized flies.
Differences in neural inertia and sleep, as well as log(EC50)s for induction and emergence, were analyzed with one-way ANOVAs followed by Bonferroni correction for multiple comparisons or Student's t-tests (unpaired, two-tailed) where applicable.
|
10.1371/journal.pbio.1001792 | The Bacterial Effector HopX1 Targets JAZ Transcriptional Repressors to Activate Jasmonate Signaling and Promote Infection in Arabidopsis | Pathogenicity of Pseudomonas syringae is dependent on a type III secretion system, which secretes a suite of virulence effector proteins into the host cytoplasm, and the production of a number of toxins such as coronatine (COR), which is a mimic of the plant hormone jasmonate-isoleuce (JA-Ile). Inside the plant cell, effectors target host molecules to subvert the host cell physiology and disrupt defenses. However, despite the fact that elucidating effector action is essential to understanding bacterial pathogenesis, the molecular function and host targets of the vast majority of effectors remain largely unknown. Here, we found that effector HopX1 from Pseudomonas syringae pv. tabaci (Pta) 11528, a strain that does not produce COR, interacts with and promotes the degradation of JAZ proteins, a key family of JA-repressors. We show that hopX1 encodes a cysteine protease, activity that is required for degradation of JAZs by HopX1. HopX1 associates with JAZ proteins through its central ZIM domain and degradation occurs in a COI1-independent manner. Moreover, ectopic expression of HopX1 in Arabidopsis induces the expression of JA-dependent genes, represses salicylic acid (SA)-induced markers, and complements the growth of a COR-deficient P. syringae pv. tomato (Pto) DC3000 strain during natural bacterial infections. Furthermore, HopX1 promoted susceptibility when delivered by the natural type III secretion system, to a similar extent as the addition of COR, and this effect was dependent on its catalytic activity. Altogether, our results indicate that JAZ proteins are direct targets of bacterial effectors to promote activation of JA-induced defenses and susceptibility in Arabidopsis. HopX1 illustrates a paradigm of an alternative evolutionary solution to COR with similar physiological outcome.
| Bacterial plant pathogens secrete toxins and inject effector proteins into the host cells to promote infection, and the identification of the individual functions of these molecules is essential to understand the infective process. Remarkably, some Pseudomonas strains have evolved a sophisticated strategy for manipulating hormonal balance by producing the toxin coronatine (COR), which mimics the plant hormone jasmonate-isoleucine (JA-Ile). The JA-Ile pathway plays a key role in plant immunity by activating defenses against fungal pathogens, while promoting bacterial growth by inhibiting the salicylic acid (SA)-dependent defenses required for Pseudomonas resistance. Here, we report that the effector HopX1 from a Pseudomonas syringae strain that does not produce COR exploits an alternative evolutionary strategy to activate the JA-Ile pathway. We show that HopX1 encodes a cysteine protease that interacts with and promotes the degradation of key JA pathway repressors, the JAZ proteins. Correspondingly, ectopically expressing HopX1 in the model plant Arabidopsis induces the expression of JA-dependent genes, and natural infection with Pseudomonas producing HopX1 promotes bacterial growth in a similar fashion to COR. Our results highlight a novel example by which a bacterial effector directly manipulates core regulators of hormone signaling to facilitate infection.
| Pseudomonas syringae is a widespread bacterial pathogen that causes disease on a broad range of economically important plant species. In order to infect, P. syringae produces a number of toxins and uses a type III secretion system (TTSS) to deliver effector proteins into eukaryotic cells [1],[2]. This mechanism is essential for successful infection by both plant- and animal-associated bacteria as bacterial mutants deficient in the TTSS are no longer pathogenic [3]. Effectors contribute collectively to pathogenesis inside the host cell by targeting host molecules and defeating plant defenses, which are based on two tiers of recognition by the innate immune system [4]. The first branch is triggered by the recognition of highly conserved microbe-associated molecular patterns (MAMPs) by host cell transmembrane proteins that function as pattern recognition receptors (PRRs), which in turn, activate MAMP-triggered immunity (MTI) [4]. The second branch recognizes type III effectors inside the plant cell via nucleotide-binding site-leucine-rich repeat (NB-LRR) resistance (R) proteins [4]. This leads to activation of effector-triggered immunity (ETI), and is characteristically associated with programmed cell death known as the hypersensitive response (HR). Accumulating evidence suggests that a primary function of microbial effectors is suppression of both MTI and ETI to avoid pathogen recognition during the infection process [5]. However, despite the fact that elucidating effector action is essential to understanding bacterial pathogenesis, the molecular function and host targets of the vast majority of effectors remain largely unknown.
Plant immunity relies on a complex network of small-molecule hormone signaling pathways [6]. Classically, salicylic acid (SA) signaling mediates resistance against biotrophic and hemi-biotrophic microbes such as P. syringae, whereas a combination of jasmonic acid (JA) and ethylene (ET) pathways activates resistance against necrotrophs such as the fungal pathogen Botrytis cinerea [6]. SA and JA/ET defense pathways generally antagonize each other and thus, elevated resistance against biotrophs is often correlated with increased susceptibility to necrotrophs, and vice versa [7]. The collective contribution of these two hormones during plant-pathogen interactions is crucial to the success of the interaction. Remarkably, some P. syringae strains have evolved a sophisticated strategy for manipulating hormonal homeostasis by producing coronatine (COR), a mimic of the bioactive jasmonate hormone, JA-isoleucine (JA-Ile) [8]. COR contributes to disease symptomatology by inducing chlorotic lesions [9]–[11], facilitates entry of the bacteria into the plant host by stimulating the opening of stomata [12],[13], and promotes bacterial growth by inhibiting SA-dependent defenses required for P. syringae resistance, because of its activation of the antagonistic JA pathway [14],[15]. COR, as the JA-Ile phytohormone, is perceived through a receptor complex formed by the F-box protein CORONATINE-INSENSITIVE 1 (COI1) and JASMONATE ZIM DOMAIN (JAZ) proteins [16]–[18]. COI1 is the F-box component of an SCF-(Skip-cullin-F-box)-type E3 ubiquitin ligase required for all JA-dependent responses tested so far [8],[19]–[22]. JAZ co-receptors are COI1 substrates that negatively regulate the JA-signaling pathway by directly interacting with and repressing transcription factors (TFs) that control JA-regulated genes [16]–[18],[23],[24]. Repression of TFs by JAZ is mediated by a general co-repressor machinery involving TOPLESS (TPL) and TPL-related proteins that interact with JAZ repressors through the adaptor protein NINJA [25]. The JAZ family of JA-repressors consists of 12 members in Arabidopsis that have emerged as central modulators of JA signaling [17],[18],[26]. Under stress conditions, COR or JA-Ile promotes the formation of JAZ-COI1 complexes, triggering JAZ degradation via the 26S proteasome [16]–[18]. This leads to de-repression of the TFs that initiate the transcription of JA-dependent genes, and repression of SA-dependent defenses against the bacteria. Thus, COR acts as a potent virulence factor in plants by triggering the degradation of JAZs. Acquisition of COR by bacterial pathogens has been of tremendous adaptive importance during host-pathogen evolution because it has allowed bacteria to manipulate the host hormonal network to promote susceptibility.
COR is produced by several bacterial strains distributed throughout the P. syringae phylogeny, but is particularly common among P. syringae pv. tomato strains such as DC3000 (Pto DC3000) [27],[28]. Interestingly, strains such as P. syringae pv. tabaci (Pta) 11528 that do not produce COR can still open stomata, suggesting that other virulence factors, probably type-III effectors, are used to activate the JA pathway instead of COR [12],[29],[30]. This is supported by several studies suggesting that COR and effectors act synergistically to induce JA responses [31]–[35]. Indeed, several effectors have been shown to modulate the expression of JA-inducible genes [33]. Moreover, gene expression profiles indicate that several JA-regulated genes are still induced in coi1 mutants upon P. syringae infection, indicating that JA responses are activated downstream or independently of COI1 [34]. Therefore, bacterial effectors should target other components of the JA pathway downstream of COI1 to activate JA responses, the best candidates being JAZ repressors. Recently, Mukhtar and colleagues developed a large-scale map of physical interactions between proteins from the reference plant Arabidopsis thaliana and effector proteins from P. syringae and the obligate biotrophic oomycete Hyaloperonospora arabidopsidis (Hpa) [36]. The experiment yielded a map of 6,200 interactions, and showed that pathogens from different kingdoms deploy independently evolved virulence effectors that interact with a limited set of highly connected cellular hubs to facilitate their diverse life-cycle strategies. Strikingly, two out of the five most significantly targeted plant hub proteins by effectors (namely RESPONSE TO LOW SULPHUR [LSU1, AT3G49580] and an unknown kinesin light-chain related protein [AT3G27960]), physically interact with several JAZ proteins [36]. Thus, it is plausible that pathogens may attempt to manipulate the JA pathway directly through JAZ proteins.
In previous studies, we generated a draft genome sequence of Pta 11528 and used a functional screen to infer its repertoire of T3SS effectors [30]. This led to the identification of Pta 11528 proteins with homology to previously described effectors. Since Pta 11528 does not produce COR, we hypothesized that it might have followed an alternative evolutionary strategy to activate the JA pathway by developing effector proteins that target JA signaling components. As noted above, we hypothesized that JAZ repressors would be the best candidate targets. Therefore, to study whether any such effector proteins target JAZ repressors, we developed a screen to analyze the stability of JAZ proteins in the presence of each Pta 11528 effector. We found that the Pta 11528 effector HopX1 encodes a cysteine protease that interacts with and degrades JAZ proteins in a COI1-independent manner. Ectopic expression of HopX1 in Arabidopsis induced the expression of JA-dependent genes, compromised the induction of SA-marker gene PR1 upon SA treatment, and complemented the growth of a COR-deficient Pto DC3000 strain during natural bacterial infections. Moreover, Pto DC3000 COR− growth increased by about one log (colony forming units [cfu]/cm2) when naturally expressing hopX1, but not its catalytic mutant version, indicating that HopX1 can effectively promote susceptibility when delivered by the natural TTSS. This increase in susceptibility was similar to the effect of supplementing bacteria with 2 µM of COR and independent of COI1. Altogether, the results indicate that similar to COR, HopX1 acts within the plant cell to promote activation of jasmonate-induced defenses and bacterial disease bypassing the need of JA-Ile perception.
It has been suggested that P. syringae pv. tabaci activates JA responses such as stomatal aperture without producing COR [12],[29],[30], which suggests that effectors from this strain could provide the same function by targeting components of the JA pathway. Whereas stomatal closure is part of a plant innate immune response triggered upon pathogen perception to restrict bacterial invasion, plant pathogenic bacteria have evolved specific virulence factors such as COR to promote stomata opening in order to circumvent such innate immune responses [12],[13]. To investigate whether live Pta 11528 bacteria can re-open plant stomata without producing COR, we first incubated host Nicotiana benthamiana leaves with virulent Pta 11528 bacteria. In this experiment, Pto DC3000 was used as a positive control, and a non-pathogenic TTSS defective strain Pta 11528 hrpV− unable to secrete effector proteins into the plant cell served as a negative control [37]. Incubation of N. benthamiana leaves with Pta 11528 hrpV− for 5 hours induced stomatal closure as described previously (Figure S1) [12]. In contrast, virulent Pta 11528 or Pto DC3000 bacteria maintained stomatal apertures similar to mock treatments. This indicates that effector proteins from Pta 11528 might play a role in re-opening of stomata as COR does, probably through action on the JA-signaling pathway.
To identify Pta 11528 effector proteins that could target JAZs, we analyzed a Pta 11528 effector library constructed in a binary vector within the T-DNA region, in which individual genes are expressed from the 35S promoter as genetic fusions to three C-terminal hemagglutinin (HA) epitope tags (Table S1). We transiently co-expressed JAZ5-HA with individual effector genes from this library or an empty vector (EV) control in N. benthamiana by agroinfiltration. Using this approach, we identified HopX1 as a Pta 11528 effector capable of compromising JAZ5 accumulation (Figure S2). To confirm this result and to exclude a potential effect of the HA tag, we developed an independent form of HopX1 with an N-terminal green fluorescent protein (GFP) fusion. We then transiently co-expressed an EV construct or GFP-hopX1 under the control of the 35S promoter with 35S:JAZ5-HA in N. benthamiana leaf tissue. Western blot analysis showed that GFP-HopX1 accumulated in N. benthamiana (Figure 1A). Similar to previous results, JAZ5-HA protein was detectable when co-expressed with the EV control but not (or only weakly) with GFP-HopX1, despite the fact that GFP-hopX1 co-expression did not affect JAZ5-HA mRNA expression levels (Figure 1B). These results indicate that HopX1 compromised the accumulation of JAZ5 protein when transiently co-expressed in N. benthamiana without affecting gene expression levels.
The Arabidopsis JAZ family contains 12 members grouped into four major phylogenetic clades (Figure S3) [17],[18],[26]. To test if HopX1 could compromise the accumulation of JAZ proteins other than JAZ5, we transiently co-expressed the 12 JAZ-HA genes individually with GFP-hopX1 or an EV control in N. benthamiana, and analyzed JAZ accumulation by Western blotting. We successfully detected eight out of 12 JAZ proteins when co-infiltrated with the EV control (Figure 1C). Interestingly, HopX1 compromised the accumulation of all eight JAZ proteins detected, indicating that HopX1 activity is not restricted to JAZ5 but targets the whole JAZ family (Figure 1C). We also analyzed the effect of HopX1 on the stability of additional JA related proteins such as the JA receptor COI1 and the downstream TF MYC2, which is a direct target of JAZ repressors [17]. HopX1 did not alter COI1 or MYC2 protein levels compared to the EV control (Figure 1D and 1E). Altogether, these results indicate that HopX1 compromises the accumulation of the JAZ family of JA-repressors in a specific manner.
HopX1 family members from different P. syringae strains are modular proteins that contain a putative cysteine-based catalytic triad and a novel conserved N-terminal domain [38]. The putative catalytic triad is required for effector function and consists of cysteine (C), histidine (H), and aspartic acid (D) residues conserved with cysteine proteases [38]. In order to determine whether these domains are conserved in HopX1 from Pta 11528, we performed a PSI-BLAST search with the HopX1Pta 11528 protein sequence. BLAST analysis revealed a near identical match with HopX1 from P. syringae pv. phaseolicola bacterial strains such as race 4 (99% identity), but showed significantly less homology with the respective Pto DC3000 homolog (72% identity) (Figure S4A). The putative catalytic triad residues were strictly conserved in HopX1Pta 11528, as was the novel N-terminal domain of unknown function typical of this family of effectors (Figure S4A). This conservation suggests that the domains may play an important role in the activity of this effector inside the plant cell, and that HopX1Pta 11528 might have cysteine protease activity.
To determine if HopX1 has cysteine protease activity in vitro, we used a kit designed for the detection of protease activity (serine, aspartic, cysteine, and metalloproteinases) using fluorometry based on the hydrolysis of a labeled casein general substrate [39]. As previously described by Nimchuck and colleagues [38], we did not detect any protease activity when purified recombinant HopX1 protein fused to maltose binding protein (MBP) was incubated with the casein-labeled substrate in vitro, indicating that this recombinant protein may be inactive or that it might lack a co-factor (Figure S4B). However, we detected significant protease activity when the casein substrate was incubated with HopX1-HA immunopurified directly from stable transgenic Arabidopsis plants expressing the hopX1 gene from a dexamethasone-inducible promoter (DEX) (Figure 2A), suggesting that HopX1 expressed in planta has protease activity. To test whether this activity of HopX1 required its conserved cysteine protease catalytic triad, we substituted the conserved Cys-179 residue within this domain for an alanine residue to generate the HopX1C179A mutant. HopX1C179A-HA inmunoprecipitated from transgenic Arabidopsis plants did not show any proteolytic activity compared to negative controls. The trypsin enzyme used as a positive control in these experiments showed much higher activity on the casein substrate than HopX1 (Figure 2A). These data indicate that HopX1 has protease activity, but seems to operate suboptimally on a general substrate in vitro. To test whether HopX1 may have evolved specific substrate selectivity, we incubated inmunoprecipitated HopX1 and HopX1C179A from Arabidopsis with recombinant MBP-JAZ5 expressed and purified from Escherichia coli cells with or without protease inhibitors. The amount of MBP-JAZ5 diminished significantly when incubated with HopX1 but not with HopX1C179A or buffer in the absence of protease inhibitors, but not in its presence (Figure 2B). Thus, HopX1, but not HopX1C179A, is capable of inducing JAZ5 degradation in vitro suggesting that the effector indeed acts as a protease on the JAZ5 substrate.
To confirm that degradation of JAZ proteins by HopX1 requires its putative cysteine protease activity in vivo, we next tested the effect of the C179A mutation on JAZ5-HA accumulation when the proteins were co-expressed transiently in N. benthamiana leaves (Figure 2C). JAZ5-HA was detectable in the presence of HopX1C179A, but not HopX1, indicating that this catalytic residue is indeed critical for the effect of HopX1 on JAZ5 levels (Figure 2C). Similarly, only wild-type HopX1, but not HopX1C179A, compromised the accumulation of additional JAZs such as JAZ1, JAZ2, JAZ9, and JAZ10 when transiently co-expressed in N. benthamiana leaves. However, HopX1 did not alter MYC2 protein levels (Figure S4C). To exclude that JAZ5 degradation is a general property of cysteine proteases, we analyzed the effects of previously described cysteine proteases such as HopC1 [40] and HopN1 [41] or an independent HopAD1 effector on the stability of JAZ5 when coexpressed transiently in N. benthamiana. As expected, JAZ5-HA accumulation was compromised in the presence of HopX1, but not when co-expressed with the EV control or with HopC1 or HopN1 proteins (Figure S5). The HopAD1 effector could not be detected in this assay. Taken together, these data suggest that degradation of JAZ proteins by HopX1 is specific and depends on a cysteine protease enzymatic activity that requires a conserved cysteine within the proposed catalytic triad.
Increased JA-Ile levels promote binding of JAZs to SCFCOI1 and subsequent degradation of JAZ repressors via the ubiquitin/26S proteasome pathway [16]–[18]. Therefore, degradation of JAZs by HopX1 might be direct (through its protease activity) or an indirect effect mediated by JA-Ile synthesis and COI1. To investigate if JAZ degradation by HopX1 is direct or indirect, we first analyzed whether HopX1-induced degradation of JAZ5 was dependent on the 26S proteasome pathway by using the proteasomal inhibitor MG132. In HopX1 coexpression experiments in N. benthamiana, JAZ5 was not detected in the presence of MG132, indicating that degradation does not require the proteasome (Figure S6). We next checked whether HopX1-induced degradation of JAZ proteins occurred in a COI1-dependent or independent manner. To test this, we first used a stable transgenic N. tabacum line silenced for expression of the NtCOI1 gene [42]. Reverse transcription (RT)-PCR analysis confirmed that control N. tabacum plants (Line VC, transformed with EV) accumulated NtCOI1 mRNA, whereas NtCOI1 transcripts were undetectable in N. tabacum plants silenced for the NtCOI1 gene (Line L18) (Figure S7A). Interestingly, NtCOI1-silenced N. tabacum plants produced few seeds, a phenotype reminiscent of infertile Arabidopsis coi1-1 plants (Figure S7B) [19]. We next analyzed the ability of HopX1 to trigger JAZ5 degradation in both EV- and NtCOI1-silenced plants when transiently co-expressed in N. tabacum, a species that also allows facile transient gene expression assays [43]. Strikingly, GFP-HopX1 compromised the accumulation of JAZ5 in both EV- and NtCOI1-silenced plants to the same extent (Figure 3A). This suggests that HopX1 triggers the degradation of JAZ proteins in a COI1-independent manner.
JAZ proteins are characterized by three sequence motifs, namely a relatively conserved N-terminal (NT) motif and the two highly conserved ZIM (central) and Jas (C-terminal) domains (Figure 3D) [17],[18],[23],[26],[44]. The ZIM domain mediates homo- and heteromeric interactions between Arabidopsis JAZ proteins [21],[45] and interacts with the general adaptor protein NINJA [25], whereas the C-terminal Jas domain is responsible for the interaction with COI1 and the TFs [21],[24],[46]. Dominant-negative JAZ variants lack part of the C-terminal Jas domain [16],[17],[22]. Consistently, these truncated JAZ forms (JAZΔJas) are resistant to COI1-dependent degradation after jasmonate treatment, and plants overexpressing them are JA-insensitive [17],[18]. To confirm whether degradation of JAZ proteins by HopX1 is COI1-independent, we analyzed the effect of HopX1 on three JAZΔJas proteins, namely JAZ1ΔJas, JAZ2ΔJas, and JAZ7ΔJas. We detected JAZ1ΔJas, JAZ2ΔJas, and JAZ7ΔJas proteins when co-expressed individually with the EV control in N. benthamiana leaf tissue (Figure 3B). However, none of these JAZΔJas protein forms accumulated when co-expressed with GFP-HopX1 (Figure 3B), confirming that HopX1 triggers the degradation of JAZ proteins in a COI1-independent manner.
All of our previous results imply that HopX1 targets JAZ proteins directly. Therefore, we next examined the ability of HopX1 and HopX1C179A to interact with all 12 Arabidopsis JAZ proteins in pull-down (PD) experiments. To do this, we used recombinant JAZ proteins fused to MBP and cell extracts of either wild-type plants (as negative controls), or transgenic plants expressing HopX1-HA or HopX1C179A-HA from the DEX inducible promoter. These experiments included protease inhibitors so that degradation of JAZs would not diminish a potential interaction with HopX1. As shown in Figure 3C, all full-length MBP–JAZ proteins tested interacted with wild-type HopX1-HA, but not with the mutant HopX1C179A-HA form. Similarly, MBP-HopX1 purified from E. coli cells co-immunoprecipitated with JAZ5-GFP when the recombinant effector protein was incubated with N. benthamiana plants extracts transiently expressing the JAZ5 transgene (Figure S8). We could also detect weak interaction with MBP-HopX1C179A but to a lesser extent than MBP-HopX1. Overall, this indicates that HopX1 interacts with JAZ proteins.
To further determine the requirements for HopX1 interaction with JAZ proteins, we expressed the JAZ deletion mutants JAZ51–91 (NT domain), JAZ592–163 (ZIM domain), and JAZ5164–274 (Jas domain) fused to MBP in E. coli and purified these fragments for PD analysis with cell extracts of transgenic plants expressing HopX1-HA (Figure 3D). As shown in Figure 3E the N-terminal fragment alone, or the C-terminus containing the Jas domain, did not interact with HopX1. However, the JAZ592–163 derivative containing the ZIM domain was sufficient for interaction with HopX1. Notably, the mutant version HopX1C179A could not be pulled down with any fragment of JAZ5, as we observed previously with full-length JAZ proteins (Figure 3E). The results indicate that HopX1 interacts directly with JAZ proteins though the central ZIM domain, whereas the other conserved domains seem to be dispensable. The data support our previous results showing that HopX1 triggers the degradation of all JAZ proteins, as the ZIM domain is present in all members of the JAZ family of repressors.
JAZ proteins are translated in the cytoplasm and localized predominantly in the nucleus [17],[18]. We next examined in which subcellular compartment the degradation of JAZs proteins by HopX1 takes place. To test this, we expressed GFP or GFP-HopX1 in N. benthamiana leaves and analyzed its subcellular localization using confocal microscopy. As reported previously [38], the fluorescent signal corresponding to GFP-HopX1 accumulated mainly in the cytoplasm 48 hours post inoculation (HPI) (Figure 3F). We could also detect GFP-HopX1 fusion protein within the nucleus in cross-middle nuclear sections using confocal microscopy analysis (Figures 3F and S9A), indicating that a pool of the effector enters the nucleus. To confirm this result, we further performed separation of nuclear extracts from the cytoplasmic fraction of N. benthamiana leaves transiently expressing GFP alone or GFP-HopX1 fusion protein. Crude fractions enriched for nuclei contained detectable levels of full length GFP-HopX1 although to a lesser extent than cytoplasmic fractions, confirming the microscopy results (Figure S9B). These results indicate that HopX1 could potentially degrade JAZ proteins both in the cytoplasm and nucleus.
The results described above suggest that HopX1 could be responsible for, or at least participate in, the induction of JA-dependent responses by Pta 11528. To test if HopX1 is sufficient for this activation, we analyzed JA marker gene expression in stable transgenic Arabidopsis lines expressing hopX1 or hopX1C179A genes from a dexamethasone-inducible promoter (DEX). These transgenic lines were constructed in Arabidopsis accession Aa–0, an ecotype that does not trigger cell death in response to certain HopX1 alleles [38]. As JA markers, we chose genes induced early after JA treatment such as JAZ5, JAZ10, and JAZ12. As shown in Figures 4A and S10, induction of hopX1 expression by DEX treatment for 36 hours strongly up-regulated all three JAZ marker genes in transgenic hopX1 Arabidopsis, whereas transcript levels remained low in transgenic plants expressing hopX1C179A or an EV control. Moreover, transgenic plants expressing hopX1 developed chlorotic symptoms after DEX treatment, in contrast with hopX1C179A plants (Figure 4B and 4C). Development of chlorosis was correlated with losses in chlorophyll content (Figure S11), a hallmark response of the JA pathway [47]. Similarly, transient expression of hopX1 by DEX treatment in N. benthamiana for 3 days, but not of hopX1C179A or an EV control, also induced strong chlorotic symptoms in the infiltrated area (Figure S12). Therefore, HopX1 triggers the activation of JA-dependent gene expression and JA-related phenotypes in a cysteine catalytic triad-dependent manner when ectopically expressed in Arabidopsis.
Crosstalk between JA and SA signaling pathways plays an important role in the regulation and fine-tuning of induced defenses against pathogens. The SA and JA/ET defense pathways generally antagonize each other, and induction of the JA pathway by P. syringae strains counteracts SA-dependent defenses [7]. Thus, we next determined whether activation of the JA-pathway by HopX1 could interfere with SA-dependent gene expression. We analyzed the expression of the SA marker gene PR1 in DEX-inducible transgenic Arabidopsis Aa–0 plants expressing hopX1 or hopX1C179A pre-treated with 1 mM of SA or a mock solution for 24 hours. Treatment with SA strongly induced PR1 expression in non-induced hopX1 or hopX1C179A transgenic plants (Figure 4D). Strikingly, DEX induction of the hopX1 effector gene for 24 hours strongly reduced subsequent SA induction of PR-1 expression (Figure 4D). Pre-induction of the hopX1C179A gene interfered weakly with PR-1 expression, but to a much lesser extent than wild-type HopX1. Overall, these results suggest that HopX1 activates the JA pathway to suppress SA-dependent defense responses to induce plant susceptibility.
On the basis of these results, we hypothesized that HopX1 could contribute to bacterial pathogenicity by mimicking COR-induced susceptibility. The jasmonate mimic COR is a bacterial toxin that contributes to bacterial invasion of the apoplast by Pto DC3000 [12]. Consistently, coronatine-deficient (COR−) Pto DC3000 mutants are less virulent on Arabidopsis plants when surface-inoculated [12]. To test if HopX1 contributes to bacterial pathogenicity during natural infections when delivered by the bacterial TTSS we compared bacterial replication of a Pto DC3000 COR− strain expressing hopX1Pta11528 or the hopX1C179A gene on Arabidopsis Col-0 plants infected by spray inoculation. This type of inoculation mimics natural infection conditions and is one of the most sensitive techniques to assess plant susceptibility to bacterial pathogens [48]. Pto DC3000 COR− growth was increased by about one log (cfu/cm2) when expressing hopX1 compared to the same strain containing hopX1C179A or an empty vector construct (Figure 4E). Strikingly, differences in bacterial growth promoted by HopX1 were abolished when DC3000 COR− strains expressing EV or the hopX1 or hopX1C179A genes were supplemented with 2 µM of COR (Figure 4E). This result supports the idea that HopX1 and COR act redundantly. Similar results were obtained when we compared bacterial replication on transgenic Arabidopsis Aa–0 plants expressing hopX1 or hopX1C179A infected with Pto DC3000 or the isogenic Pto DC3000 COR− strain by spray inoculation (Figure S13). These results indicate that HopX1 can complement the deficiency in COR production of Pto DC3000 COR−, and supports a key role for the effector catalytic triad in the activation of the jasmonate pathway.
We further studied whether Pto DC3000 COR− growth promotion by HopX1 was COI1-dependent by performing similar experiments in Arabidopsis plants lacking the COI1 receptor gene (coi1-30). As described previously, external application of COR did not restore Pto DC3000 COR− growth on coi1-30 plants (Figure 4F) [10],[11]. In contrast, expression of hopX1, but not hopX1C179A or EV, enhanced the growth of Pto DC3000 COR− in coi1-30 plants, similar to previous results in Col-0 plants (Figure 4F). Moreover, we examined whether JAZ proteins were degraded by HopX1 in vivo after bacterial infections on transgenic Arabidopsis Col-0 plants expressing the dominant negative JAZ1ΔJas-HA variant from the 35S promoter. As described above, JAZ proteins lacking the C-terminal Jas domain do not interact with COI1 and cannot be ubiquitinated and degraded by the proteasome, thus promoting JA-insensitivity [17],[18],[21],[24],[46]. JAZ1ΔJas-HA levels were strongly reduced in total plant extracts 24 hours after infiltration of Pto DC3000 COR− bacteria expressing hopX1, whereas infiltration of buffer or Pto DC3000 COR− expressing either hopX1C179A or EV did not cause any significant changes (Figure 4G). Consistent with the fact that JAZ1ΔJas-HA is resistant to COI1-dependent degradation by bacterial COR, challenge with Pto DC3000 did not alter JAZ1ΔJas-HA levels (Figure 4G). Finally, we investigated whether Pto DC3000 COR− bacteria expressing hopX1 could re-open plant stomata. To do this, we incubated N. benthamiana leaves with Pto DC3000 as a positive control or Pto DC3000 COR− bacteria expressing hopX1, hopX1C179A, or an EV control. Incubation of N. benthamiana leaves with Pto DC3000 COR− expressing an EV for 5 hours induced stomatal closure whereas stomata of N. benthamiana leaves incubated with Pto DC3000 remained open similar to mock treated control leaves as it was described previously (Figure S14) [12]. Strikingly, only Pto DC3000 COR− bacteria expressing hopX1, but not hopX1C179A, could re-open plant stomata of N. benthamiana leaves. This indicates that HopX1 from Pta 11528 plays a role in re-opening of stomata as COR does. These data are consistent with functional redundancy between HopX1 and the phytotoxin COR. Taken together, our data indicate that HopX1 acts inside the plant cell to promote activation of the JA pathway and induce susceptibility in Arabidopsis.
Levels of resistance in whole plants are influenced by systemic signals mediated, in many cases, by plant hormones. The importance of the role of hormones in biotic interactions is underlined by the increasing number of pathogenic microbes that are known to produce phytohormones or phytohormone mimics to perturb hormonal homeostasis and promote disease [6]. To date, production of cytokinins (CKs) [49], abscisic acid (ABA) [50], auxin [51], JA [52], and ET [53] has been reported in various bacterial or fungal species [6]. Remarkably, some pathogens produce hormone mimics. This is the case of some strains of P. syringae that produce COR, a mimic of the bioactive jasmonate JA-Ile, but synthesized via an unrelated biosynthetic pathway involving ligation of coronamic acid (cma) to the polyketide coronafacic acid (cfa) [2],[6]. Like JA-Ile, COR functions as an SA antagonist to promote virulence via suppression of host defenses. Notably, COR is produced by only a few P. syringae pathovars, whereas all Pseudomonas inject an array of effector proteins into the host cell that collectively promote disease by targeting and altering host cellular activities. The identification of their cellular targets is crucial to understand virulence. Recent data indicate that some effectors might impinge the JA pathway. For instance, Cui and colleagues showed that the effector AvrB indirectly perturbs JA signaling by interfering with the Arabidopsis mitogen-activated protein kinase MAP kinase 4 (MPK4) [35]. Moreover, several studies suggest that even in strains producing COR, effectors act synergistically with COR to induce the JA pathway [31]–[34]. Thus, it seems plausible that Pseudomonas manipulates the plant hormonal network through bacterial effectors to induce susceptibility. Here we have uncovered a novel molecular mechanism by which a bacterial strain (Pta 11528), which does not produce COR, activates the JA pathway to promote susceptibility through the degradation of JAZ repressors by the effector HopX1. In this work, we analyzed a Pta 11528 effector library containing ten identified effectors whereas the full Pta 11528 effector repertoire is predicted to secrete a suite of about 30 virulence effector proteins into the host cytoplasm [30]. Thus, it is plausible that additional effectors among the Pta 11528 effector repertoire not tested in this work may also destabilize JAZ proteins in a redundant manner with HopX1 to ensure activation of JA signaling and bacterial pathogenesis.
Expression of hopX1, but not hopX1C179A, induced expression of early JA-responsive genes while reducing SA-mediated induction of the SA-marker gene PR1, a hallmark response of the JA pathway. These phenotypes are evidently associated with the virulence function of the effector, as the presence of HopX1 promotes Pto DC3000 growth both in stable transgenic Arabidopsis plants expressing the effector under the control of a inducible promoter, or when the effector is delivered naturally by the TTSS of the bacteria. These results highlight a novel bacterial strategy to subvert the antagonistic relationship between host SA and JA signaling pathways. JA-Ile activates plant responses by promoting physical interaction between the E3 ligase COI1 and JAZ repressors. This interaction leads to JAZ polyubiquitination (poly-Ub) and subsequent degradation by the 26S proteasome, releasing TFs from repression [16]–[18]. Remarkably, HopX1 compromises the accumulation of JAZ repressors in NtCOI1-silenced plants, and triggers the degradation of dominant-negative JAZ variants lacking the C-terminal Jas domain, which is required for the interaction with COI1 [17],[18],[26]. Therefore, HopX1-triggered degradation of JAZ proteins occurs in a COI1-independent manner, thus allowing activation of the pathway in the absence of the hormone. The ability of HopX1 to promote degradation of JAZ proteins and activate JA responses suggests that this is a strategy to promote disease. PD and co-immunoprecipitation experiments demonstrated that HopX1 interacts with JAZ proteins through its central ZIM domain, conserved in all 12 Arabidopsis JAZ repressors and their JAZΔJas variants. The interaction with the ZIM domain is consistent with HopX1-triggered degradation of all full-length JAZs and truncated derivatives tested in our assays, and indicates that the effect of HopX1 on JAZs is direct or at least through the same protein complex. JAZ proteins are translated in the cytoplasm and then localized mainly, but not exclusively, in the nucleus [17],[18],[54]. For example, JAZ1 can be detected in both cytosol and nucleus [55]. Moreover, Withers and colleagues recently reported a MYC2-dependent mechanism for nuclear import of cognate JAZs transcriptional repressors in Arabidopsis [54], which implied an equilibrium of JAZ proteins between the nucleus and the cytoplasm. Thus, it is conceivable that HopX1 could potentially degrade JAZ proteins both in the cytoplasm and nucleus.
Interaction and degradation of JAZs by HopX1 suggest that this effector has proteolytic activity. HopX1 family members are modular proteins that contain a putative cysteine-based catalytic triad and a novel conserved N-terminal domain [38]. The putative catalytic triad consists of cysteine (C), histidine (H), and aspartic acid (D) and is similar to that utilized by diverse enzyme families such as cysteine proteases and peptide N-glycanases (PNGases), which are all members of the transglutaminase (TGase) protein superfamily [38]. While previous reports failed to detect protease activity in vitro using general substrates, genetic approaches demonstrated the functional relevance of the conserved catalytic triad of HopX1 family members [38]. For example, mutation of this putative catalytic triad of various HopX family members abolished avirulence activity on R2-expressing bean cultivars and also prevented initiation of cell death in Arabidopsis following transient expression assays [38]. We found that HopX1 has protease activity on both general and specific substrates when the effector was directly immunopurified from plants and that this activity required Cys-179 within the conserved catalytic triad. Previously, Coaker and colleagues reported that the cysteine protease effector AvrRpt2 requires host general folding catalyst cyclophilins like Arabidopsis ROC1 to activate AvrRpt2 cysteine protease activity once the effector is translocated into the plant cell [56],[57]. Activated AvrRpt2 undergoes autoprocessing of its N-terminus to yield the mature protease that cleaves the host target RIN4 in planta [58]. Thus, it is possible that HopX1 also requires a host eukaryotic chaperone into the plant cell for the activation of the protease activity in a similar fashion as AvrRpt2. These results, together with the ability of HopX1 to interact with JAZs in PD and co-immunoprecipitation assays, suggest a model in which HopX1 acts on JAZs directly or indirectly through a third plant protein in the same complex. Noteworthy, HopX1C179A shows markedly reduced affinity for JAZ proteins in PD and co-immunoprecipitation assays compared to wild-type HopX1 (Figures 3C, 3D, and S8). Thus, it is possible that the lack of activity of HopX1C179A on JAZs could be due not exclusively to a missing enzymatic activity but also to a compromised ability to interact with its host targets, or a combination of both simultaneously. In this regard, we can also not completely exclude that HopX1 activity on JAZs could be due to an additional plant protease that co-purifies between the same protein complex. Despite previous reports showing that the cysteine protease effector AvrPphB cleaved host targets into smaller fragments [58],[59], we did not see smaller bands in our assays including free MBP. We speculate that this could be due to the small size of JAZ proteins (25–35 KDa). However, it is also possible that HopX1 degrades JAZ proteins at multiple sites and triggers target degradation in a similar fashion to the cysteine protease effector AvrRpt2 on its 30-kDa host substrate protein RIN4 [58],[59]. On the other hand, it is possible that the immunoprecipitation of HopX1-HA or C179A-HA from transgenic plants contains additional proteases that could affect MBP stability. This would explain why in Figure 2B MBP-JAZ5 protein purified from E. coli is more stable in the presence of general cocktail protease inhibitors. Despite this, unspecific degradation would occur in all samples to the same extent and therefore, the effect we observed in MBP-JAZ5 in the presence of HopX1 is likely to be specific. Moreover, we found that substitution of the conserved Cys-179 amino acid of HopX1 to alanine also compromised HopX1-mediated JAZ5 degradation in vivo. Similarly, activation of JA-dependent gene expression and JA-related phenotypes occurred in a Cys-179 dependent manner when expressed ectopically in Arabidopsis. Furthermore, in contrast to wild-type HopX1, HopX1C179A delivered by the TTSS was unable promote growth of Pto DC3000 COR− bacteria. Overall, these data suggest that HopX1 family members are proteolytic enzymes of the cysteine protease family, and that this activity is required for function.
Several JAZ pre-mRNAs are subject to alternative splicing, which, at least for JAZ10, results in truncated JAZΔJas proteins that are resistant to proteasomal degradation because they are unable to interact with COI1, thereby causing dominant JA-insensitive phenotypes [26],[45],[60]. The alternative splicing of JAZ genes provides a feedback mechanism to efficiently repress the JA signal output and reduce the fitness costs associated with over-stimulation of the signaling pathway [45],[60]. Degradation of JAZΔJas variants by HopX1 suggests that this effector should promote a sustained activation of the JA pathway, favoring susceptibility. In agreement with this hypothesis, Pto DC3000 COR− expressing hopX1 promoted bacterial replication in Arabidopsis plants (Figure 4E and 4F). Moreover, Pto DC3000 COR−, which does not produce COR, grew better in the presence of the effector (DEX in Figure S13) than the wild-type strain producing COR in the absence of the effector (mock in Figure S13). Therefore, our results suggest that evolution has shaped this effector–host interaction to maximize the activation of the JA response pathway through direct targeting of most forms of JAZs. Besides the 12 JAZ repressors, six additional Arabidopsis proteins contain a ZIM domain. Collectively, these 18 proteins are known as the TIFY proteins [23],[61]. It remains to be determined whether HopX1 targets any of these ZIM domain containing proteins, and if they also play a role in plant defense against bacteria.
Notably, HopX homologs are found in diverse phytopathogenic bacteria including Pto DC3000. BLAST analysis revealed a near identical match of HopX1Pta 11528 with HopX1 from P. syringae pv. phaseolicola bacterial strains such as race 4 (Pph race 4) (99% identity) but significantly less homology with the respective Pto DC3000 homolog (72% identity) (Figure S4A). Several lines of evidence suggest that the Pto DC3000 and Pta 11528 alleles of HopX1 play different roles in the plant cell. Firstly, HopX1Pto DC3000 triggers cell death in several Arabidopsis ecotypes including Col-0 and Ws-0 whereas HopX1Pph race 4 (which is almost identical to the Pta 11528 allele) does not [38]. Secondly, HopX1Pto DC3000 did not compromise the accumulation of JAZ5 when co-expressed transiently in N. benthamiana (Figure S15). Consistently, transgenic Arabidopsis Col-0 plants expressing the dominant negative JAZ1ΔJas variant infected with Pto DC3000 did not alter JAZ1ΔJas levels (Figure 4G). This indicates that these plants are resistant to both HopX1Pto DC3000 and COI1-dependent degradation by bacterial COR. The ability of HopX1Pta 11528 to trigger JAZ degradation and activation of JA responses may reflect specialized functions among the broad divergence across the HopX family members. It is tempting to speculate that redundancy between COR and HopX1 has allowed strains that produce the phytotoxin to evolve HopX1 for other functions and in doing so, lose its JA-inducing activity. Overall, HopX1 exemplifies a bacterial strategy for pathogenicity in which the JA pathway is targeted through direct degradation of JAZ repressors to promote pathogenesis.
Finally, emerging data suggest that bacterial pathogens have evolved effectors to manipulate important plant hormones that regulate defense response such as SA and JA [6]. For, example, the P. syringae effector HopI1 directly targets Hsp70 in choloroplasts to suppress SA accumulation [62] whereas the effector AvrB promotes JA signaling likely through an indirect mechanism via the activation of MPK4 [35]. Very recently, the P. syringae effector HopZ1a was reported to interact and acetylate several JAZ proteins through a putative acetyltransferase activity [55]. HopZ1a-mediated acetylation induces JAZ1 degradation through an undefined mechanism that is dependent on COI1. This leads to activation of JA-dependent gene expression and plant susceptibility [55]. It is possible that post-translational modifications of JAZ1 induces or represses 26S proteasome degradation triggered by COI1. In this study, we found that hopX1 encodes a cysteine protease, activity that is required for degradation of JAZs by HopX1. HopX1 associates with JAZ proteins through its central ZIM domain and degradation occurs in a proteasome- and COI1-independent manner, highlighting the different strategies used by bacterial effectors to target similar host components. Importantly, ectopic expression of HopX1 in Arabidopsis induces the expression of JA-dependent genes and represses SA-induced markers, whereas delivery of HopX1 by the natural TTSS of Pseudomonas partially re-opens stomata during the infection process and promotes susceptibility to a similar extent as the addition of COR. These results highlight a novel molecular mechanism by which bacterial effectors directly manipulates core regulators of hormone signaling to facilitate infection. Indeed, oomycete pathogens were also found to produce effectors that interact with JAZ3 [36]. The recent findings highlight the JA receptor complex as a major common and critical hub of suppression for diverse pathogens during the arms race between plants and pathogens.
Arabidopsis Aa–0 plants were transformed as described previously [63] using the binary vector pTA7002 [64] containing the hopX1 or the hopX1C179A gene fused to a C-terminal HA under the control of a dexamethasome (DEX)-inducible promoter. Homozygous single-insertion transformants were selected in the T3 generation and HopX1 or HopX1C179A accumulation was confirmed by Western blots after induction with 30 µM DEX with 0.01% Silwet L-77. To generate transgenic plants expressing JAZ1ΔJas-HA in Col-0 background, JAZ1ΔJas was amplified with Expand High Fidelity polymerase (Roche) using Gateway-compatible primers (AtJAZ1 ΔJas, 5′-ggggacaagtttgtacaaaaaagcaggcttcATGTCGAGTTCTATGGAATGTTCTG-3′ and 5′-ggggaccactttgtacaagaaagctgggtcCTTGGCTAAGCTATTAGCGGT-3′). PCR products were cloned into pDONR207 with a Gateway BP II kit (Invitrogen) and sequence verified. This plasmid, a Gateway LR II kit (Invitrogen), and the pGWB14 [65] destination vector was used to generate 35S:JAZ1ΔJas-HA. These constructs were transferred to Agrobacterium tumefaciens strain C58C1 by freeze thawing and then transformed in Col-0 plants by floral dipping method [66]. Hygromycin-resistant plants were selected and their T2 progenies propagated for subsequent analysis. The knockout line coi1-30 [67] and stable N. tabacum lines silenced for an EV (line VC) or the NtCOI1 gene (line18) [42] were previously described.
N. benthamiana and N. tabacum plants were grown in controlled environment chambers at an average temperature of 24°C (range 18°C–26°C), with 45%–65% relative humidity under long day conditions (16 h light). In contrast, A. thaliana plants were grown in controlled environment chambers at an average temperature of 22°C (range 16°C–24°C), with 45%–65% relative humidity under short day conditions (10 h light). N. benthamiana and N. tabacum plants were grown for three to four weeks prior to A. tumefaciens-mediated transformation. Arabidopsis plants of four to six weeks were usually analyzed in bacterial growth assays.
The E. coli strain DH5α was used for cloning, and small- and large-scale plasmid isolation. A. tumefaciens C58C1 was used for transient expression in N. benthamiana. Pseudomonas strains used in this study were Pseudomonas syringae pv. tomato (Pto) DC3000, Pseudomonas syringae pv. tabaci (Pta) 11528, a non-pathogenic TTSS defective Pta 11528 hrpV− bacteria unable to secrete effector proteins into the plant cell [37] and the coronatine-deficient Pto DC3000 strain (Pto DC3000 COR−) which is a Pto DC3000 AK87 mutant that carries mutations in cmaA (coronamic acid A) and cfa6 (coronafacic acid 6) [11]. HopX1 or HopX1C179A were cloned into pCPP5040, a derivative of the broad-host-range vector pML123 [68], which expresses insert genes from the nptII promoter, and generates protein products for expression in Pseudomonas species [41]. This vector was a kind gift from Emilia Lopez Solanilla and Pablo Rodriguez Palenzuela. These plasmids were introduced into Pto DC3000 COR− by triparental mating in which pRK600 was used as a helper plasmid.
Leaf discs from 4- to 5-week-old N. benthamiana young leaves were exposed to white light for 1 hour while submerged in a solution containing 50 mM KCl, 10 µm CaCl2, and 10 mm MES-KOH, (pH = 6.1) to induce stomatal aperture. Subsequently, leaf discs were immersed in buffer or bacterial suspension at 5×108 cfu/ml (optical density at 600 nm [OD600] = 1) in 50 mM KCl, 10 µm CaCl2, and 10 mm MES-KOH (pH = 6.1). The samples were incubated under the same conditions for 5 hours. Abaxial leaf surfaces were observed with a microscope (Leica DMR), and stomatal aperture was measured using ImageJ software.
Statistical significance based on t test analysis was developed by GraphPad Prism program. Seven independent samples were used to analyze the significance of bacterial growth results.
Seventeen independent stomata were used to analyze stomatal aperture in each condition.
Growth and transient expression conditions were as described previously [69] using the A. tumefaciens strain C58C1. For transient gene expression, A. tumefaciens was syringe infiltrated in N. benthamiana or N. tabacum leaves at OD600 = 0.4, whereas for transient gene co-expression assays both constructs were infiltrated at OD600 = 0.3. Samples were collected after two days. The genes expressed in this paper were 35S:hopX1-HA, DEX:hopX1-HA, DEX:hopX1C179A-HA, 35S:GFP, 35S:GFP-hopX1, 35S:GFP-hopX1C179A, 35S:JAZ-HA (Arabidopsis cDNA coding sequences cloned in constructs for all 12 JAZs), 35S:COI1-GFP, 35S:MYC2-HA, 35S:JAZ1ΔJas-HA, 35S:JAZ2ΔJas-HA, and 35S:JAZ7ΔJas-HA. We also used the Pto DC3000 effector genes 35S:hopX1Pto DC3000-HA, 35S:HopC1Pto DC3000-HA, 35S:HopAD1Pto DC3000-HA, and 35S:HopN1Pto DC3000-HA. For inhibitory assays in planta, MG132 (Sigma) inhibitor was co-infiltrated with A. tumefaciens at a concentration of 100 µM.
Total proteins were extracted from leaf tissue by homogenization in extraction buffer (100 mM Tris-HCl [pH 7.5], 150 mM NaCl, 5 mM EDTA, 5% glycerol, 10 mM DTT, 2% PVPP, 1 mM PMSF, protease inhibitors cocktail [Roche], and 0.5% Triton X-100). In all experiments protein samples were equilibrated to equivalent concentrations of total proteins. Extracted proteins were fractioned by 8%–10% SDS-PAGE, transferred onto HybondTM-P membranes (Amersham) and incubated with anti-HA-horseradish peroxidase (Roche) or anti-GFP-horseradish peroxidase antibody (Milteny Biotec). Immunodetection was performed with ECL chemiluminiscence reagent (GE Healthcare) or Supersignal West Femto (Thermo Scientific).
MBP-JAZ fusion proteins were generated as previously described [21]. Ten-day-old Arabidopsis wild-type Col-0 seedlings and lines expressing DEX:hopX1-HA or DEX:hopX1C179A-HA were induced with 30 µM DEX plus 0.01% Silwet L-77 or a mock solution for 24 hours. Seedlings were ground in liquid nitrogen and homogenized in extraction buffer containing 50 mM Tris-HCl, (pH 7.4), 150 mM NaCl, 10% glycerol, 0.2% NP-40, 1 mM DTT, 1 mM phenylmethylsulphonyl fluoride, 5 mM MgCl2, 50 mM MG132 (Sigma-Aldrich), and complete protease inhibitor (Roche). After centrifugation (16,000 g at 4°C), the supernatant was collected. For in vivo PD experiments, 6 µg of resin-bound MBP fusion protein was added to 250 µg of pre-equilibrated total protein extract and incubated for 1 h at 4°C with rotation. After washing, samples were denaturalized, loaded on 8% SDS-PAGE gels, transferred to nitrocellulose membranes, and incubated with anti-HA-horseradish peroxidase (Roche). A 5 µl aliquot of MBP-fused protein of each sample was run into SDS-PAGE gels and stained with Coomassie brilliant blue to confirm equal protein loading.
Immunoprecipitation of HopX1-HA and HopX1C179A-HA from stable transgenic Arabidopsis plants using the anti-HA affinity matrix (Roche) was performed according to the manufacturer's instructions with some modifications (Anti-HA Affinity Matrix, catalogue number 11 815 016 001, Roche). Briefly, transgenic Arabidopsis seedlings expressing hopX1-HA and hopX1C179A-HA or wild-type Col-0 used as a negative control were induced by spraying with 30 µM DEX plus 0.01% Silwet L-77 for 24 hours. The material was ground in liquid nitrogen and homogenized in extraction buffer containing 100 mM Tris-HCl (pH 7.5), 150 mM NaCl, 5 mM EDTA, 5% glycerol, 10 mM DTT, 2% PVPP, 1 mM PMSF, protease inhibitors cocktail (Roche), and 0.5% Triton X-100. The samples were centrifuged twice at 16,000 g at 4°C. The supernatant was incubated for 5 hours (4°C, with rotation) with the anti-HA affinity matrix (Roche), and then washed once with 1 ml of extraction buffer and three times with 1 ml of cool TBS (50 mM Tris-HCl [pH 7.5], 150 mM NaCl, [pH = 7.4]) to remove all protease inhibitors from the anti-HA affinity matrix. Elution of HA proteins from the anti-HA affinity matrix was performed by incubating with 1 mg/ml of HA peptide (Roche) in TBS buffer for 30 minutes at 37°C with strong shaking. Supernatant was recovered by centrifugation and transferred to a fresh tube for protease activity assays. Immunoprecipitated HopX1-HA and HopX1C179A-HA was pre-equilibrated for similar amounts of protein and maintained at 4°C until used.
In vitro Protease Fluorescent Detection kit to determine protease activity was performed according to the manufacturer's instructions (PF0100-1KT, Sigma). Briefly, 10 µl of immunoprecipitated HopX1-HA or HopX1C179A-HA (or MBP-HopX1 for Figure S4B) was incubated with 20 µl of Incubation buffer and 20 µl of fluorescein isothiocyanate (FITC)-casein substrate overnight in the dark at 37°C with moderate shaking. The reaction was stopped by treatment with 150 µl of 10% trichloroacetic acid (TCA) for 30 min at 37°C in the dark. Samples were then centrifuged, and supernatant was recorded for fluorescence intensity with excitation at 485 nm and monitored for the emission wavelength of 535 nm (485/535 nm).
For in vitro experiments regarding protease activity of HopX1 on JAZs, we incubated 20 µl of immunoprecipitated HopX1-HA or HopX1C179A-HA with 5 µl of recombinant JAZ5 fused to MBP expressed and purified from E. coli cells in a total volume of 50 µl TBS (with or without general protease inhibitors, Roche). The reaction was incubated overnight at 37°C with moderate shaking. To stop the reactions all samples were denaturalized, and then loaded on 10% SDS-PAGE gel, transferred to a nitrocellulose membrane, and incubated with anti-MBP (AbCAM). Immunodetection was performed with ECL chemiluminiscence reagent (GE Healthcare)
HopX1 was placed into pENTR/D-TOPO (Invitrogen) using the 5′-ggggacaagtttgtacaaaaaagcaggcttcATGAGAATTCACAGTGCTGGTCA-3′ and 5′-ggggaccactttgtacaagaaagctgggtcTTATCTTCGTGGAGGCATGCCTTTAGACG-3′ primers, and then recombined into the gateway binary vector pGWB6 to create constructs that express N-terminal GFP fusion proteins. This was then transiently expressed in N. benthamiana leaves by Agrobacterium-mediated transformation. Localization of GFP fusion was visualized with sequential laser scanning confocal microscopy, using a Leica Confocal SP5 with sequential imaging at 488 nm excitation and 505–525 nm emission (green/GFP) and 633 nm excitation and 660 nm emission (red/chlorophyll).
Plant cell fractionation was performed by using the CelLytic PN Isolation/Extraction kit for plant leaves (CELLYTPN1-1KT, Sigma) according to the manufacturer's instructions.
The chlorophyll content of leaves was measured by acetone extraction according to Arnon [70]. Briefly, 100–150 mg tissue per sample was extracted for two hours in 400 µl of acetone 80% (v/v) at 4°C in the dark with constant shacking. The homogenate was centrifuged at 3,000 rpm for 2 minutes. The supernatant was saved (V1) and 0.2 ml of the supernatants were diluted in a known volume of acetone 80% until the absorbance of the extract at 663 nm and 645 nm was read between 0.2–0.8 (V2) using a SpectraMax Absorbance Microplate Reader. The concentration of chlorophyll a (Ca), b (Cb), and total chlorophyll (Ct) was calculated using Arnon's equations: Ca = 12.7 * A663−2.63 * A645, Cb = 22.9 * A645−4.68 * A663, and Ct = Ca+Cb * [V1 * (V2+0.2)/(0.2 * P]. Ca, Cb are expressed in µg * ml−1, P in grams of fresh leave tissue, V in ml, and Ct in µg * ml−1 of fresh weight.
Quantitative RT-PCR for JA-dependent gene expression experiments were performed with RNA extracted from 10-day-old seedlings grown on liquid MS media that were treated with 5 µM DEX for 36 h or a mock solution. For quantitative RT-PCR analysis of PR1 expression in stable transgenic Arabidopsis lines expressing the hopX1 or hopX1C179A genes, 10-day-old seedlings grown on liquid MS media were DEX-induced for 24 hours and then treated with 1 mM SA or a mock solution for additional 24 hours. For each experiment, three biological replicates, consisting of tissue pooled from 15 to 20 plants were taken. For semi-quantitative RT-PCR of AtJAZ5 expression levels when co-expressed with HopX1 in N. benthamiana, discs from six independent leaves were collected and frozen in liquid nitrogen two days after infiltration. For semi-quantitative RT-PCR analysis of stable NtCOI1-silenced N. tabacum plants, discs from six independent leaves of four weeks old plants were collected and frozen in liquid.
RNA extraction and cleanup was done using Trizol reagent (Invitrogen) followed by RNeasy mini kit (Qiagen) and DNase digestion to remove genomic DNA contamination. cDNA was synthesized from 0.5 to 1 µg of total RNA with the high-capacity cDNA reverse transcription kit (Applied Biosystems). Two microliters from one-tenth diluted cDNA was used to amplify selected genes. For quantitative PCR analysis, Power SYBR Green was used for gene amplification (Applied Biosystems). Quantitative PCR was performed in 96-well optical plates in a 7300 Real Time PCR system (Applied Biosystems). Data analysis shown was done using three technical replicates from one biological sample; similar results were obtained with two additional independent biological replicates.
Primers for genes used here are as follows: AtJAZ10, 5′-GAGAAGCGCAAGGAGAGATTAG-3′ and 5′-CTTAGTAGGTAACGTAATCTCC-3′; AtJAZ5, 5′-AAAGATGTTGCTGACCTCAGTG -3′ and 5′-CCCTCCGAAGAATATGGTCA-3′; AtJAZ12, 5′-CATCTAATGTGGCATCACCAG-3′ and 5′-TGCCTCCTTGCAATAGGTAGA-3′; AtPR1, 5′-AAGTCAGTGAGACTCGCATGTGC-3′ and 5′-GGCTTCTCGTTCACATAATTCCC-3′; AtACTIN8, 5′-CCAGTGGTCGTACAACCGGTA -3′ and 5′- TAGTTCTTTTCGATGGAGGAGCTG-3′; HopX1, 5′-TAGCAAGCTTCGCTTACG-3′ and 5′-GTTTCACGCGTAACCTTG-3′; NtCOI1, 5′-GAAGATCTTGAATTGATGGC-3′ and 5′-CCCAGAAGCATCCATCTCAC-3′; and NtαTubulin, 5′-AGTTGGAGGAGGTGATGATG-3′ and 5′-TATGTGGGTCGCTCAATGTC-3′.
Transgenic Arabidopsis plants expressing hopX1 and hopX1C179A were induced by spraying with 30 µM DEX plus 0.01% Silwet L-77 or a mock solution for 24 hours, and then infected with selected bacteria. All bacterial growth assays in Arabidopsis were performed by spray inoculation as described [48]. Briefly, overnight bacterial cultures were pelleted and resuspended in sterile 10 mM MgCl2. Plants were sprayed with a bacterial suspension containing 108 (cfu)/ml bacteria (OD600 = 0.2) with 0.04% Silwet L-77. Leaf discs were harvested after two days and ground in 10 mM MgCl2. Serial dilutions of leaf extracts were plated on LB agar with appropriate antibiotics. Each data point represents the average of seven replicates, each containing two leaf discs from different plants. Error bars indicate standard errors of the mean (SEM). These experiments were repeated at least three times with similar results, and representative results are shown.
Immunoblots showing JAZ1ΔJas degradation in Pto DC3000 COR− expressing hopX1 infected Col-0 Arabidopsis plants were performed after syringae infiltration of the indicated bacterial suspension at 108 (cfu)/ml (OD600 = 0.2) into the leaves. Samples were collected after 24 hours for protein analysis.
|
10.1371/journal.pcbi.1005569 | Active Vertex Model for cell-resolution description of epithelial tissue mechanics | We introduce an Active Vertex Model (AVM) for cell-resolution studies of the mechanics of confluent epithelial tissues consisting of tens of thousands of cells, with a level of detail inaccessible to similar methods. The AVM combines the Vertex Model for confluent epithelial tissues with active matter dynamics. This introduces a natural description of the cell motion and accounts for motion patterns observed on multiple scales. Furthermore, cell contacts are generated dynamically from positions of cell centres. This not only enables efficient numerical implementation, but provides a natural description of the T1 transition events responsible for local tissue rearrangements. The AVM also includes cell alignment, cell-specific mechanical properties, cell growth, division and apoptosis. In addition, the AVM introduces a flexible, dynamically changing boundary of the epithelial sheet allowing for studies of phenomena such as the fingering instability or wound healing. We illustrate these capabilities with a number of case studies.
| We present a detailed analysis of the Active Vertex Model to study the mechanics of confluent epithelial tissues and cell monolayers. The model combines the commonly used Vertex Model for describing epithelial tissue mechanics with the active matter dynamics extensively studied in soft matter physics. We utlise an exact mathematical mapping that enables a very efficient numerical implementation using standard methods for simulating particle-based models. System sizes accessible to this model allow us to probe the dynamical motion patterns that occur in tissues over a range of length- and time-scales previously inaccessible to available simulation tools. Our model also includes a number of essential features required to properly describe actual biological systems such as cell growth, cell division and aptotsis, as well as the dynamic boundary of the epithelial sheet. This allows us to study phenomena such as the finger-like protrusions in cell monolayers and processes related to wound healing. The model is implemented into the SAMoS simulation software package and is publically available under a non-restrictive open source licence.
| Collective cell migration [1, 2] in epithelial tissues is one of the key mechanisms behind many biological processes, such as the development of an embryo [3], wound healing [4, 5], tumour metastasis and invasion [6]. Due to their layered, tightly connected structure [7], epithelial tissues also serve as an excellent model system to study cell migration processes. Over several decades [8] extensive research efforts have been devoted to understanding molecular processes that lead to cell migration [9] and, at larger scales, on how cell migration drives complex processes at the level of the entire tissue, such as morphogenesis. With recent advances in various microscopy techniques combined with the development of sophisticated automatic cell tracking methods, it is now possible to study collective migration patterns of a large number of cells over extended periods of time with cell-level resolution, both in vitro and in vivo. Traction force microscopy [10] experiments revealed that collective cell motion is far more complex than expected [11–13]. It is often useful to draw parallels between the collective behaviour of tissues and systems studied in the physics of colloids, granular materials and foams as these can provide powerful tools for understanding complex cell interactions in biological systems. For example, in a homogenous cell monolayer, one observes large spatial and temporal fluctuations of inter-cellular forces that cannot be pinpointed to a specific location, but cover regions that extend over several cells [14]. These are reminiscent of the fluctuations observed in supercooled colloidal and molecular liquids approaching the glass transition [11] and include characteristic features of the dynamical and mechanical response, such as dynamical heterogeneities and heterogeneous stress patterns, that were first observed in glasses, colloids and granular materials and that have extensively been studied in soft condensed matter physics [15]. It has also been argued that the migration patterns are sensitive to the expression of different adhesion proteins [16] as well as to the properties of the extracellular environment [17], such as the stiffness of the substrate [18, 19]. These observations lead to the development of the notion of plithotaxis [12], a universal mechanical principle akin to the more familiar chemotaxis, which states that each cell tends to move in a way that maintains minimal local intercellular shear stress. While plausible, it is yet to be determined whether plithotaxis is indeed a generic feature in all epithelial tissues.
Equally fascinating are the experiments on model systems that study cell migration in settings designed to mimic wound healing [5, 20–23]. For example, the existence of mechanical waves that span the entire tissue and generate long-range cell-guidance have been established in Madin-Darby Canine Kidney (MDCK) epithelial cell monolayers [23]. Subtle correlations between purse-string contractility and large-scale remodelling of the tissue while closing circular gaps have also been identified [22]. Finally, a mechanism dubbed kenotaxis has been proposed [20], which suggests that there is a robust tendency of a collection of migrating cells to generate local tractions that systematically and cooperatively pull towards the empty regions of the substrate.
On the developmental side, in pioneering work, Keller et al. [24] constructed a light-sheet microscope that enabled them to track in vivo positions of each individual cell in a zebrafish embryo over a period of 24h. A quantitive analysis [25] of the zebrafish embryo was also able to relate mechanical energy and geometry to the shapes of the aggregate surface cells. Another extensively studied system that allows detailed tracking of individual cells is the Drosophila embryo [26–30]. In recent studies that combined experiments with advanced data analysis, it was possible to quantitatively account for shape change of the wing blade by decomposing it into cell divisions, cell rearrangements and cell shape changes [31, 32]. Finally, it has recently become possible to track more than 100,000 individual cells in a chick embryo over a period exceeding 24 hours [33]. This was achieved by developing an advanced light-sheet microscope and state-of-the-art data analysis techniques designed to automatically track individual cells in a transgenic chick embryo line with the cell membranes of all cells in the embryonic and extra embryonic tissues labelled with a green fluorescent protein tag. All these experiments and advanced data analysis techniques provide unprecedented insights into the early stages of embryonic development, making it possible to connect processes at the level of individual cells with embryo-scale collective cell motion patterns.
While there have been great advances in our understanding of how cells regulate force generation and transmission between each other and with the extracellular matrix in order to control their shape and cell-cell contacts [9], it is still not clear how these processes are coordinated at the tissue-level to drive tissue morphogenesis or allow the tissue to maintain its function once it reaches homeostasis. Computer models of various levels of complexity have played an essential role in helping to answer many of those questions [34]. One of the early yet successful approaches has been based on the cellular Potts model (CPM) [35, 36]. In the CPM, cells are represented as groups of “pixels” of a given type. Pixels are updated one at a time following a set of probabilistic rules. Pixel updates in the CPM are computationally inexpensive [37], which allows for simulations of large systems. In addition, the CPM extends naturally to three dimensions [38]. While very successful in describing cell sorting as well as certain aspects of tumour growth [39], CPM has several limitations, the main one being that the dynamics of pixel updates is somewhat artificial and very hard to relate to the dynamics of actual cells. This problem has been to some extent alleviated by the introduction of the subcellular element model (ScEM) [40, 41]. ScEM is an off-lattice model with each cell being represented as a collection of 100–200 elements—spherical particles interacting with their immediate neighbours via short-range soft-core potentials. Therefore, ScEM is able to model cells of arbitrary shapes that grow, divide and move in 3D. The main disadvantage of ScEM is that it is computationally expensive (off-lattice methods in general require more computations per time step compared to their lattice counterparts), and without a highly optimised parallel implementation, applications of the ScEM are limited to a few hundred cells at most, which is not sufficient to study effects that span long length- and time-scales.
Particle-based models have also been very successful at capturing many aspects of cell migration in tissues [42–46]. However, when it comes to modelling confluent epithelial layers with the resolution of individual cells, one of the most widely and successfully used models is the Vertex Model (VM). The VM originated in studies of the physics of foams in the 1970s and was first applied to model monolayer cell sheets in the early 1980s [47]. Over the past 35 years it has been implemented and extended numerous times and used to study a wide variety of different systems [48–53]. The VM is at the core of the cell-based [54] CHASTE [55], a versatile and widely used software package. The VM assumes that all cells in the epithelium are roughly the same height and thus that the entire system can be well approximated as a two-dimensional sheet. The conformation of the tissue in the VM is computed as a configuration that simultaneously optimises area and perimeter of all cells. While the model is quasi-static in nature, it captures remarkably well many properties of actual epithelial tissues. There have been numerous attempts to introduce dynamics into the vertex model [47, 51, 53, 56, 57]. However, there are limitations associated with each approach. To the best of our knowledge, most dynamical versions of the vertex model seem to neglect fluctuations with a notable recent exception [53]. In contrast, recent traction microscopy experiments [14] suggest that these fluctuations might be a crucial ingredient in understanding collective cell migration. Finally, we point out a technical point that makes the implementation of VM somewhat challenging. Namely, in order to capture topology changing moves, such as cell neighbour exchanges, i.e., T1 transitions, one has to perform rather complex mesh restructuring operations [51, 58] that require complex data structures and algorithms and that inevitably add to the computational complexity of the model.
Building upon the recently introduced Self-Propelled Voronoi (SPV) model [59], in this paper we apply the ideas introduced in the physics of active matter systems [60] to the VM. This allows us to construct a hybrid Active Vertex Model (AVM) that is able to accurately describe the collective migration dynamics of a large number of cells. The AVM is implemented within SAMoS [61], an off-lattice particle-based simulation software developed specifically to study active matter systems. One of the key advantages of the hybrid approach presented in this study is that it not only enables studies of very large systems, but also introduces a very natural way to handle the T1 transitions, thus removing the need for complex mesh manipulations that are of uncertain physical and biological meaning.
Owing to its origins in the physics of foams [62], in the VM cells are modelled as two-dimensional convex polygons that cover the plane with no holes and overlaps, i.e., the epithelial tissue is represented as a convex polygonal partitioning of the plane (Fig 1). The main simplification compared to models of foams is that most implementations of the VM assume that contacts between neighbouring cells are straight lines. However, we note that there have been several recent studies where this assumption has been removed and cell-cell junctions were allowed to be curved [63, 64]. In addition, neighbouring cells are also assumed to share a single edge, which is a simplification compared to real tissues, where junctions between two neighbouring cells consist of two separate cell membranes that can be independently regulated. Typically, three junction lines meet at a vertex, although vertices with a higher number of contacts are also possible [58]. The model tissue is therefore a mesh consisting of polygons (i.e., cells), edges (i.e., cell junctions), and vertices (i.e., meeting points of three or more cells).
An energy is associated to each configuration of the mesh. It can be written as
E V M = ∑ i = 1 N K i 2 A i - A i 0 2 + ∑ i = 1 N Γ i 2 P i 2 + 2 ∑ μ , ν Λ μ ν l μ ν , (1)
where N is the total number of cells, Ai is the area of the cell i, while A i 0 is its reference area. Ki is the area modulus, i.e. a constant with units of energy per area squared measuring how hard it is to change the cell’s area. Pi is the cell perimeter and Γi (with units of energy per length squared) is the perimeter modulus that determines how hard it is to change perimeter Pi. lμν is the length of the junction between vertices μ and ν and Λμν is the tension of that junction (with units of energy per length). 〈μ, ν〉 in the last term denotes the sum is over all pairs of vertices that share a junction. Note that the model allows for different cells to have different area and perimeter moduli as well as reference areas, allowing for modelling of tissues containing different cell types. Finally, for convenience we introduced a prefactor 2 in the last term in Eq (1) to compensate for double counting of cell junctions when switching from a sum over junctions to a sum over cells in the force calculation discussed below.
Some authors write the perimeter term as ∑ i Γ ˜ i / 2 ( P i - P i 0 ) 2, where P i 0 is a reference perimeter, and omit the last term in Eq (1) or completely omit the P2 term [47, 58]. Under the assumption that the values of Λμν for all junctions of the cell i are the same, i.e., Λμ,ν ≡ Λi, the last term in Eq (1) becomes Λi ∑〈μ,ν〉 lμ,ν = ΛiPi. Therefore, if we identify Λ = - Γ ˜ P i 0, it immediately becomes clear that the descriptions in Eq (1) and the model with the preferred perimeter are identical to each other. Note that this is true up to the constant term 1 / 2 Γ ˜ i ( P i 0 ) 2, which is unimportant as it only shifts the overall energy and does not contribute to the force on the cell (see below). The description in Eq (1) is slightly more general as it allows for the junctions to have different properties depending on the types of cells that are in contact. Retaining the P i 2 term is also advisable in order to prevent the model from becoming unstable if the area modulus is too small.
It is straightforward to express cell area and cell perimeter in terms of vertex coordinates. Therefore, vertex positions together with their connectivities uniquely determine the energy of the epithelial sheet, hence the name Vertex Model. The main assumption of the VM is that the tissue will always be in a configuration which minimises the total energy in Eq (1). Determining the minimum energy configuration is a non-trivial multidimensional optimisation problem and, with the exception of a few very simple cases, it can only be solved numerically. A basic implementation of the VM therefore needs to use advanced multidimensional numerical minimisation algorithms to determine the positions of vertices that minimise Eq (1) for a given set of parameters Ki, Γi and Λμν. Most implementations [51, 53, 58], also introduce topology (connectivity) changing moves to model events such as cell rearrangements.
There have been several attempts to introduce dynamics into the VM [47, 51, 56], including a recent study that introduced stochasticity into the junction tension [53]. The idea behind such approaches it to write equations of motion for each vertex as
γ d r μ d t = F μ , (2)
where γ is a friction coefficient and rμ is the position vector of vertex μ. Fμ is the total force on vertex μ computed as the negative gradient of Eq (1) with respect to rμ, i.e., Fμ = −∇rμ EVM. We note that the exact meaning of friction in confluent epithelial tissues is the subject of an ongoing debate that is beyond the scope of this study. Here, as in the case of most models to date, we assume that all effects of friction (i.e., between neighbouring cells as well as between cells and the substrate and the extracellular matrix) can be modelled by a single constant. While this may appear to be a major simplification, as we will show below, the model is capable of capturing many key features of real epithelial tissues. Eq (2) is a first order equation since the mass terms have been omitted. This so-called overdamped limit is commonly applied to biological systems, since the inertial effects are typically several orders of magnitude smaller than the effects arising from the cell-cell interactions or random fluctuations produced by the environment. Note that the force on vertex μ depends on the position of its neighbouring vertices, resulting in a set of coupled non-linear ordinary differential equations. In most cases those equations can only be solved numerically.
While the introduction of dynamics alleviates some of the problems related to the quasi-static nature of the VM, one still has to implement topology changing moves if the model is to be applicable to describing cell intercalation events. This can lead to unphysical back and forth flips of the same junction and has only recently been analysed in full detail [58].
It is important to note that the VM in its original form is a quasi-static model. In other words, it assumes that at every instant in time, the tissue is in a state of mechanical equilibrium. This is a strong assumption, which is in line with many biological systems, especially in the case of embryos where cells actively grow, divide and rearrange. As a matter of fact, biological systems are among the most common examples of systems out of equilibrium. Therefore, while it is able to capture many of the mechanical properties of the tissue, the VM is unable to fully describe the effects that are inherently related to being out of equilibrium. Many such effects are believed to be behind the collective migration patterns observed in recent experiments. In addition, in many dynamical implementations of the VM the effects of both thermal and non-thermal random fluctuations originating in complex intercellular processes and interactions with the environment are either completely omitted or not very clear. While for a system out of equilibrium the fluctuation-dissipation theorem [65] does not hold, and the relation between random fluctuations and friction is not simple, it is even more important to note that fluctuations can have non-trivial effects on the collective motion patterns [53].
Here we take an alternative approach and build a description similar to the recently introduced SPV model [59]. The idea behind the SPV is that instead of treating vertices as the degrees of freedom, one tracks positions of cell centres. Forces on cell centres are, however, computed from the energy of the VM, Eq (1). The core assumption of the model is that the tissue conformations correspond to the Voronoi tessellations of the plane with cell centres acting as Voronoi seeds. We recall that a Voronoi tessellation is a polygonal tiling of the plane based on distances to a set of points, called seeds. For each seed point there is a corresponding polygon consisting of all points closer to that seed point than to any other. This imposes some restrictions onto possible tissue conformations, i.e., not all convex polygonal tessellations of the plane are necessarily Voronoi, but it has recently been argued that Voronoi tessellations can predict the diverse cell shape distributions of many tissues [66]. Furthermore, the exact details of the tessellation are not expected to play a significant role in the large scale behaviour of the tissue, which this model aims to describe. We, however, note that recently an interesting model based on a generalised Voronoi description has been proposed [67].
In the original implementation of Bi, et al. [59] the Voronoi tessellation is computed at every time step. The vertices of the tessellation are then used to evaluate forces at all cell centres, that are, in turn, moved in accordance to those forces and the entire process is repeated. While conceptually clear, this procedure is numerically expensive as it requires computation of the entire Voronoi diagram at each time step. This limits the accessible system size to several hundred cells.
Here, we instead propose an alternative approach based on the Delaunay triangulation. The Delaunay triangulation for a set of points P in the plane is a triangular tiling, DT (P), of the plane with the property that there are no points of P inside the circumcircle of any of the triangles in DT (P) [68]. A property of a Delaunay triangulation that is key for this work is that it is possible to construct a so-called dual Voronoi tessellation by connecting circumcenters of its triangles. This establishes a mathematical duality between Delaunay and Voronoi descriptions. This duality is exact and quantities, such as the force, expressed on the Voronoi tiling have an exact map onto quantities expressed on its dual Delaunay triangulation. Although being non-linear (see Sec. “Force on the cell centre”), this map is relatively simple, and therefore fast to compute. An important property of the Voronoi-Delaunay duality is that continuous deformations of one map into continuous deformations of the other. In other words, smooth motion of a cell’s centre will correspond to a smooth change in that cell’s shape. This is crucial to ensuring that during the dynamical evolution the cell connectivity changes continuously, a feature that is essential for accurately modelling T1 transitions. The main advantage of working with the Delaunay description is that while the Voronoi tessellation has to be recomputed each time cell centres move, it is possible to retain the Delaunay character of a triangulation via local edge flip moves (Fig 2c), which drastically increases the efficiency of the Delaunay based approach and enables us to simulate systems containing tens of thousands of cells.
Before we introduce the Active Vertex Model (AVM), we pause to make a comment about the notation. In the following, we will always use Latin letters to denote cells, i.e. positions of their centres, and Greek letters to denote vertices of the dual Voronoi tessellation, i.e., meeting points of three or more cells. Therefore, vertices of the VM will always carry Greek indices (Fig 2a).
In order to validate the model and compare it with the results of similar models proposed in the literature, as well as to show its use in modelling actual biological tissues, we now apply the AVM to several problems relevant to the mechanics of epithelial tissue layers.
We start by illustrating one of the key processes observed in epithelial tissues, the T1 transition. As detailed in Fig 2, in the AVM T1 transitions are handled through an edge flip in the Delaunay triangulation. An edge flip only happens when, in the notation of Fig 2c, we have α + β = δ + γ = 180°. Then both triangles are circumscribed by the same circle passing through its combined four vertices. The location of the T1 transition coincides with the centre of this circle. Due to the continuous connection between the position of sites of the Delaunay triangulation and its dual Voronoi tessellation, we always approach this point smoothly, i.e. a junction between two cells will smoothly shrink to a point, the T1 transition will occur, and then it will expand in a new direction. This process arises naturally in the AVM model, in stark contrast with many currently available implementations [85] that require a cut-off criterion on the edge length of a cell before a T1 transition can occur. It also avoids discontinuous jumps at finite edge length, bypassing the T1 point altogether, also a feature of a number of models, notably those based on sequential energy minimisation [49, 81, 86]. Next to its high computational efficiency, the ability to smoothly go through a T1 transition without the need for any additional manipulations of either the Delaunay triangulation or the Voronoi tessellation is one of the key advantages of the AVM approach.
In Fig 5, we illustrate a T1 transition in the bulk, in a region of phase space where the system exhibits liquid-like behaviour, but with very slow dynamics (see next section). The edge linking cells 2 and 4 (in red) slowly shrinks to a point, and then rapidly expands in the opposite direction. This feature points to dynamics akin to certain models of sheared materials [87], where the active driving pulls the material over an energy barrier from one minimum to the next. It is somewhat different from the activated dynamics which has been proposed for the SPV [86], which would predict a series of fluctuations through which the barrier between minima is ultimately crossed.
We now explore different modes of collective behaviour, i.e., phases, of the tissue based on the values of parameters of the original VM (Ki, Γi and Λμ,ν), and AVM-specific parameters such as the activity fa, the orientational correlation time τr, and the boundary line tension λ. In order to keep the number of independent parameters to a minimum, it is again convenient to rewrite the energy of the VM, Eq (1), in a scaled form [49, 59]. We first choose Ki = 1 and set A i 0 = π as an area scale. For simplicity, we assume that all perimeter and junction tensions are the same, i.e., we set Γi ≡ Γ and Λμ,ν ≡ Λ for all i, μ and ν. Then, as discussed below Eq (1), we can complete the square on the second and third terms in Eq (1) and obtain the scaled VM potential
E V M = ∑ i = 1 N 1 2 A i - A 0 2 + ∑ i = 1 N Γ 2 P i - P 0 2 , (20)
where P0 = −Λ/Γ and A0 = π. The first term in Eq (20) penalises changes in the cell area, while the second term penalises changes of the perimeter. There is no reason for the preferred area A0 to be generically compatible with the preferred perimeter P0. This sets up a competition between the two terms in Eq (20), giving a natural scale that is determined by the relative ratio of Γ P 0 2 to K A 0 2. In other words, if K A 0 2 > Γ P 0 2, the cell will try to optimise its area at the expense of paying a penalty for not having the most optimal perimeter, and the opposite if K A 0 2 < Γ P 0 2.
Bi, et al. [59] introduced the dimensionless shape factor p 0 = P 0 A 0, which controls the ratio of the cell’s perimeter to its area through the target area A0 and target perimeter P0. The value of p0 then determines whether the area or the perimeter term in Eq (20) wins and effectively sets the preferred shape of each cell: cells of different shapes have different values of p0. For example, regular hexagons, pentagons, squares and triangles correspond to p0 = 3.722, p0 = 3.812, p0 = 4.0 and p0 = 4.559, respectively.
Remarkably, one observes [49, 59, 86] a transition between a solid-like behaviour of the tissue, where cells do not exchange neighbours, and liquid-like behaviour, where neighbour exchanges do occur, at p0 = 3.812, a value that corresponds to a regular pentagon. At present, the biological significance of this observation is not clear, but it appears to be a robust feature of many experimental systems [88]. In order to make the comparison between the AVM and the SPV model, we also adopt p0 as a main parameter that controls the preferred cell shape.
In order to initialise the simulation, in each run we start by placing soft spheres with slightly polydisperse radii in a circular region. We then use SAMoS to minimise the energy of a soft sphere packing in the presence of a fixed boundary. This ensures that initially, cells are evenly spaced without being on a grid. We also fix the packing fraction to ϕ = 1, ensuring that the average cell area of the initial configuration is 〈A〉 = A0. The boundary is either fixed (referred to as “fixed system”), or allowed to fluctuate freely (“open system”). Fig 6 shows a representative set of the states that we observe. We run the simulation for either 100,000 time steps with step size δt = 0.01 in the unstable region (e.g., Fig 6e), or 250,000 time steps with δt = 0.025 in the solid-like region (e.g., Fig 6c). For these systems with N = 1000 cells in the interior, this takes between 10–40 minutes on a single core of a modern Intel Xeon processor depending on the number of rebuilds of the Delaunay triangulation that are necessary (more in the liquid-like phase).
The unit of time is set by γ/Ka2, where a ≡ 1 is the unit of length. We note that Bi, et al. use a = A 0 as the unit of length. This is possible as long as cells are not allowed to grow, i.e., when A0 does not change in time. The AVM allows for the cells to change their size and therefore we need to choose a different unit of length. In our case, a is the range of the of soft-core repulsion between cell centres (see S1 Appendix, Eq. (35)).
At low values of p0, we find a system that prefers to be in a state with mostly hexagonal cells, unless the active driving fa is very high. Open systems will shrink at this point so that all cells are close to their target P0, as shown in Fig 6a. Consistent with this, larger values of the perimeter modulus Γ lead to stronger shrinking. For fixed systems, this route is blocked, and instead there is a strong inward tension on the boundaries and a gradient in local density, as shown in Fig 6b.
In agreement with the results of Bi, et al. [59], we find that at low p0 < 3.81 and low values of driving fa, cells do not take an organised pattern and do not exchange neighbours. Recast in the language of solid state physics, the tissue is in an amorphous solid or glassy state. In Fig 6c we show such a state for a fixed boundary. In order to characterise the physical properties of this state, we measure the dynamical time scale of cell rearrangements through a standard tool of the physics of glassy systems, the self-intermediate scattering function [15]
F q , t = exp i q · r ( t ) - r ( 0 ) . (21)
F (q, t) measures the decay of the autocorrelation of cell-centre positions r(t) at a particular wave vector, q, taken usually to be the inverse cell size q ≡ |q| = 2π/a. The long-time decay of F (q, t) is characterised by the so-called alpha-relaxation time τα at which F (q, t) has decayed by half. When the system solidifies, i.e. when neighbour exchanges stop, r(t) remains constant and hence τα diverges [15], and stays infinite within the solid phase. In Fig 7a–7c, we show the phase diagram of τα as a function of p0 and fa, for several systems with different boundary conditions.
In Fig 7a we show regions of solid-like and liquid-like phases in a system with fixed boundaries, at Γ = 1, and a low noise value of τr = 0.01. We find a boundary of the solid-like phase that stretches from p0 ≈ 3.81 at small fa to a maximum activity fa beyond which the system is fluid at all p0. This is qualitatively, but not quantitatively consistent with the results of Bi, et al., who find a transition line at roughly twice our fa values. Several factors are likely implicated in this discrepancy. Our systems, at N = 1000 cells, are more than twice as large as the N = 400 systems considered by Bi, et al., and finite system size effects seem to play an important role, as shown below. We measure τα at a value of 1/q corresponding to displacements of one cell size. However, even though displacements are large, we have evidence that this may not be sufficient to induce T1 transitions and therefore fluidise the system. Finally, fixed boundaries were used here and the periodic boundaries of Bi, et al. are likely not strictly equivalent.
The influence of the type of boundary conditions is very significant. In Fig 7b, we show the phase diagram for the same Γ = 1 and τ r - 1 = 0 . 01 as in Fig 7a, except with open boundary conditions and boundary line tension λ = 0. Separately, for Γ = 0.1, we have also confirmed that the value of the boundary line tension does not significantly affect the onset of the solid-like regime. We find a significantly lower maximum fa for the transition, fa = 0.03, a factor of 10 compared to the fixed case. We note that the effect also persists at τ r - 1 = 0 . 1, but is less pronounced. While we do not have a full explanation for this result, we do note that fluctuations of the boundary allow for rearrangements that are otherwise strongly suppressed by the fixed boundary. For example, the system in Fig 6d shows significant boundary fluctuations. It is liquid-like with τα ≈ 10, whereas the equivalent fixed system has τα ≈ 100. In view of the significant role of the boundary, we expect a strong system-size dependence.
At very high p0 and low active driving, we observe a systematic increase of τα (especially visible in Fig 7c). This unexpected result is accompanied by structural changes in the cell patterns that we observe. Fig 6f shows a liquid system at p0 = 3.95, near the relative minimum τα for a given fa. The distribution of cell centres appears random. In contrast, as can be seen in Fig 6g, at very high p0 = 4.85, cells arrange themselves into rosette shapes, where many vertices meet in a point. Rosettes are a feature of many developmental systems [89], so it is interesting to see that they do appear in the AVM context. Cell centres also arrange themselves in equidistant chains, hinting at a connection to one of the various pattern-formation instabilities studied in nonlinear dynamics. We note that this regime is numerically delicate, and the addition of the soft repulsive core between cell centres (see S1 Appendix, Eq. (35)) is necessary to make simulations stable. This simulation is performed deep inside the liquid regime where cell junctions are not necessarily accurately represented by straight lines and the applicability of the model is questionable. At present it is, therefore, not clear whether the configuration shown in Fig 6g is an artefact of the model or it indeed has biological significance. One should proceed with caution when assigning biological interpretation to any configuration with very large values of p0, i.e. for p0 ≳ 4.6.
In parts of the phase diagram, we observe a fingering instability [90–93] where regions a few cells wide migrate outward from the centre, as shown in Fig 6e. When τα drops below approximately 10, we observe that the fluctuations of the boundary already present in Fig 6d become unconstrained. This is a mechanically unstable regime: Eventually, these cells will detach, a process we are not yet able to model due to the topological change that it would imply (see, Fig 3). We have observed that fluctuations need to reach a threshold of approximately >5% of a length increase in the boundary to break through to an unconstrained growth, otherwise the system remains stable, see e.g. Fig 6d–6g. We then associate a time scale τinst with reaching this threshold and use it to measure the degree of instability: a small time scale denotes a rapid growth rate of fluctuations. Fig 7c shows τα, and Fig 7d shows τinst for the same open system with Γ = 0.1, τ r - 1 = 0 . 1 and boundary line tension λ = 0.1. We note that the transition line between solid-like and fluid-like states is low, at fa = 0.03. At and below fa = 0.1, the boundary of the system is stable, and above this threshold, the instability becomes more pronounced with increasing fa and smaller τα. The physical mechanism responsible for the instability involves a subtle interplay of fa, boundary line tension (stronger line tension suppresses the instability), the noise level τ r - 1 (lower τ r - 1 enhances the instability), and p0. The instability resembles observations of finger formation in MDCK monolayers [90, 91]. Existing models link it to either leader cells [93, 94], a bending instability [92], or an active growth feedback loop [95], while here it emerges naturally. A detailed account of this phenomenon will be published elsewhere.
Division and death processes are important in any living tissue, for example, cell division and ingression processes play essential roles during development. Therefore, as noted in Sec. “Cell Growth, Division and Death”, the AVM is equipped to handle such processes. It is important to note, however, that the removal of one cell during apoptosis or ingression and the addition of two new cells during division in the AVM causes a discrete change in the Voronoi tessellation which implies a discontinuous change of the local forces derived from the VM. We have simulated the growth of a small cluster of cells to assess whether this discrete change in geometry can lead to any instabilities in the model. These test runs did not reveal any artefacts due to discontinuities in the force caused by the division events.
In order to illustrate the growth process, we choose a shape factor, p0 = 3.10, corresponding to Γ = 1 and Λ = −5.5 and no active driving, i.e., we set fa = 0. This puts the system into the solid-like phase where T1 events are absent. Our simulation runs for 106 time steps at δt = 0.005, corresponding to 5000 time units, starting from 37 cells and stopping at about 24,000 cells. To balance computational efficiency with a smooth rate of division, cells are checked for division every 25 time steps. We show snapshots of different stages of the tissue growth in Fig 8a–8e.
We note that the numerical stability of the simulation that involves growth is quite sensitive to the values of the parameters used in the AVM. For example, divisions of highly irregularly shaped cells, as commonly observed in the high p0 regime, can put a significant strain on the simulation and even cause a crash. Helpfully, some of these problems are alleviated by the soft repulsive potential between cell centres (see S1 Appendix, Eq. (35)). This in addition to a smaller time step is used to mediate the impact of cell divisions for growing systems with high p0.
In Fig 8f we show the tissue size as a function of the simulation time. In this simulation there are no apoptosis or cell ingression events and, as expected, the tissue size grows exponentially. However, at long times, the growth slows down and deviates from exponential growth. This is easy to understand, as the centre of the tissue is prevented from expanding by the surrounding cells. The effect can be seen in Fig 8e, where cells located towards the centre have, on average, smaller areas and in Fig 8g, which shows a clear pressure buildup in the centre. This suggests that in the later stages, the simulated tissue is not in mechanical equilibrium any more. The pressure is computed as the trace of the Hardy stress tensor, defined as [96]
σ ^ w H = 1 2 ∑ α , β α ≠ β ∫ s = 0 1 - f α β ω ( 1 - s ) r α + s r β - r ⊗ r α - r β d s , (22)
where ω is a smoothing function and the sum is over all junctions and triangulation edges. The reason is that we can determine the shape and area of each cell using just the scalar lengths, {rαβ}, of these two sets of edges. We eliminate the integral by choosing,
ω ( r ) = 1 / A D if r ∈ D 0 otherwise (23)
where D is the region for which the stress is calculated and A D is the area of D. For convenience we choose D to coincide with individual cells. Finally, f α β = - ∂ E V M ∂ r α β r ^ α β, where EVM is defined in Eq (1) and r ^ α β is the unit-length vector along the junction or triangulation edge denoted by (α, β). This particular choice of ω and D reduces our calculation to that of the more common virial stress [96], but we note that the Hardy stress is a useful tool for coarse-graining. A full discussion regarding coarse-graining of stress calculations in the AVM will be published elsewhere. We also see clear heterogeneities in the local pressure shown in Fig 8g. In Fig 8h, we show the radial pressure profile in the tissue at the end of the simulation. From the figure it is also evident that the there is a substantial pressure buildup close to the centre of the tissue as well as that angular averaging substantially reduces local pressure fluctuation notable in Fig 8g. The origin of these effects warrants a detailed investigation and will be addressed in a later publication, we note however that stress inhomogeneities are a persistent feature of the epithelial cell monolayers that have been investigated by traction force microscopy [12, 97]. A similar pressure buildup has also been investigated in a growing cell aggregate [98] and in a mathematical model of nonuniform growth in a layer of tissue [99].
Finally, in Fig 8i we show the distribution of the number of neighbours for this model system. The observed distribution is in qualitative agreement with observations on several tissues both in animals and plants (e.g., Drosophila wing disc, Xenopus tail epidermis, Hydra outer epidermis, Anagallis arvensis meristem, cucumber epidermis) reported in the literature [100, 101]. We note that while the relative values of the percentages of nearest neighbours are dependent on the parameters used in the model, we observed that the overall shape of the distribution is maintained over a several sets of parameters chosen in different regions of the parameter space. A detailed study of this distribution and its precise dependence on the model parameters is, however, beyond the scope of this study.
The AVM is equipped to allow for cell-specific parameters, which enables us to investigate tissues with locally varying mechanical properties. A commonly studied example of the effects such heterogeneities is cell sorting. As an example we show simulations that display sorting of two distinct cell types. We achieve this by setting the junction tension Λ for each pair of cell-cell and cell-boundary contacts. All our simulations consist of 1000 cells with half chosen randomly to be of the “red” type and the others being of the “blue” type. In these simulations, boundaries have been kept fixed. We observe sorting behaviour akin to that found in other commonly used tissue models [35, 102]. Using r, b and M to denote red, blue and the boundary respectively, we start by fixing K = 1 and Γ = 1 and set −6.8 = Λrr < Λrb < Λbb = −6.2, corresponding to p0 in the range 3.58–3.93. We, however, note that p0 is a quantity defined “per cell” and one should understand it in this context only as a rough estimate whether a given cell type is in the solid or fluid phase. All cells are subject to small random fluctuations of their position which allows for T1 transitions that can bring initially distant cells into contact. We set the active driving, fa, to zero. We have chosen different values for Λrr and Λbb to reflect the idea that the surfaces of these cells have different adhesive properties [103]. Note that the Λ parameter for a particular contact is proportional to its energy per unit length. Sorting of cells into groups of the same type occurs when the energy of two red-blue contacts is greater than the energy of one red-red contact and one blue-blue contact, corresponding to Λrb > (Λrr + Λbb)/2, see Fig 9a–9c. In this regime, for cells of the same type it is energetically favourable for the new contact to elongate while local red-blue contacts are shortened. Conversely, if Λrb < (Λrr + Λbb)/2, then cells maximise their red-blue contacts forming a “checkerboard” pattern (Fig 9d). The final pattern is not without defects, the number and location of which depend on the initial conditions. The tissue boundary consists of contacts between cell centres and boundary particles so ΛrM and ΛbM need also to be specified to reflect the way in which the cell types interact with the extracellular matrix or surrounding medium. Initially we set ΛrM = ΛbM = −6.2 and observe that blue cells cover the boundary enveloping red cells because this facilitates lower energy red-blue and red-red contacts being formed. If we incentivise red-boundary contacts by setting ΛrM < Λrr + ΛbM − Λrb = −6.6 we make red-boundary contacts preferable [102]. This case is shown in Fig 9e for ΛrM = −6.8.
Finally, we note that all simulations for mechanically heterogeneous systems were performed without any active driving. While a detailed study of the effects of active driving on cell sorting is beyond the scope of this study, we note that a limited set of test runs suggest, as expected, that including activity can reduce the time over which cells sort but that very high activity leads to mixing.
We now briefly turn our attention to the effects of several models of cell polarity alignment. So far, we have assumed torque τi = 0 in Eq (14), i.e., we are in the situation where the the polarisation vector of each cell is independent of the surrounding cells and its direction diffuses randomly over time. In biological systems, it is known that a cell’s polarity responds to the surrounding and many forms of polarity alignment have been proposed. Here, we highlight two alignment mechanisms that are compatible with the current understanding of cell mechanics. In Fig 10a–10c, we have used the alignment model defined in Eq (9) that assumes that the polarity vector ni of cell i aligns with its velocity vi. The torque term in Eq (14) is then given by τi = −Jv vi × ni, where Jv is the alignment strength. This model was first developed for collectively migrating cells (modelled as particles) [74], and it exhibits global polar migration, i.e. a state in which all particles align their velocities and travel as a flock. In dense systems of active particles confined to a finite region, velocity alignment has been shown to be intimately linked to collective elastic oscillations [75]. It is remarkable that the main hallmarks of this active matter dynamics are also observed in the model tissue. In Fig 10a, we show velocity alignment dynamics for a confined system in the solid-like phase; here the collective oscillations are very apparent. They are strikingly reminiscent of the collective displacement modes observed in confined MDCK cell layers [104, 105].
In Fig 10b, we apply the same dynamics, but now to a system that is in the liquid-like phase, with fixed boundaries. Here, the collective migration wins, but the confinement to a disk with fixed boundaries forces the cells into a vortex-shaped migration pattern. Finally, in Fig 10c, in the absence of confinement, we recover the collective polar migration of the cell patch.
In Fig 10d, we show the effects of aligning the cell’s polarity to the largest principal axis of the cell shape tensor, defined in Eq (10). This type of alignment also leads to collective motion in an unconfined system, however there are significant fluctuations as the allowed cell patterns are highly frustrated by the constraint to remain in a Voronoi tesselation.
These preliminary results serve as a showcase of the non-trivial effects of cell-cell alignment on the collective behaviour of the entire tissue. A more detailed account of the effects of different alignment models will be published elsewhere.
In the previous discussions, all examples assumed a patch of cells with the topology of a disk. However, the AVM is not restricted to the circular geometry and can be applied to systems of arbitrary shapes, including domains with complex connectivity. Such situations often arise when modelling experimental systems where cells surround an obstacle, or in studies of wound healing. In Fig 11 we present a gallery of non-circular shapes that can be readily studied using the AVM as it is implemented in SAMoS. The annular geometry shown in Fig 11a would be suited for modelling wound healing problems as well as situations where cells migrate in order to fill a void. A common experimental setup where cell colonies are prepared as rectangular strips [90] is shown in Fig 11b, where three separate patches grow towards each other. Finally, in Fig 11c we show an example of yet another very interesting situation [106], where cells are grown in a confined region of space.
In this paper we have introduced the Active Vertex Model. It is a hybrid model that combines ideas from the physics of active matter with the Vertex Model, a widely used description for modelling confluent epithelial tissues. Active matter physics is a rapidly growing field of research in soft condensed matter physics, and it is emerging as a natural framework for describing many biophysical processes, in particular those that occur at mesoscales, i.e., at the scales that span multiple cells to the entire tissue. Our approach is complementary to the recently introduced Self-Propelled Voronoi model [59], for it allows modelling of systems with fixed and open, i.e. dynamically changing boundaries as well as cell-cell alignment, cell growth, division and death. The AVM has been implemented into the SAMoS software package and is publicly available under a non-restrictive open source license.
The AVM utilises a mathematical duality between Delaunay and Voronoi tessellation in order to relate forces on cell centres to the positions of the vertices of the dual lattice, i.e. meeting points of three of more cells—a natural description of a confluent epithelial tissue. This not only allows for a straightforward and efficient implementation using standard algorithms for particle-based simulations, but provides a natural framework for modelling topological changes in the tissue, such as intercalation and ingression events. In other words, in the AVM T1 transition events arise spontaneously and it is not necessary to perform any additional steps in order to ensure that cells are able to exchange neighbours.
Furthermore, our implementation of the AVM is very efficient, allowing for simulations of systems containing tens of thousands of cells on a single CPU core, thus enabling one to probe collective features, such as global cell flow patterns that span length-scales of several millimetres. In addition, the AVM is also able to handle multiple cell types and type specific cell contacts, which allows simulations of mechanically heterogeneous systems.
All these features make the AVM a strong candidate model to address many interesting biological and biophysical problems related to the mechanical response of epithelial tissues, especially those that occur at large length and time scales that are typically only accessible to continuum models. Unlike in the case of continuum models where relating parameters of the model to the experimental systems is often difficult and unclear, the AVM retains the cell-level resolution, making it simpler to connect it to the processes that occur at scales of single cells. The AVM is, however, not designed to replace continuum models, but to serve as the important layer that connects the complex molecular processes that occur at the cell level with the global collective behaviour observed at the level of the entire tissue.
A natural question to ask is: What is the advantage of the AVM compared to other particle based models, e.g. [43, 44], that are computationally far less demanding? There is clearly a tradeoff between computational costs and the required level of details necessary to address a specific biological question. It is, however, not a priori clear where the right balance between the two lies. The key advantage of the AVM compared to other cell-centre based models is that keeping track of both cell centres and cells themselves allows for straightforward extensions of the model to include effects such as active forces on vertices, active non-linear response of cell junctions, etc. that would be impossible to implement and parametrise in purely cell-centre based models.
With that in mind, there are, of course, still many ways the model can be improved. For examples, it would be very interesting to augment the AVM to include the effects of biochemical signalling. This would require solving a set of differential equations for signals in each time step, and then supplying those solutions to the mechanics part of the model. Adding such functionality would substantially increase the computational cost of simulations, however at the same time it would allow for detailed studies of the coupling between chemical and mechanical signalling. These are believed to be essential for developing a full understanding of the mechanics of epithelial tissues. Given the layout of the AVM and its implementation, implementing such functionality would be straightforward.
Furthermore, in the current version of the AVM activity is introduced in a very rudimentary manner, via assuming that cells self-propel in the direction of their polarity vector. This is a very strong assumption that would need a much stronger experimental support than currently available. It is also possible that a far more sophisticated model would be required to fully capture cell motility. However, one also needs to keep in mind that there is a tradeoff between being as biologically accurate as possible and retaining a sufficient level of simplicity to be able to efficiently perform simulations of large systems. With all this in mind, we argue that the our simulations clearly show that even this simplistic model of self-propulsion is capable of capturing many features of real systems, and that it can serve as a good starting point for building biologically more accurate descriptions.
Another potentially very interesting feature that is currently not supported would be splitting and merging of the boundaries, that is, allowing for topological changes of the entire sheet such as those depicted in Fig 3c. This would allow us to study detachment of a part of the tissue or opening apertures as well as the opposite problem of closing holes and gaps. The latter is of great importance for studying problems related to wound healing. Unfortunately, setting up a set of general rules on how to automatically split a boundary line or merge two boundaries into a single one is not a simple task from a point of view of computational geometry. The problem is further complicated if those rules are also to be made biologically plausible, which is essential for the model to be relevant to actual experiments.
We conclude by noting that recently there have been several interesting attempts at extending the Vertex Model description to three dimensions [85, 107–112]. While in principle possible, extending the AVM to a curved surface or making it fully three-dimensional would be quite involved. Being able to study curved epithelial sheets would be of great interest to systems where curvature clearly cannot be ignored, e.g., as is the case in modelling intestinal crypts [113–116]. While there is nothing in the description of the AVM that is unique to the planar geometry, there are several technical challenges associated with directly porting it onto a curved surface. Most notably, building a Delaunay triangulation on an arbitrary curved surface is not a simple task. In addition, quantities such as bending rigidity that are naturally defined on triangles would have to be properly mapped onto contacts lines between neighbouring cells. This is not straightforward to do. Developing a fully three-dimensional version of the AVM would be an even a greater challenge since the duality between Delaunay and Voronoi descriptions central to this model has no analogue in three dimensions.
We hope that the AVM will provide a useful and complementary tool for probing the aspects of the epithelial tissue mechanics that are not available to other methods, as well as serve as an independent validation for the results obtained by other methods.
|
10.1371/journal.pcbi.1003598 | Inference of RNA Polymerase II Transcription Dynamics from Chromatin Immunoprecipitation Time Course Data | Gene transcription mediated by RNA polymerase II (pol-II) is a key step in gene expression. The dynamics of pol-II moving along the transcribed region influence the rate and timing of gene expression. In this work, we present a probabilistic model of transcription dynamics which is fitted to pol-II occupancy time course data measured using ChIP-Seq. The model can be used to estimate transcription speed and to infer the temporal pol-II activity profile at the gene promoter. Model parameters are estimated using either maximum likelihood estimation or via Bayesian inference using Markov chain Monte Carlo sampling. The Bayesian approach provides confidence intervals for parameter estimates and allows the use of priors that capture domain knowledge, e.g. the expected range of transcription speeds, based on previous experiments. The model describes the movement of pol-II down the gene body and can be used to identify the time of induction for transcriptionally engaged genes. By clustering the inferred promoter activity time profiles, we are able to determine which genes respond quickly to stimuli and group genes that share activity profiles and may therefore be co-regulated. We apply our methodology to biological data obtained using ChIP-seq to measure pol-II occupancy genome-wide when MCF-7 human breast cancer cells are treated with estradiol (E2). The transcription speeds we obtain agree with those obtained previously for smaller numbers of genes with the advantage that our approach can be applied genome-wide. We validate the biological significance of the pol-II promoter activity clusters by investigating cluster-specific transcription factor binding patterns and determining canonical pathway enrichment. We find that rapidly induced genes are enriched for both estrogen receptor alpha (ER) and FOXA1 binding in their proximal promoter regions.
| Cells express proteins in response to changes in their environment so as to maintain normal function. An initial step in the expression of proteins is transcription, which is mediated by RNA polymerase II (pol-II). To understand changes in transcription arising due to stimuli it is useful to model the dynamics of transcription. We present a probabilistic model of pol-II transcription dynamics that can be used to compute RNA transcription speed and infer the temporal pol-II activity at the gene promoter. The inferred promoter activity profile is used to determine genes that are responding in a coordinated manner to stimuli and are therefore potentially co-regulated. Model parameters are inferred using data from high-throughput sequencing assays, such as ChIP-Seq and GRO-Seq, and can therefore be applied genome-wide in an unbiased manner. We apply the method to pol-II ChIP-Seq time course data from breast cancer cells stimulated by estradiol in order to uncover the dynamics of early response genes in this system.
| Transcription mediated by RNA polymerase II (pol-II) is an essential process in the expression of protein-coding genes in eukaryotes. Transcription is dependent upon a number of sequential and dynamic events, such as recruitment of pol-II to the transcriptional start site, activation of pol-II through phosphorylation of its C-terminal domain, elongation of the nascent transcript through the transcribed region and termination [1]. Each of these steps may be rate-limiting and can therefore affect the level of gene expression. In this paper, we describe a simple probabilistic model of transcription whose parameters can be inferred using time-series data such as pol-II ChIP-Seq data [2] or nascent transcript measurement by GRO-Seq that reports markers of transcriptional activity [3]. This model can be used to identify transcriptionally engaged genes, estimate their transcription rates and infer transcriptional activity adjacent to the promoter. The transcriptional dynamics of estrogen responsive genes in a breast cancer cell line were described by fitting this model to pol-II ChIP-seq time course datasets.
Chromatin immunoprecipitation, in conjunction with massively parallel sequencing (ChIP-seq) evaluates interactions between proteins and DNA, and, for example, can be used to monitor the presence of pol-II on DNA. Estimating the amount of pol-II associated with a transcribed gene provides a measure of transcriptional activity [2]. Sequential measurement of pol-II occupancy on genes released from transcriptional blockade, for example, in response to stimuli, reveal a wave of transcription moving through the body of the responding transcript.
A number of studies have attempted to determine the rate of transcription through modelling the dynamics of pol-II. Darzacq et al. fit a mechanistic model of pol-II transcription to nascent RNA data at a single locus and obtained a transcription speed of 4.3 kilobases per minute [4]. Wada et al. activated transcription of genes greater than 100 kbp in length and estimated the transcription speeds using a model that measures an intronic RNA signal through taking advantage of co-transcriptional splicing. They obtain an average transcription rate of 3.1 kbp min−1 [5]. Singh and Padget (2009) reversibly inhibit transcription to determine the transcription rate of 9 genes, all of which were greater than 100 kbp which had an average transcription rate of 3.79 kbp min−1 [6]. The data used in these studies have good temporal resolution (e.g. samples every 7.5 min in [5]) and reliably allow fitting of mathematical models or the direct measurement of transcription speed, however, only for a limited set of long genes. In contrast, high throughput data sets such as ChIP-Seq, can be used to uncover transcription dynamics genome-wide but typically have much lower temporal resolution, motivating the development of alternative modelling approaches that report genome-wide transcription rates.
One way around the low temporal resolution of typical high-throughput time course data is to employ a non-parametric model of the biological signals of interest. In many cases we expect these signals to vary continuously and smoothly in time, when averaged over a cell population, and a Gaussian process model provides a convenient non-parametric model in such cases [7]. Gaussian processes have recently found applications in a range of biological system models [8]–[11].
Here we present a Gaussian process model of transcription dynamics which can be fitted to genome-wide pol-II occupancy data measured using ChIP-Seq. The model describes the movement of pol-II through the gene body and combines a flexible model of promoter-proximal pol-II activity with a reliable estimate of transcription speed. By identifying genes which fit the model well, we provide a useful method to identify actively transcribed genes. The model does not assume a constant transcription speed and can therefore identify variable rates of transcription, for example due to transcriptional pausing. Model parameters are inferred using either maximum likelihood (ML) estimation or via Bayesian inference using Markov chain Monte Carlo (MCMC) sampling. The Bayesian approach provides confidence intervals for parameter estimates and can incoporate priors that capture domain knowledge, e.g. the expected range of transcription speeds, based on previous experiments.
We fit our model to a pol-II ChIP-Seq time course dataset from MCF7 breast cancer cells stimulated with estradiol. The model is used to identify the set of transcriptionally engaged genes and estimate their mean transcription rate and transcriptional activity near the promoter. By clustering promoter activity profiles, potential co-regulated groups of genes are identified, particularly those that respond rapidly to estrogen signalling. Subsequent characterisation of transcription factor (TF) binding sites in proximity to the promoters of genes within clusters provides a means of classifying groups of promoters that are responsive to the binding of specific combinations of TFs. Additionally, publically available ChIP-Seq datasets of TF profiles from the same system were used to identify cluster-specific patterns in TF-binding. The rates of transcription estimated by our model are consistent with the literature [4], [5] but with the advantage that our method allows the computation of transcription speeds genome-wide.
Our methodology has a number of advantages. We do not require data with high temporal resolution, making it feasible to model transcriptional dynamics genome-wide using ChIP-Seq or GRO-Seq time course data. We infer transcription rates for all genes in an unbiased manner and by using Bayesian parameter estimation we are able to associate our transcription rate estimates with confidence intervals. Our model is non-parametric and therefore does not make very strong assumptions about the temporal changes in transcriptional activity. Fitting the model genome-wide allows us to identify and filter out transcripts where pol-II does not travel down the gene body. This provides a principled method to identify responsive genes, in particular, early acting estrogen responsive genes in the specific application considered here. Since our model does not enforce a uniform transcription speed over the entire gene body, we can take into account phenomena such as pol-II pausing which would result in a non-uniform transcription speed. We also use this model to infer the promoter activity of transcriptionally engaged genes, to identify co-regulated gene modules downstream of estrogen signalling.
Visualizing pol-II ChIP-seq reads mapped to transcriptional units at multiple time points following the addition of estradiol to MCF7 cells reveals the motion of pol-II through the gene body of estrogen responsive genes (see Figure 1). Computing the average pol-II occupancy over successive gene segments describes the motion of the transcription wave. Thereafter, fitting a model capable of smoothly interpolating between observed time points and by determining the time taken for pol-II to move from one gene segment to the next determines if pol-II is transcriptionally engaged on a given transcript and the speed at which it is moving through this transcriptional unit. We use a convolved Gaussian process to model the relationship between the pol-II signal at different regions of the gene and across time. Model parameters are determined using maximum likelihood (ML) or Bayesian inference via Markov chain Monte Carlo (MCMC) to determine genes of interest and moreover, in the case of MCMC, determine confidence intervals for our parameter estimates.
A Gaussian process (GP) is a distribution over the space of functions. This distribution is completely specified by a mean function and a covariance function . A function is said to be drawn from a Gaussian process if at any finite collection of points has a multivariate Gaussian distribution with mean vector and covariance matrix specified by and , respectively. GPs provide a powerful framework for non-parametric regression [7]. If a function is assumed to be drawn from a GP with known mean and covariance function, we can infer the function value and associated uncertainty at unobserved locations given noise-corrupted observations. GPs have recently been applied in modelling biological systems, e.g. modelling protein concentrations as latent variables in differential equation models of transcriptional regulation [8], [9] and modelling spatial gene expression [11].
Here we introduce a novel application of GPs to modelling the spatio-temporal dynamics of pol-II occupancy during transcription. Convolved GPs allow the modelling of correlations between multiple coupled data sources. In our case these data sources are the pol-II occupancy over time collected at different locations along the transcribed region of a gene. Modelling the data as a convolved process borrows information from these different data sources in estimating the model parameters and inferring the underlying signal in the data. Also, we find that convolved GPs are necessary to account for changes in the shapes of signals observed at different regions of the gene. In linear systems theory, the output of a linear time-invariant system whose impulse response is is given by the convolution of the input and , that is . If different sets of observations are believed to be related, they can be modeled as the outputs of different linear systems in response to a single input. If this input is modeled as a GP, then it will form a joint GP together with all the outputs and data from one output stream will be useful in inferring the rest [12]–[20]. In our case, incorporating the data from multiple spatially separated regions of the genes allows us to infer an underlying function that links all these regions. This proves useful as a summary of the transcription dynamics of the gene and we show that it provides useful insights into potential coregulation.
A key component of our method involves the estimation of delay between time series observed at different segments of the gene. The study of time delay between related time series has received attention from a number of researchers for a long time [22]. The application areas range from signal processing to astronomy [23]. The classic approach to time delay estimation involves computing the cross-correlation between the related time series and determining the value of delay for which this function is maximised. Consider two signals and given by(9)where and are uncorrelated noise processes. The cross-correlation function is given by where denotes the expectation operator. The value of that maximises yields an estimate of the delay . When the signals are sampled at equally spaced time points with spacing between samples, the discrete time equivalent of is readily estimated. Let , the discrete cross-correlation is estimated as
The delay is estimated by finding the value of for which is maximised. The corresponding delay estimate is . However, this approach does not work well when the time series are unevenly sampled as is the case in several astronomical and biological studies. A number of techniques have been developed to handle unevenly sampled time series including the discrete correlation function (DCF) [24], and the more recent kernel based approaches [25], [26]. The DCF is computed as follows, for all the time differences are binned into discrete bins of size . The DCF at is given by [24], [25](10)where(11)and and are the variances of the observation streams while and are observation error variances.
In the kernel based approach of [25], the underlying function of equation (Equation 9) is modelled as the sum of a fixed number of kernels centered at the observation times. That is(12)where(13)
The value of that minimises the estimation error is the delay estimate. Our implementation follows that presented in [25] where we assumed a fixed kernel width. This kernel width is determined by leave one out cross-validation.
We used synthetic data and previously published experimental data to assess our novel method's performance. To generate the synthetic data, the underlying function of equation (Equation 9) was given as a sum of Gaussian kernels. That is
N was fixed at 20 and the observation interval . , and were generated at random with , and . A random delay was used to generate the observations which were corrupted by additive Gaussian noise with . To determine the effect of number of observations on the quality of inference we compute the Median Normalised Square Error (MNSE) of the estimated delay as a function of the number of observations for 50 random realisations of the the signals. We also investigated the effect of distorting the shape of the observed signals by introducing convolution. In real signals the restriction that the shape remains unchanged sometimes leads to poor results. The parameters of the smoothing kernel in equation (1) were generated at random with and .
To assess performance of our method on a well characterised real-world dataset we obtained a dataset from Singh and Padgett [6] where the delay in appearance of pre-mRNA signal at exon-intron junctions was used to compute estimates of transcription speed for 9 genes. To generate the data, transcription was reversibly inhibited in vivo using 5,6-dichlorobenzimidazole 1-beta-D-ribofuranoside (DRB) and the pre-mRNA measured after the inhibitor was removed. As verified by the authors, the kinetics of pol-II and pre-mRNA are similar hence we expect good performance on this dataset to indicate applicability of our method to pol-II ChIP-seq data.
To demonstrate an application to pol-II ChIP-Seq data, we apply our model to investigate the transcriptional response to Estrogen Receptor signalling. ChIP-seq was used to measure pol-II occupancy genome-wide when MCF-7 breast cancer cells are treated with estradiol (E2). Cells were put in estradiol free media for three days. This is defined media devoid of phenol red (which is estrogenic) containing 2% charcoal stripped foetal calf serum. The charcoal absorbs estradiol but not other essential serum components, such as growth factors. This results in basal levels of transcription from E2 dependent genes. The cells are then incubated with E2 containing media, which results in the stimulation of estrogen responsive genes. The measurements were taken at logarithmically spaced time points 0, 5, 10, 20, …, 320 minutes after E2 stimulation.
Raw reads were mapped onto the human genome reference sequence (NCBI_build37) using the Genomatix Mining Station (software version 3.2.1). The mapping software on the Mining Station is an index based mapper that uses a shortest unique subword index generated from the reference sequence to identify possible read positions. A subsequent alignment step is then used to get the highest-scoring match(es) according to the parameters used. We used a minimum alignment quality threshold of 92% for mapping and trimmed 2 basepairs from the ends of the reads to account for deterioration in read quality at the 3′ end. The software generates separate output files for uniquely mapped reads and reads that have multiple matches with equal score. We only used the uniquely mapped reads. On average about 66% of all reads could be mapped uniquely. The data are available from the NCBI Gene Expression Omnibus under accession number GSE44800.
Time series of pol-II occupancy over various segments of genes were computed in reads per million (RPM) [27] using BEDtools [28], [29]. The genes were divided into 200 bp bins and the RPM computed for each bin. The occupancy in a particular gene segment was the mean RPM of the bins in that segment. Here, the gene is divided into five segments each representing 20% of the gene.
We first applied our methodology to synthetic data in order to compare its performance to other methods. We investigated the performance of five methods, namely cross-correlation (Corr), DCF, the kernel approach of [25] (Kern), a GP approach with no convolution (GP-NoConv), and the convolved GP approach developed in this paper (GP-Conv). Tables 1 and 2 show the MNSE for the different delay estimation methods as a function of the number of observations for synthetic data without convolution and with convolution respectively. Note that the kernel and DCF methods require an estimate of the noise variance and in this simulation study we provide the algorithms with the true value, but that would not be known in practice. We see that when no convolution is introduced, the kernel method performs well but is outperfomed by both GP methods. When convolution is introduced the kernel method appears to break down and as expected the GP-Conv outperforms the other techniques.
We next applied the model to pre-mRNA data from Singh and Padgett [6] where the delay in appearance of pre-mRNA signal at exon-intron junctions was used to compute estimates of transcription speed for 9 genes. Figure 3(A) shows the pre-mRNA signal for the SLC9A9 gene (the same data shown in Figure 4d of [6]). The delays read from these plots were used in [6] to determine transcription speeds. Figure 3 (B–D) shows the fit obtained using the kernel method, GP-NoConv and GP-Conv respectively. Table 3 shows the delays read off the plots as well as values obtained using the five delay estimation algorithms for different regions of the nine genes presented in [6]. In each row the delay estimate with the lowest normalised square error is highlighted. Table 4 shows the MNSE for the five delay estimation algorithms for all the genes. We see that the convolved GP method developed in this paper outperforms the other techniques. This method has the added advantage of inferring a latent function which links all the observations and which can be used for downstream analysis. Also, when analysis is genome-wide, reading delays off individual plots is not feasible and furthermore when the sampling intervals are irregularly spaced assigning delays manually would be error prone. These results serve to justify the use of the convolved GP method introduced in this paper.
We applied our method to a ChIP-Seq time-course dataset measuring pol-II occupancy genome-wide when MCF-7 cells are treated with estradiol (E2). For our initial experiment, we considered 3,064 genes which exhibit significant increase of pol-II occupancy between 0 and 40 minutes after E2 treatment. These genes were determined by counting the number of pol-II tags on the annotated genes in the RefSeq hg19 assembly at 0 and 40 minutes after E2 treatment and computing the ratio of these counts. We keep those genes where this quantity is greater than one standard deviation above the mean. For these 3,064 genes, we filtered out genes less than 1000 bp in length and computed model fits using the ChIP-seq time series data for the remaining 2623 genes. The estimation of the parameters for a given gene was performed using maximum likelihood with fixed at zero, and the values constrained to be equal. Intuitively, one would expect the values of delay to be non-decreasing. We therefore keep only those genes where this natural ordering is preserved for further analysis. We also discard genes with and since these are generally seen to be poor fits. Small values of arise when the data is best modelled as a noise process while large values model constant profiles which are not interesting in our analysis. This left us with 383 genes which we consider a conservative set of genes where there is evidence of engaged transcription and where the model parameters can be confidently estimated. To rank these genes we compared the log marginal likelihood of the model fit to that obtained if we assume independence between the segments, which is equivalent to setting the off-diagonal blocks in equation (8) to the zero matrix.
Figure 2 (A–F) shows the inferred pol-II time profile and histogram of the samples of the delay parameters for three of the top 10 genes found to fit the model well. We note that a relatively small number of activated genes fit the model well. This is primarily because for shorter genes the pol-II occupancy quickly rises over the whole gene such that the temporal resolution of the data cannot capture the wave as it traverses the gene body. With a closer or more evenly spaced time course we would expect a good fit for a greater proportion of activated genes.
Figure 4 (A) shows the linear regression plots using the delay samples for the TIPARP gene. Figure 4 (B) shows the histogram of speed samples from which we can compute the confidence interval for the speed estimate. The 95% confidence interval is indicated in Figure 4 (B) by the red triangle markers (cf. Table 5). Table 5 shows the average transcription speeds for the top 10 genes computed using the samples of the delay parameters. Figure 5 shows a box plot of the average transcription speeds computed using the samples of the delay parameters for these genes.
The advantage of fitting each of the delay parameters independently instead of enforcing a linear relationship is that it allows us to take into account phenomena such as pol-II pausing and provides a means to filter genes where the values of estimated delay are not naturally ordered. Visual inspection of the inferred time series of the top ranked genes is consistent with a ‘transcription wave’ traversing the gene. The transcription wave is especially evident in the longer genes MYH9 and RAB10. This motivates a closer look at long genes. Table 6 shows the average transcription speeds computed using the samples of the delay parameters for the 23 long genes found to fit the pol-II dynamics model well. Grouping these genes according to the magnitude of the median transcription speed allows us to compare our results to those presented previously. From Table 6 we see that 12 (52%) of these genes have average transcription speeds between 2 and 4 kb per minute, a range that includes speeds previously reported in the literature [5], [6].
In this work we have presented a methodology for modelling transcription dynamics and employed it to determine the transcriptional response of breast cancer cells to estradiol. To capture the movement of pol-II down the gene body, we model the observed pol-II occupancy time profiles over different gene segments as the delayed response of linear systems to the same input. The input is assumed to be drawn from a Gaussian process which models the pol-II activity adjacent to the gene promoter. Given observations from high-throughput data such as pol-II ChIP-Seq data, we are able to infer this input function and estimate the pol-II activity at the promoter. This allows us to differentiate transcriptionally engaged pol-II from pol-II paused at the promoter and yields good estimates of transcriptional activity.
In addition to estimating the transcriptional activity at the promoter, inferring the pol-II occupancy time profiles over different gene segments allows us to compute the transcription speed. We expect the delay parameters of different gene segments to be non-decreasing and this provides a natural way to determine genes that are being actively transcribed in response to E2.
Clustering the inferred promoter activity profiles allows us to investigate the nature of the response and group genes that are likely to be co-regulated. We found that the four clusters significantly enriched for both ER and FOXA1 binding within 40 kb according to public ChIP-Seq data were those that showed the earliest peak in pol-II activity at the promoter. ER and FOXA1 ChIP peaks in the neighbourhood of these genes were also more likely to be overlapping than the average for ChIP-identified binding events of these TFs genome-wide. This observation provides some support for the previously proposed role of FOXA1 as a mediator of early transcriptional response in estrogen signalling. These results also show that our method can help regulatory network inference. The inferred promoter activity profiles pinpoint the times of transcriptional activation very accurately without confounding transcriptional delays. As genes with similar inferred promoter activity profiles are likely to have similar TF binding profiles, they are likely to be co-regulated as well. The promoter profiles should therefore lead to more accurate predictions of regulator-target relationships using time-course-based methods (e.g. [9]) than using expression time course data.
As well as modelling transcriptional speed and transcriptional activity profiles, the proposed modelling approach may have other useful applications. For example, recent research has uncovered a link between transcription dynamics and alternative splicing [41]. It is believed that aberrant splicing can cause disease and a number of studies have tried to understand the mechanisms of alternative splicing [42]. The proposed model can potentially be used to identify transcriptional pausing events, and such results could be usefully combined with inference of splice variation from RNA-Seq datasets from the same system. Also, with the increasing availability of high-throughput sequencing data exploring multiple layered views of the transcription process and its regulation, the convolved modelling approach developed here has the potential to be usefully applied to more complex coupled spatio-temporal datasets.
|
10.1371/journal.pntd.0002175 | Dual African Origins of Global Aedes aegypti s.l. Populations Revealed by Mitochondrial DNA | Aedes aegypti is the primary global vector to humans of yellow fever and dengue flaviviruses. Over the past 50 years, many population genetic studies have documented large genetic differences among global populations of this species. These studies initially used morphological polymorphisms, followed later by allozymes, and most recently various molecular genetic markers including microsatellites and mitochondrial markers. In particular, since 2000, fourteen publications and four unpublished datasets have used sequence data from the NADH dehydrogenase subunit 4 mitochondrial gene to compare Ae. aegypti collections and collectively 95 unique mtDNA haplotypes have been found. Phylogenetic analyses in these many studies consistently resolved two clades but no comprehensive study of mtDNA haplotypes have been made in Africa, the continent in which the species originated.
ND4 haplotypes were sequenced in 426 Ae. aegypti s.l. from Senegal, West Africa and Kenya, East Africa. In Senegal 15 and in Kenya 7 new haplotypes were discovered. When added to the 95 published haplotypes and including 6 African Aedes species as outgroups, phylogenetic analyses showed that all but one Senegal haplotype occurred in a basal clade while most East African haplotypes occurred in a second clade arising from the basal clade. Globally distributed haplotypes occurred in both clades demonstrating that populations outside Africa consist of mixtures of mosquitoes from both clades.
Populations of Ae. aegypti outside Africa consist of mosquitoes arising from one of two ancestral clades. One clade is basal and primarily associated with West Africa while the second arises from the first and contains primarily mosquitoes from East Africa
| The authors are all medical entomologists who have worked in the field for more than 30 years. Over the past 20 years we have primarily worked on Aedes aegypti, the primary mosquito vector of Dengue and Yellow Fever Viruses. Twelve years ago, we began using mitochondrial markers to study relationships among Ae. aegypti populations. Since that time, 14 publications and 4 datasets have used the same markers and collectively 95 unique mtDNA haplotypes have been found. Haplotype phylogenies have consistently identified two clades. However, it wasn't until we combined efforts with our African colleagues that we realized that the two clades largely correspond with West and East Africa. Aedes aegypti populations from throughout the world are “mixtures” of mosquitoes from these two original clades. We plan to continue this effort to determine whether the composition of Ae. aegypti populations affects their ability to transmit arboviruses and also if mitochondrial haplotypes differ between mosquitoes with or without the newly discovered chromosomal inversions.
| Aedes aegypti, the ‘yellow fever mosquito’, is the primary vector to humans of the four serotypes of dengue flaviviruses (DENV1-4) and the yellow fever flavivirus (YFV). Dengue is a major public health problem in the tropics, causing millions of dengue fever and hundreds of thousands of dengue hemorrhagic fever cases annually [1]. In endemic areas the annual number of cases has risen steeply since the 1950s [2]. With multiple serotypes circulating in endemic areas, 100 million infections of dengue fever (DF) occur annually, including up to 500,000 cases of the more severe form of disease called dengue hemorrhagic fever (DHF) with a case fatality rate of up to 5% [3]. Despite the development of a safe, effective YFV vaccine, yellow fever remains an important health risk in sub-Saharan Africa and tropical South America [4], [5]. The WHO estimates that there are 200,000 cases and 30,000 deaths attributable to YFV infection each year, most of which occur in Africa [6].
There are two recognized subspecies of Ae. aegypti s.l., the presumed ancestral form, Ae. aegypti formosus (Aaf), a sylvan mosquito supposedly limited to sub-Saharan Africa; and Ae. aegypti aegypti (Aaa), found globally in tropical and subtropical regions typically in association with humans. The designation of Ae. aegypti s.l. subspecies arose from observations made in East Africa in the late 1950's that the frequency of pale “forms” of Ae. aegypti was higher in populations in and around human dwellings than in adjacent forests [7], [8]. The implied correlation between color and behavior prompted Mattingly to revisit the biology and taxonomy of Ae. aegypti [9]. He described formosus (Walker) as a subspecies of Ae. aegypti that was restricted to sub-Saharan Africa and in West Africa “is the only form known to occur except in coastal districts and in one or two areas of limited island penetration.” However, this latter statement was based only on two collections, one from Ghana and the other from Burkina Faso. He also suggested that Aaf most frequently breeds in natural containers such as tree holes, and feeds primarily on wild animals. Mattingly also stated that in addition to the dark-scaled parts of the body being generally blacker, “ssp. formosus never has any scales on the first abdominal tergite.” The type form of Aaa was alternatively defined as “either distinctly paler and browner (at least in the female) than ssp. formosus or with pale scaling on the first abdominal tergite or both.” He also suggested that Aaa breeds in artificial containers provided by humans, will breed indoors, and has a preference for feeding on human blood [9].
The subsequent studies of Tabachnick, Powell, Munstermann and Wallis [10]–[21] on the population genetics and vector competence of Ae. aegypti s.l. showed that global collections fell into two clades. One clade contained Aaa from East Africa, South America and the Caribbean suggesting that these New World populations were derived from East Africa. The other clade contained Asian and Southeastern U.S. Aaa and a basal branch containing Aaf from both East and West Africa suggesting an independent New World and Asian introduction. Their parallel work on vector competence [11]–[13] showed that West African Aaf had lower competence for YFV than other global collections of Aaf and Aaa. A more recent study examined 24 worldwide collections of Ae. aegypti s.l. at 12 polymorphic microsatellite loci [22]. Two distinct genetic clusters were identified: one included all domestic populations outside of Africa and the other included both domestic and forest populations within Africa.
Fourteen papers published since 2000 [23]–[37] and 4 unpublished datasets on GenBank used sequence variation in the mitochondrial NADH dehydrogenase subunit 4 (ND4) gene to describe patterns of gene flow among Ae. aegypti s.l. collections within and among countries outside Africa (Table 1). For example, the first paper in Table 1 was a population genetic analysis of gene flow among 10 Aedes aegypti collections from seven cities along the northeastern coast of Mexico [23]. A total of 574 mosquitoes were examined and 9 novel ND4 haplotypes were discovered. Using Tamura-Nei distance [38] and neighbor joining [39], haplotypes were placed into two clades with 90% and 99% support. Table 1 documents that to date 95 novel ND4 haplotypes have been discovered and that two phylogenetic patterns were consistently noted: either mtDNA haplotypes were distributed on two well supported clades (pattern 1), seen in three published datasets [23], [25], [35], and three unpublished datasets (GB1, GB2, GB4) or as a basal group (more similar to the outgroup species) from which a second well supported derived (less similar to the outgroup species) clade arises (pattern 2 - publications [24], [28], [30], [32]–[34], [36]. These patterns are not limited to the mitochondrial ND4 gene. A study in Brazil utilized the mitochondrial Cytochrome Oxidase I (COI) gene to examine gene flow among 163 mosquitoes in 14 collections [40]. Their phylogenetic analysis identified two clades with 81 and 96% bootstrap support. Based upon comparison with GenBank COI sequences [41] from an Ae. aegypti strain collected from Kenya, another from West Africa and a third Aaf strain; they designated one clade as “East African” and the other as “West African.” A study in Argentina that included collections from Brazil, Paraguay, Uruguay and Bolivia utilized Restriction Fragment Length Polymorphism (RFLP) analysis of the ND4, ND5, COI and COII mitochondrial genes and identified three clades [34]. However, since that study did not include sequence data for these four genes they could not be compared to sequences in the present study. A combined study of Aaa, Aaf, Ae. albopictus and Ae. mascarensis from islands in the southwest Indian Ocean examined phylogenetic relationships within and among all four taxa [41]. Bayesian phylogenetic analysis clearly differentiated two clades; one (labeled GR1) had a credibility value of 0.81 and contained all mosquitoes identified as Aaf while a second clade (GR2) had a credibility value of 0.86 and contained all Aaa mosquitoes. Aaa and Aaf were monophyletic with Ae. mascarensis immediately basal. A study using microsatellites and the mitochondrial ND4 and COI genes in Bolivia detected two clades [36] with credibility values of 0.75–0.76.
Despite the large numbers of studies that have detected these two mitochondrial clades, no studies have been made of the clades in continental Africa. Assumptions about the African origin of Ae. aegypti s.l. are based upon the observation that 58 species of the subgenus Stegomyia are also endemic to Africa [42] and the greatest genetic diversity in allozymes markers [14], [15], [18] and microsatellites [22] in Ae. aegypti s.l. are found in African collections.
It is currently unclear if there is an association between the two well documented mitochondrial clades in the literature and the Aaa and Aaf subspecies or if the clades are differentially associated with East versus West Africa. To address this deficiency, the present study examines ND4 haplotypes among 426 Ae. aegypti s.l. collected at 10 locations in Senegal, West Africa and seven novel haplotypes collected in 7 locations in Kenya in East Africa. A comparison of these sequences was then made with the 95 existing haplotypes detected and reported globally in the literature (Table 1).
Over three years (2005–2008) Ae. aegypti larvae were collected from 10 locations in Senegal (Table 2). These were raised to adults in a field laboratory, bloodfed and eggs were collected. Eggs were transported to Colorado State University where they were hatched and reared to adults. Immediately following eclosion, males and females were classified as either Aaa or Aaf using McClelland's [7] scale pattern system. Mosquitoes with any white scales on the first abdominal tergite of the adult were designated Aaa. If the first abdominal tergite was completely lacking in white scales then the individual was designated Aaf. Adults were allowed to mate and oviposit. DNA was then extracted from each individual using the salt extraction protocol [43] suspended in 300 µl of TE buffer (10 mM Tris-HCl and 1 mM EDTA, pH 8.0), and stored at −80°C. The same procedures were followed with F1 mosquitoes collected in East Africa (Table 2).
DNA was also purified from five other species to serve as outgroups: Ae. (Stegomyia) metallicus (Edwards) (JX427526), Ae. (Stegomyia) luteocephalus (Newstead) (JX427527), Ae. (Stegomyia) unilineatus (Theobald) (JX427530), Ae. (Fredwardius) vittatus (Bigot) (JX427529), and Ae. (Zavortinkus) longipalpis (Grunberg) (JX427528). All were collected near Kedougou, Senegal and identified using four taxonomic keys [42], [44]–[46]. The Ae. (Stegomyia) albopictus (Skuse) sequence was EF153761.
Initially degenerate primers were developed for PCR using the only mosquito mtDNA sequences available in 2000 (An. gambiae, An. albimanus) [23]. These were ND4+ (5′-GTDYAT TTATGATTRCCTAA-3′) and ND4−(5′-CTTCGDCTTCCWADWCG TTC-3′). Although these primers had been used in five prior studies [23]–[25], [31], [32] they failed to amplify any products using template DNA from Senegal Ae. aegypti. New primers were designed once the Ae. aegypti mitochondrial genome (EU352212) became available. They were ND4sb+ (5′-TTATGATTGCCAAAGGCTCAT-3′), and ND4sb− (5′-CTTCGTCTTCCTATTCGTTC-3′). The new ND4 primers were optimized on a gradient thermal cycler and had an optimal annealing temperature of 52°C. Amplification failures with African template DNA and the ND4+/− primers probably occurred because these primers were degenerate and because the primer annealing site for ND4+ varied in the Senegal mitochondrial genomes. The size of the amplified product was 387 bp.
These new primers were used to amplify ND4 from the 426 mosquitoes shown in Table 2. PCRs were 25 uL in volume and used Commercial GoTaq (BioRad, Hercules, CA). Single Strand Conformation Polymorphism (SSCP) analysis was performed on amplified PCR products to identify unique haplotypes for each location [43]. The sensitivity and specificity of SSCP were evaluated by sequencing at least two PCR products for each perceived unique SSCP pattern. PCR products were purified using minElute PCR purification kits (Qiagen, Valencia, CA). DNA concentration was determined on a Nanodrop spectro-photometer (N-1000) (ThermoFisher Scientific, Wilmington, DE). Purified DNA was loaded onto a 96 well semi-skirt plate with either the forward or reverse ND4 primers for each sample. The plates were then sent to the Colorado State University sequencing facility http://www.pmf.colostate.edu/dna_sequencing.html. PCR products from 92 mosquitoes were sequenced.
NUMTS (Nuclear mtDNA) [47] have been previously reported in Ae. aegypti [33], [48]. Because true mitochondrial genomes are haploid, NUMTs are most readily identified by scanning sequences for heterozygous sites (double peaks). To detect NUMTs in the present study, forward and reverse trace files were aligned and tested for heterozygotes using Geneious software (http://www.geneious.com/). No heterozygous sites were detected in any sequences gathered in the present study. However, this approach is not definitive because a NUMT may be entirely homozygous. Three NUMTs were found in GenBank sequences (AF203367, AF203368, AF334847) previously submitted from the senior authors' laboratory [23], [24].
Sequences were aligned using ClustalW http://www.genome.jp/tools/clustalw/. Primer sequences were removed from the 5′ and 3′ ends. Aligned sequences were analyzed with RAxML [49] to identify duplicate sequences. A total of 16 unique haplotypes were found among the 92 sequences and 15 of these were new. The published haplotype (DQ176837) [28] appeared 63 times in 92 sequences and was previously found in Guinea, Uganda, and Singapore.
All phylogenetic analyses employed Maximum Likelihood with bootstrap analysis using RAxML [49]. Bootstrap support was evaluated with 1000 pseudoreplicates to test the consistency of the derived clades. To test the ML phylogeny, a Bayesian analysis of the same dataset was performed using MrBayes3.2 [50]. Trees were drawn using TreeGraph2 [51]. Distance/Neighbor-joining and Maximum Parsimony trees were not derived because most of the datasets had already been subjected to these analyses in the original publications (Table 1).
The first dataset analyzed contained the 34 Ae. aegypti haplotypes found to date in Africa. These were comprised of the 15 new unique Senegal haplotypes from the present study and one Senegal haplotype collected in Dakar in a previous study [28] (labeled in red in Figure 1). Seven novel haplotypes from Kenya and one from Uganda are labeled in blue, three from Cameroon [29] appear in black and seven haplotypes that appeared in collections from Africa and other global locations in various other studies (Table S1) appear in green. Figure 1 is a ML tree with % bootstrap support and clade credibility scores (a posteriori probabilities from Bayesian analysis) appearing over branches with >50% support or credibility scores >0.5.
There are six patterns to note in this phylogeny. First, based upon use as outgroups of four related subgenus Stegomyia species and two additional African subgenera, two clades are identified. This clade has a moderate 72% bootstrap support with maximum likelihood analysis and a clade credibility value of 0.75 in the Bayesian analyses. In addition, these same clades were independently detected in seven of the fifteen published studies and in two of the three unpublished GeneBank datasets. One of the clades is basal (more similar to the outgroups) while the second clade is derived (less similar to the outgroups) from the basal clade. Hereafter these are referred to as the “basal” and “derived” clades. Second, all 15 new Senegal haplotypes occur in the basal clade while the one haplotype collected in Dakar belongs to the derived clade. Third, two of the eight east African haplotypes, one from Kenya and one from Uganda appear in the basal clade but six are in the derived clade. Fourth, the basal clade contains two globally distributed haplotypes. AF203348 has been found independently in 9 studies from Mexico, Brazil, Venezuela, Thailand, Tahiti, Cambodia, Singapore, Myanmar, and Kenya (Table S1) while DQ176837 has been found independently in five studies from Guinea, Uganda, Singapore, Cameroon, Brazil, Myanmar, and Senegal. Fifth, the derived clade has three basal branches represented by global haplotypes. EU650411 and EU650417 have been found in Brazil, Senegal, and the USA [35] while AF203356 has been found in 8 studies from Mexico, Brazil, Venezuela, USA, Senegal and Myanmar. Within the derived clade, AY906841 has been collected from Brazil and Kenya while DQ440274 has appeared in 5 studies from Senegal, Venezuela, and Thailand. The sixth pattern is that there was no difference in subspecies composition between the two clades. All Kenyan mosquitoes lacked scales on the first abdominal tergite (were Aaf) but occurred on the same clade with mosquitoes previously identified as Aaa. Similarly, we have previously shown [52] that Ae. aegypti s.l. from northwest Senegal mosquitoes are composed mostly of Aaa, while those from southeastern Senegal are mostly pure Aaf. Yet all Senegal mosquitoes collected in the present study occur on the basal clade.
The results in Fig. 1 prompted us to examine all of the 215 Ae. aegypti s.l. ND4 sequences currently on GenBank (Tables 1 & S1). After removing redundant sequences, 95 unique ND4 haplotypes remained. The ML and Bayesian phylogenies containing all 117 (95+15(Senegal) +7(Kenya)) haplotypes and outgroups appear in Figures S1 and S2 respectively. The same six patterns noted in Figure 1 are repeated in these two full analyses. Of the 65 haplotypes that occur in the basal clade, 19 are from Africa (15 from Senegal, 2 from Cameroon and 1 each from Kenya and Uganda), 8 are from North America, 16 are from South America, and 13 are from Southeast Asia. The basal group contains 6 global or widely distributed haplotypes (AF203348, DQ176837 from Fig. 1, DQ176845, DQ176848, and AF203346 from the New World, and EF153747 from the New World and Thailand). The three NUMTs in GenBank appear at the base of the basal group.
Of the 52 haplotypes that occur in the derived clade, 8 are from Africa (1 from Senegal, 1 from Cameroon and 6 from Kenya), 7 are from North America, 15 are from South America, and 13 are from Southeast Asia. The derived group contains 8 global or widely distributed haplotypes (EU650411, EU650417, AF203356, AY906841 and DQ440274 from Fig. 1, AF203344, AF334863, and AF334860 from the New World). No NUMTs were found in the derived clade.
The phylogeny displayed in Figure 1, the phylogenetic analyses of all mtDNA ND4 haplotypes reported to date (Figs. S1 & S2) in addition to the fourteen independently derived phylogenies that appear in publications (Table 1) all support an hypothesis that Ae. aegypti populations from around the world consist of mosquitoes that arise from one of two matrilineages. Outgroups consisting of four related subgenus Stegomyia species and two additional African subgenera, consistently indicate that one of the Ae. aegypti matrilineages is basal while the second matrilineage arises from the first.
The purpose of this study was to trace the African origins of these two clades. Key observations are that all but one of the ND4 haplotypes from Senegal occur on the basal matrilineage whilst haplotypes from East Africa arise predominantly on the second, derived matrilineage (Fig. 1). However, samples from Kenya are only from the Rabaï area. Mbarakani, Bengo and Rabaï are approximately 100 m apart and this cluster is 4 km from Changombe. Further, Rabaï is 14 km inland from Mombassa on the coast. Mombassa is the second largest city of Kenya and a major port. Thus, as with Dakar in Senegal, Mombassa could easily be a place where Ae. aegypti immigrate through human commerce. It would be very interesting to sample Ae. aegypti from other locations further inland in East Africa to assess this possibility.
This pattern prompted us to re-examine all of the 215 Ae. aegypti s.l. ND4 sequences in GenBank (Table S1). Phylogenetic analyses of the 95 unique haplotypes confirmed that all but one West African haplotype occurred on the basal matrilineage. This matrilineage also contained many globally distributed haplotypes. Conversely, most east African haplotypes occurred on the derived matrilineage which also contained many globally distributed haplotypes. The phylogenies presented here demonstrate that Ae. aegypti populations outside Africa consist of “mixtures” of mosquitoes from both the basal and derived matrilineages.
Figure 1 is ambiguous as to whether Aaf or Aaa (sensu Mattingly) was the ancestor because basal haplotypes were detected in mosquitoes with and without scales on the first abdominal tergite [52]. This result is not surprising given that McClelland's 1974 study [53] also found collections of almost pure Aaf in Pensacola, Key West and Miami, Florida. Conversely, collections from Kenya, Nigeria, Tanzania, Senegal, Ghana, Burkina Faso, Sri Lanka, Calcutta, Jamaica, and Miami Airport contained diverse mixtures of Aaf and Aaa mosquitoes. Inferences about subspecies composition and West versus East African origins cannot be inferred from the earlier allozymes studies [10]–[21] nor from the recent microsatellite study [22] because they did not use McClelland's [53] scoring scheme nor did they include an outgroup.
The current study is unique in providing the first mitochondrial ND4 data from West Africa and definitively associating the two clades reported in the literature with West and East Africa. We strongly emphasize that the hypotheses and patterns described in this paper are not novel. Bracco et al [28] made 36 collections throughout the New World (Brazil, Peru, Venezuela, Guatemala, US), three from Africa (Guinea, Senegal, Uganda), and three from Asia (Singapore, Cambodia, Tahiti). They also detected two clades and concluded that “three percent of nucleotide divergence between these two clades is suggestive of a gene pool division that may support the hypothesis of occurrence of two subspecies of Ae. aegypti in the Americas.” Later the two clades were actually labeled as East and West African albeit based on only three haplotypes from long established laboratory strains [40]. Most recently a combined analysis of ND4 and CO1 also associated one clade (lineage 2) with West Africa [41].
We have recently discovered multiple chromosome inversions in Ae. aegypti s.l. [54] and M. Sharakova obtained direct visual evidence with Fluorescent In Situ Hybridization (FISH) for inversions on each arm of the third chromosome (unpublished). An obvious question arises as to how these inversions correspond to mosquitoes from the two clades and to the global versus African Ae. aegypti microsatellite clades [22].
|
10.1371/journal.ppat.1007775 | An ortholog of Plasmodium falciparum chloroquine resistance transporter (PfCRT) plays a key role in maintaining the integrity of the endolysosomal system in Toxoplasma gondii to facilitate host invasion | Toxoplasma gondii is an apicomplexan parasite with the ability to use foodborne, zoonotic, and congenital routes of transmission that causes severe disease in immunocompromised patients. The parasites harbor a lysosome-like organelle, termed the "Vacuolar Compartment/Plant-Like Vacuole" (VAC/PLV), which plays an important role in maintaining the lytic cycle and virulence of T. gondii. The VAC supplies proteolytic enzymes that contribute to the maturation of invasion effectors and that digest autophagosomes and endocytosed host proteins. Previous work identified a T. gondii ortholog of the Plasmodium falciparum chloroquine resistance transporter (PfCRT) that localized to the VAC. Here, we show that TgCRT is a membrane transporter that is functionally similar to PfCRT. We also genetically ablate TgCRT and reveal that the TgCRT protein plays a key role in maintaining the integrity of the parasite’s endolysosomal system by controlling morphology of the VAC. When TgCRT is absent, the VAC dramatically increases in volume by ~15-fold and overlaps with adjacent endosome-like compartments. Presumably to reduce aberrant swelling, transcription and translation of endolysosomal proteases are decreased in ΔTgCRT parasites. Expression of subtilisin protease 1 is significantly reduced, which impedes trimming of microneme proteins, and significantly decreases parasite invasion. Chemical or genetic inhibition of proteolysis within the VAC reverses these effects, reducing VAC size and partially restoring integrity of the endolysosomal system, microneme protein trimming, and invasion. Taken together, these findings reveal for the first time a physiological role of TgCRT in substrate transport that impacts VAC volume and the integrity of the endolysosomal system in T. gondii.
| Toxoplasma gondii is an obligate intracellular protozoan parasite that belongs to the phylum Apicomplexa and that infects virtually all warm-blooded organisms. Approximately one-third of the human population is infected with Toxoplasma. Toxoplasma invades host cells using processed invasion effectors. A lysosome-like organelle (VAC) is involved in refining these invasion effectors to reach their final forms. A T. gondii ortholog of the malarial chloroquine resistance transporter protein (TgCRT) was found to be localized to the VAC membrane. Although the mutated version of the malarial chloroquine resistance transporter (PfCRT) has been shown to confer resistance to chloroquine treatment, its physiologic function remains poorly understood. Comparison between the related PfCRT and TgCRT facilitates definition of the physiologic role of CRT proteins. Here, we report that TgCRT plays a key role in affecting the integrity and proteolytic activity of the VAC and adjacent organelles, the secretion of invasion effectors, and parasite invasion and virulence. To relieve osmotic stress caused by VAC swelling when TgCRT is deleted, parasites repress proteolysis within this organelle to decrease solute accumulation, which then has secondary effects on parasite invasion. Our findings highlight a common function for PfCRT and TgCRT in mediating small solute transport to affect apicomplexan parasite vacuolar size and function.
| Toxoplasma gondii uses polypeptide invasion factors to efficiently invade host cells. These proteins are stored in two unique sets of organelles in Toxoplasma parasites, micronemes and rhoptries. Microneme proteins undergo a series of proteolytic cleavage steps within the parasite's endosomal system, followed by further trimming and intramembrane cleavage on the parasite surface [1,2]. Proper maturation and secretion of microneme proteins are crucial for efficient invasion of parasites [3–5].
Microneme protein maturation is conducted by several proteases. During intracellular trafficking, microneme proteins are first cleaved by aspartyl protease 3 (TgASP3) in a post-Golgi compartment [5]. A cathepsin L-like protease (TgCPL) was also shown to help process some microneme proteins in the endosome-like compartment (ELC) of the parasite [4]. The mature proteins then pass through the micronemes and undergo further trimming and intramembrane cleavage on the parasite surface. More specifically, a subtilisin ortholog, TgSUB1, was shown to trim some microneme proteins including microneme protein 2 (TgMIC2) and TgMIC2-associated protein (TgM2AP) on the parasite surface [3]. Subsequently, an integral membrane protease, rhomboid 4 (TgROM4), intramembranously cleaves transmembrane microneme proteins to release them from the cell surface. TgROM4 substrates include TgMIC2 and apical membrane antigen 1 (TgAMA1) [5–8]. Overall, precise control of proteolytic activities within the parasite’s endosomal system and on the plasma membrane is critical for processing parasite invasion effectors.
Among these proteases, TgCPL and TgSUB1 are both localized to, or transit through, the parasite’s endolysosomal system. TgCPL is located in a lysosome-like organelle, termed Vacuolar Compartment (VAC) or Plant-Like Vacuole (PLV) (hereafter referred to as VAC) in Toxoplasma parasites [4,9,10]. Our previous studies showed that the genetic ablation of TgCPL causes defects in parasite invasion and acute virulence [4,11]. TgCPL becomes activated in the VAC and a portion of TgCPL is delivered to the juxtaposed ELC for maturation [4]. TgSUB1 is a micronemal protease and contains a GPI anchor necessary for membrane association [3]. TgSUB1 was shown to be activated in a post-ER compartment and to transit through the parasite’s endolysosomal system before trafficking to micronemes [12]. Deletion of TgSUB1 leads to inefficient trimming of microneme proteins on the parasite surface, thereby resulting in defects in invasion and virulence [3]. Hence, maintaining integrity of the parasites’ endolysosomal system is critical for regulating the distribution and activity of endolysosomal proteases.
In addition to VAC dysfunction resulting in reduced invasion, replication, and virulence [4,11], parasites with impaired VAC proteolytic function are unable to turn over autophagosomes during chronic infection and thereby cannot survive in host brain tissues [13]. Despite its importance, the VAC has not been well characterized. Only a few proteins have been localized to the VAC/PLV [4,10,14–17]. The VAC exists as a prominent intact organelle during initial infection and subsequently fragments during intracellular replication, based on staining of TgCPL, a major luminal protease in the VAC [4]. It is unknown how parasites regulate these and other morphological changes that occur within the endolysosomal system. In a previous study, TgCRT expression was knocked down in Type I Toxoplasma parasites using a tetracycline-inducible system [15], and VAC swelling was observed, suggesting that the function of TgCRT is necessary to maintain normal VAC volume. Fitness defects were also seen in the TgCRT knockdown strain [15]. However, a detailed characterization of how an altered VAC affects different steps of parasite intracellular growth is still missing and the corresponding molecular mechanisms underlying the phenotypes are not understood.
Interestingly, the swollen VAC phenotype for the TgCRT knockdown mirrors the enlarged digestive vacuole (DV) phenotype for chloroquine-resistant (CQR) Plasmodium falciparum expressing CQR-associated mutant PfCRT [18]. More recently, an L272F PfCRT mutation, along with CQR-conferring mutations, was found to increase DV volume by an additional 1–2 μm3 [19]. In vitro assays using purified recombinant PfCRT, reconstituted in proteoliposomes, suggest that PfCRT transports aminoquinoline drugs, basic amino acids, and perhaps oligopeptides, likely in an electrochemically coupled fashion [20,21]. With respect to drug transport, PfCRT expressed within CQR P. falciparum appears to exhibit higher chloroquine (CQ) transport efficiency relative to PfCRTs found in chloroquine-sensitive (CQS) strains [20–22]. These findings suggest that the PfCRT mediates the transport of key osmolytes from the P. falciparum DV. Unfortunately, the inability to successfully ablate the PfCRT gene [23] limits additional analysis of function in vivo.
Here, we successfully delete the TgCRT gene in a Type I Toxoplasma parasite strain by double crossover homologous recombination. The resulting mutant, Δcrt, displayed a severely swollen VAC and aberrant colocalization of markers for the VAC and ELC. Surprisingly, this aberrant organellar organization is associated with down-regulated transcription and translation of several proteases residing in the parasite’s endolysosomal system, altering microneme secretion and resulting in defective parasite invasion and acute virulence. We also engineer successful overexpression of wild type TgCRT constructs in yeast and show that the protein mediates CQ transport. Collectively, these findings determine a novel role for maintaining endolysosomal integrity, suggest functional similarities for TgCRT and PfCRT proteins, and provide a new model system for analyzing the function of apicomplexan CRT proteins.
Previous studies have localized TgCRT in the VAC by gene epitope-tagging and immunofluorescence microscopy, and utilized an anhydrotetracycline-regulated system to reduce levels of expression of TgCRT in a Type I Toxoplasma RH strain. These results revealed that TgCRT is involved in volume control of the VAC [15]. However, incomplete depletion of TgCRT limits further characterization of its function. Additionally, lack of detailed phenotypic characterization restricts our understanding of how the swollen VAC affects parasite fitness and virulence. Here, we adopted a genetically tractable RH-derived strain, termed RHΔku80 (hereafter referred to as WT), to produce a complete TgCRT knockout (refer to S1 Text, Table 1, and S1A and S1B Fig for more details). The RHΔku80 strain lacks non-homologous end-joining DNA repair, thus enhancing homology-dependent DNA recombination [24]. Due to the increased homologous recombination efficiency, this strain has been widely used as a wild type Toxoplasma strain.
Upon generating RHΔku80Δcrt, we observed that purified extracellular Δcrt parasites exhibited large “concave” subcellular structures under differential interference contrast (DIC) microscopy, whereas WT and ΔcrtCRT strains did not display this phenotype (Fig 1A). This subcellular structure was also observed in pulse-invaded Δcrt parasites (Fig 1B). To identify the swollen structures, we stained the WT, Δcrt, and ΔcrtCRT parasites with anti-TgCPL antibodies. TgCPL is a major luminal endoprotease in the VAC of Toxoplasma [4,25]. Immunofluorescence microscopy showed that TgCPL staining co-localized with concave subcellular structures in Δcrt (Fig 1B). The TgCPL staining in Δcrt was larger than that in WT and ΔcrtCRT parasites, indicating that the VAC becomes swollen when TgCRT is absent. We quantified the VAC sizes based on TgCPL staining as described previously [13,15]. VAC diameter for the Δcrt parasites (1.12 ± 0.07 μm) is approximately 2.6-fold larger than for WT parasites (0.43 ± 0.03 μm), while the ΔcrtCRT (0.46 ± 0.02 μm) VAC was similar to that measured for WT parasites (Fig 1B). If we assume the VAC is approximately spherical, then the Δcrt parasite VAC is approximately 15-fold larger than the WT VAC. In contrast to pulse-invaded parasites, the swollen concave structure was not observed in replicated Δcrt parasites (Fig 1C). However, TgCPL staining showed differences between WT and Δcrt parasites (Fig 1C). For WT, the VAC displayed dynamic structures during replication that appeared as small fragmented puncta upon TgCPL staining [4]. However, TgCPL staining revealed fewer small fragmented punctate structures in replicating Δcrt parasites and instead featured one or more larger and intensely stained puncta (Fig 1C). Overall, we found that loss of TgCRT severely alters the morphology of the VAC in pulse-invaded as well as replicating Toxoplasma.
The VAC is a lysosome-like organelle, participating in the parasite’s endolysosomal system. It provides an environment for maturation of TgCPL and delivers activated TgCPL to its adjacent endosome-like compartment (ELC) to assist in processing microneme proteins required for parasite invasion [4]. It also serves as a digestive organelle to degrade endocytosed proteins [4,11,26]. We hypothesized that the dramatic swelling of the VAC might affect the integrity of the parasite’s endolysosomal system. We stained WT, Δcrt, and ΔcrtCRT parasites with antibodies recognizing markers of the VAC (anti-TgCPL) and of the ELC (anti-proTgM2AP or TgVP1) [4,10], and captured a series of deconvolved Z-stack images for individual co-staining. In pulse-invaded parasites, the VAC and ELC displayed distinct subcellular staining in WT and ΔcrtCRT strains, whereas in Δcrt parasites both markers partially co-localized (Fig 1C). Similarly, the non-fragmented TgCPL puncta in replicating Δcrt parasites also showed partial colocalization with both proTgM2AP and TgVP1 (Fig 1C). We quantified the colocalization of TgCPL with proTgM2AP and TgCPL with TgVP1 by measuring their Pearson’s correlation coefficient (PCC). Our analysis revealed that TgCPL showed significantly higher colocalization with proTgM2AP or TgVP1 in Δcrt parasites than WT and ΔcrtCRT strains in both stages of pulse invasion and replication (Fig 1C). A sodium/proton exchanger, named TgNHE3, was previously identified in the ELC [27]. We also co-stained WT, Δcrt, and ΔcrtCRT strains with anti-TgCPL and anti-TgNHE3 antibodies, and did not observe partial colocalization in Δcrt mutant (S2 Fig). These results mirror a previous observation that the ingested host proteins in the TgCPL-deficient parasites overlapped with proTgM2AP to a greater extent than TgNHE3 during their trafficking through the parasite’s endolysosomal system [26], suggesting that the TgNHE3 occupies a distinct subdomain of the ELC. Overall, our findings suggest that altered VAC morphology due to the absence of TgCRT affects the integrity of the parasite’s endolysosomal system.
Toxoplasma utilizes exocytosis and endocytosis via endolysosomes to release invasion effectors, and to ingest host proteins during intracellular growth, respectively [11,28–30]. We therefore characterized invasion, replication, and egress for the Δcrt strain. First, we measured the invasion efficiency of parasites at 30–120 min post-infection. At 30 min post-infection, the Δcrt mutant showed ~50% reduction in invasion compared to WT and ΔcrtCRT (Fig 2A). Differences in invasion efficiency between WT and Δcrt were reduced by ~20% at 60 min post-infection, and were not seen at 120 min post-infection (S3A Fig), suggesting that Δcrt parasites have slower invasion kinetics relative to the WT strain. To further understand the basis of this invasion deficiency, we compared parasite attachment to host cells using previously published methods [3,31]. We found that the Δcrt mutant showed ~50% reduction in host cell attachment compared to WT and ΔcrtCRT strains (Fig 2B). In contrast, we observed no significant differences in both gliding distance and types (S3B Fig). Second, we used immunofluorescence microscopy to quantify parasite replication. Infected cells were stained with DAPI and anti-TgGRA7 antibodies to define individual parasite nuclei and parasitophorous vacuolar (PV) membranes, respectively. The average number of parasites per PV was calculated for each strain to compare replication rates. There were no statistical differences in parasite replication between WT and Δcrt parasites at 28 and 40 h post-infection (Fig 2C). We also introduced NanoLuc luciferase into WT, Δcrt, and ΔcrtCRT parasites, and measured the fold-change of luciferase activity for 72 h post-infection to calculate relative growth rates. Similarly, we did not observe growth differences between WT and Δcrt at 24, 48, and 72 h post-infection (Fig 2C). Third, the egress efficiency of each strain was determined by a lactate dehydrogenase release-based assay. The parasites were incubated with 0.5 mM Zaprinast for 5 min to induce egress. The egressed parasites disrupt host cell membranes to release lactate dehydrogenase, which is subsequently quantified to extrapolate to the number of egressed PVs. We did not observe egress defects in the TgCRT-deficient parasites (Fig 2D). Last, we determined the acute virulence of Δcrt parasites in a murine model. Outbred CD-1 mice were infected with a subcutaneous or intravenous inoculum of 100 WT, Δcrt, or ΔcrtCRT parasites. Thirty percent of mice infected with the Δcrt mutant survived when mice were infected subcutaneously, whereas WT and ΔcrtCRT infections led to quantitative mortality at 10–12 days post-infection (Fig 2E). Mice receiving WT parasites by intravenous inoculation showed mortality starting at 13 days post-infection and all expired at 20 days. The Δcrt and ΔcrtCRT parasites caused death in 40% and 80% of infected mice, respectively (Fig 2E). Statistical analysis showed that mice infected with WT and Δcrt parasites have significant difference in their survival time. Seroconversion of the surviving mice was confirmed by ELISA. We also challenged surviving Δcrt mice with 1000 WT parasites by subcutaneous injection and did not observe lethality after 30 days post-challenge. These findings indicate that the pre-inoculation of Δcrt parasites conferred immunological protection against subsequent acute toxoplasmosis. Given the hyper-virulent nature of Type I Toxoplasma strain in a murine model, the Δcrt mutant dramatically lost its acute virulence compared to WT parasites. Collectively, our findings revealed that Toxoplasma parasites require the TgCRT protein for optimal invasion and acute virulence but not for replication and egress.
During infection, Toxoplasma parasites sequentially secrete proteins to facilitate host invasion. Microneme proteins are the first to be secreted. These proteins traffic through the parasite’s endolysosomal system and undergo intracellular maturation before being trimmed and released from the parasite surface by intramembrane cleavage [3–8,32]. To test which step(s) is affected in the Δcrt parasites, we probed cell lysates and excretory secretory antigen fractions (ESAs) of each strain with anti-TgMIC2, anti-TgM2AP, and anti-TgMIC5 by immunoblotting to measure abundances and secretion patterns. The migration patterns of these three microneme proteins in cell lysates were similar among the strains (Fig 3A). The abundances of the individual microneme proteins were normalized against the protein level of the Toxoplasma actin protein by densitometry and plotted for quantification. All three strains showed comparable steady-state abundances of these proteins (Fig 3A). To further evaluate abundances of secreted microneme proteins, we probed constitutive and induced ESAs with the same antibodies. The constitutive and induced ESAs were generated by incubating purified parasites in D10 medium (DMEM medium supplemented with 10% (v/v) cosmic calf serum) for 30 min at 37°C or D10 medium supplemented with 1% (v/v) ethanol for 2 min at 37°C, respectively. In the ESAs secreted by WT parasites, TgMIC2 exhibited two bands migrating at 100 kDa and 95 kDa, while TgM2AP showed 4 proteolytically processed polypeptides along with pro- and mature forms. However, TgMIC2 only existed as a 100 kDa band in Δcrt parasites. Furthermore, mature TgM2AP was not processed in the constitutive ESAs of the Δcrt strain and showed significantly reduced processing in the induced ESAs of Δcrt parasites (Fig 3B). The secreted TgMIC5 protein displayed similar migration patterns among these strains (Fig 3B). Secretion of these microneme proteins was also quantified by normalizing the relative abundances of the proteins against the protein level of secreted TgGRA7, a dense granule protein. The secretion of TgMIC2, TgM2AP, and TgMIC5 were reduced by ~80%, 50%, and 40%, respectively, in the induced ESAs of Δcrt parasites, compared to the WT strain. The differences in the amount of microneme secretion were less significant in the constitutive ESAs. Secretion of TgMIC2 and TgM2AP was decreased by ~40% and 25%, respectively, in Δcrt parasites compared to the WT strain, whereas TgMIC5 did not show a difference (Fig 3B). Lower ESA secretion induced by ethanol in the Δcrt mutant suggests that Δcrt parasites are deficient in rapid release of certain proteins from the micronemes.
To examine whether the abnormal secretion of microneme proteins alters their intracellular trafficking patterns, we stained pulse-invaded and replicated parasites with TgMIC2 and TgM2AP antibodies. Both microneme proteins trafficked to the apical end of the parasites and showed normal staining patterns (Fig 3C). Prior to secretion, some transmembrane microneme proteins are released via proteolytic cleavage by the intramembrane rhomboid protease TgROM4. Deletion of TgROM4 leads to retention of some microneme proteins on the parasite’s plasma membrane, such as TgMIC2 and TgAMA1 (Toxoplasma apical membrane antigen 1) [6–8,32]. To test whether the aberrant endolysosomal system alters the retention of microneme proteins on the surface of parasites, we stained purified, non-permeabilized extracellular parasites with anti-TgMIC2 antibodies. Immunofluorescence microscopy did not reveal excess TgMIC2 on the plasma membrane of Δcrt parasites (S4 Fig), suggesting that the reduced secretion of microneme proteins is not due to their inefficient intramembrane cleavage on the parasite’s plasma membrane.
The endosome-like compartment is involved not only in the trafficking of microneme proteins, but also rhoptry contents [33]. We stained pulse-invaded and replicated parasites with anti-TgROP7 antibodies to examine the trafficking of rhoptry proteins and the morphology of rhoptries. TgROP7 staining revealed typical rhoptry patterns located at the apical end of the parasites (Fig 3D), excluding the possibility of aberrant trafficking of rhoptry contents and possible defects in rhoptry biogenesis. Taken together, our data suggest that the invasion defects for Δcrt parasites are caused by incomplete trimming and consequent inefficient secretion of microneme proteins at the parasite’s plasma membrane, but not by altered intracellular maturation, trafficking, or intramembrane cleavage of microneme proteins, nor by altered rhoptry morphology.
The inefficient proteolytic processing of TgMIC2 and TgM2AP in Δcrt ESAs led us to investigate whether these phenotypic observations were caused by the abnormal expression patterns or subcellular trafficking of Toxoplasma subtilisin 1 (TgSUB1) in Δcrt parasites. A previous publication reported that parasites lacking TgSUB1 showed defective trimming of secreted microneme proteins, such as TgMIC2 and TgM2AP [3], which we noted resemble the altered patterns of TgMIC2 and TgM2AP products observed for the Δcrt mutant. Therefore, we quantified secreted TgSUB1 in both constitutive and induced ESAs by probing them with an anti-SUB1 antibody, previously found to specifically react against TgSUB1 and PfSUB1 [34]. Immunoblotting analysis revealed that there was no detectable TgSUB1 in the ESAs of Δcrt parasites (Fig 4A). TgSUB1 stays on the plasma membrane via its GPI-anchor prior to its release by self-shedding [12,35]. Given that there was no detectable TgSUB1 in the ESAs of the Δcrt strain, we measured the abundance of TgSUB1 on the parasite surface by immunofluorescence microscopy to assess if the Δcrt retained TgSUB1 on the plasma membrane. We stained non-permeabilized extracellular parasites with anti-SUB1 to evaluate the amount of surface-anchored TgSUB1. Similarly, no detectable TgSUB1 was observed on the plasma membrane of Δcrt parasites (Fig 4B). These data suggest that there is lower expression of TgSUB1 on the surface of Δcrt parasites.
To further dissect the basis for reduced TgSUB1 on the surface, we tested two possibilities: 1) TgSUB1 traffics aberrantly within the parasite to prevent its delivery to parasite surface, and 2) the expression level of TgSUB1 is reduced. TgSUB1 is a microneme protein that also traffics through the parasite’s endolysosomal system [3,12]. The aberrant endolysosomal system in Δcrt parasites potentially alters intracellular trafficking and/or maturation of TgSUB1 that then reduces its expression. To test these two possibilities, first, we stained pulse-invaded and replicated parasites with anti-SUB1 to examine TgSUB1 intracellular trafficking patterns. TgMIC5 localization was used as a reference for typical expected microneme staining. Surprisingly, we observed much less TgSUB1 staining in Δcrt parasites compared to the WT strain (Fig 4C). Next, we quantified abundance of TgSUB1 in parasite cell lysates and found that TgSUB1 was decreased by approximately 85% in Δcrt parasites compared to WT parasites (Fig 4D). To further understand how TgSUB1 expression is suppressed in the Δcrt mutant, we performed qPCR to measure TgSUB1 mRNA for WT, Δcrt, and ΔcrtCRT parasites. TgSUB1 transcript was reduced ~10-fold upon deletion of TgCRT (Fig 4E). Collectively, our findings suggest that the arrested overlap of the VAC and ELC dramatically decreases the abundance of TgSUB1 protein, which then alters the proteolytic processing of normally secreted micronemal protein invasion effectors, thereby reducing invasion efficiency.
The swollen VAC and its aberrant overlap with the ELC in the Δcrt parasites could conceivably lead to altered gene transcription to assist in the adaptation of these parasites. We conducted transcriptome sequencing to detect global alterations in gene transcription for Δcrt parasites relative to WT. Differential gene expression analysis identified 102 genes whose transcript levels changed greater than 1.5-fold in the Δcrt strain. Forty-six and fifty-six genes had increased and reduced transcripts, respectively (Fig 5A and S1 Table).
Four proteases were among the list of genes showing reduced transcripts in the Δcrt mutant, including a putative aminopeptidase N protein (TgAMN, TGGT1_221310), a putative Pro-Xaa serine carboxypeptidase (TgSCP, TGGT1_254010), aspartyl protease 1 (TgASP1, TGGT1_201840), and an ICE family protease-like protein (TgICEL, TGGT1_243298). We validated transcript levels for these proteases, as well as two known VAC luminal proteases (TgCPL and TgCPB), in WT, Δcrt, and ΔcrtCRT strains by qPCR. The qPCR analysis showed that the transcript levels of TgAMN, TgSCP, TgASP1, and TgCPB were decreased by 50%, 20%, 47%, and 14%, respectively, in Δcrt parasites (Fig 5B). Protein levels of TgCPL and TgCPB were quantified by immunoblotting and compared for WT, Δcrt, and ΔcrtCRT parasites. Although TgCPL transcript levels did not differ, the abundance of TgCPL protein was decreased ~40% in the Δcrt mutant (Fig 5C). TgCPB transcript levels were reduced in Δcrt, and both the pro- and mature forms of TgCPB protein were decreased relative to WT parasites (Fig 5C). Densitometry analysis showed that the expression level of mature TgCPB was reduced by ~60% in the Δcrt parasites (Fig 5C).
To determine the subcellular locations of the down-regulated proteases, we tagged endogenous TgAMN and TgSCP with 3xHA and 3xmyc epitope tags at their C-termini in WT parasites, respectively (S5 Fig and Table 1). After drug-selection, we probed cell lysates with anti-HA and anti-myc antibodies, respectively, to test expression. Immunoblotting revealed that the observed molecular masses of both proteins were similar to the predicted sizes based on the primary sequences (Fig 5D). Next, the tagged strains were co-stained with antibodies recognizing the epitope tags along with anti-TgCPL or anti-VP1 antibodies to determine their subcellular locations. Immunofluorescence microscopy revealed that both TgSCP and TgAMN were localized to the VAC/ELC (Fig 5D). TgASP1 subcellular location was also determined to be within the VAC (data were deposited in a public repository, www.toxodb.org; Dou, Z. et al., in preparation). Collectively, these data suggest that the swollen VAC in Δcrt parasites causes reduced transcription and translation of several endolysosomal proteases.
Given that CRT is a putative small solute transporter, the deletion of TgCRT potentially results in the accumulation of small nutrient molecules generated by proteolysis within the VAC, further swelling its size. Therefore, we speculated that inhibition of proteolytic activity within the swollen VAC might reduce its size. We tested this hypothesis by chemically or genetically suppressing VAC proteolysis. First, we treated WT, Δcrt, and ΔcrtCRT parasites with 1 μM LHVS, an irreversible inhibitor of TgCPL protease [25]. TgCPL is a major endopeptidase involved in the maturation of microneme proteins and digestion of host proteins [4,11]. Infected host cells were incubated with LHVS for 48 h to allow full inhibition of TgCPL. Treated parasites were liberated from host cells and used to infect new host cells for 30 min, followed by TgCPL staining to quantify the size of the VAC. As expected, LHVS-treated Δcrt parasites displayed smaller VACs than DMSO-treated Δcrt parasites (Fig 6A). To validate these findings, we genetically ablated TgCPL in Δcrt parasites to create the ΔcrtΔcpl double knockout by CRISPR-Cas9-based genome editing. The replacement of TgCPL with a pyrimethamine resistance cassette was confirmed by PCR and immunoblotting (S6A Fig and Table 1). The ΔcrtΔcpl mutant showed a significantly smaller concave structure than Δcrt parasites under DIC microscopy (S7 Fig). We also compared the size of the VAC in WT, Δcrt, and ΔcrtΔcpl based on TgCPB staining by immunofluorescence microscopy using similar methods previously described, and found that similarly, the ΔcrtΔcpl partially reversed its VAC size compared to Δcrt parasites, but still showed a larger VAC than the WT strain (Fig 6B). Moreover, we deleted TgCPB in Δcrt to test whether such phenotype of partial VAC size restoration was independent of the deletion of specific VAC/ELC-localizing proteases. TgCPB was previously identified as another known VAC-localizing protease, displaying both endo- and exo-peptidase activities [14,25]. Due to its carboxypeptidase activity, it is expected that TgCPB generates more small solutes relative to TgCPL. We used CRISPR-Cas9 genome editing to generate a ΔcrtΔcpb double knockout (S6B Fig and Table 1). The successful gene ablation was confirmed by PCR and immunoblotting (S6B Fig). The resulting ΔcrtΔcpb mutant also showed a smaller concave subcellular structure compared to the Δcrt mutant (S7 Fig). The size of the VAC in WT, Δcrt, and ΔcrtΔcpb was quantified based on the TgCPL staining as described above and the ΔcrtΔcpb parasite VAC was reduced by ~35% compared to Δcrt parasites (Fig 6B).
To test if, like TgCPL and TgCPB, residual expression of TgSUB1 in Δcrt parasites contributed to VAC swelling, we deleted TgSUB1 in Δcrt to create and validate a ΔcrtΔsub1 double knockout mutant (S6C Fig and Table 1). We found that targeted deletion of TgSUB1 in Δcrt partially reversed the swollen VAC phenotype (Fig 6B and S7 Fig). Altogether, our findings from the chemical or genetic inhibition of different proteases strongly suggest that the proteolysis within the VAC plays a key role in its swelling in parasites lacking TgCRT.
Next, we tested whether the partial restoration of VAC size was also associated with the reversal of other phenotypes observed in the Δcrt mutant. Here, we chose the ΔcrtΔcpb strain as a representative strain for further testing. ΔcrtΔcpb showed fewer parasites with partial overlap between the markers for the VAC (TgCPL) and ELC (proTgM2AP) compared to Δcrt. Approximately 44% of ΔcrtΔcpb parasites showed partial overlap between TgCPL and proTgM2AP staining compared to 62% in the Δcrt strain (Fig 6C), with both significantly higher than the 19% and 25% seen for WT and ΔcrtCRT strains, respectively. We also tested whether the partial resegregation of the VAC and ELC in both ΔcrtΔcpb and ΔcrtΔcpl mutants was associated with the recovery of TgSUB1 expression. By immunoblotting, TgSUB1 showed comparable expression in both the WT and ΔcrtΔcpb strains (Fig 6D). Similarly, such recovery of TgSUB1 expression was also seen in the ΔcrtΔcpl strain (Fig 6D). Moreover, TgSUB1 secretion was observed in both constitutive and induced ESA fractions in the ΔcrtΔcpb parasites (Fig 6E). TgM2AP and TgMIC2 were correctly cleaved by TgSUB1 in ΔcrtΔcpb, and their secretion patterns were similar to those seen in the WT strain (Fig 6E). This partial restoration of phenotypes in the ΔcrtΔcpb mutant resulted in an increase of invasion efficiency up to ~60% compared to the Δcrt strain, although invasion was still significantly lower than that of WT parasites (Fig 6F). Collectively, these data show a close association between the size of the VAC, altered morphology of the parasite’s endolysosomal system, protein abundance of TgSUB1, microneme protein secretion and processing, and parasite invasion.
Finally, we attempted to express TgCRT in S. cerevisiae yeast following previously described strategies for PfCRT [36,37]. Native TgCRT cDNA was not expressed well in S. cerevisiae yeast (S8 Fig). Via alignment with PfCRT (S9A Fig), removing the 300 N-terminal residues of the TgCRT gene that are non-homologous to PfCRT preserves all putative transmembraneous domains and inter-helical loop regions. Thus, following a previously published strategy for difficult to express PfCRT mutants [22] we created a fusion gene that replaced the 300 most N-terminal residues of the TgCRT sequence with the 111 most N-terminal residues from the S. cerevisiae plasma membrane ATPase (PMA), which harbors a yeast plasma membrane localization sequence (S9B and S9C Fig). The fusion protein was well expressed in the plasma membrane of S. cerevisiae (S8 Fig) with vacuolar-to-cytosol topology preserved, similar to the expression of PMA-PfCRT in yeast as described in a previous publication [38]. Following an approach previously described for PfCRT and PfCRT mutants [22,39], we assayed PMA-TgCRT expressing yeast for chloroquine (CQ) transport (Fig 7A). Via alignment with PfCRT (S9A Fig), TgCRT T369 corresponds to the well-studied K76 residue within PfCRT; previously, mutation of PfCRT K76 to T has been shown to increase the efficiency of CQ transport by PfCRT, although it is not the sole determining factor [22,40,41]. We expressed both WT TgCRT and T369K TgCRT in yeast to measure their transport efficiencies. Both the wild type TgCRT protein and a TgCRT T369K mutant were found to transport CQ, albeit slower than PfCRTHB3 and PfCRTDd2 derived from CQ-sensitive and resistant malaria strains, respectively, under similar conditions (Fig 7A). Also, we noted that TgCRT required higher external [CQ] (80 mM versus 16 mM for PfCRT) to achieve similar levels of transport, suggesting increased CQ Km for this transporter (Fig 7A). Analysis of PfCRT isoforms and the role that individual amino acid substitutions play in modifying transport activity has been ongoing for well over a decade (e.g. Callaghan et al. 2015 [22]). Since PfCRT—TgCRT identity is modest (S9A Fig) many more TgCRT mutagenesis studies will be required to fully define catalytically critical sites in TgCRT. Regardless, these initial data show that mutation of a TgCRT threonine, analogous to the well-known position 76 T for PfCRT [22,40,41], also affects CQ transport as is the case for PfCRT.
We also exchanged threonine for lysine in the WT TgCRT complementation construct and transfected Δcrt to examine the extent to which TgCRTT369K affects VAC size in Toxoplasma parasites. Interestingly, in contrast to full recovery of VAC size in the WT TgCRT complementation strain, TgCRTT369K only partially restored the swollen VAC (Fig 7B). These findings, along with the TgCRT transport data, strongly suggest that the swollen VAC is caused by luminal osmolyte excess, similar to findings for PfCRT as described in "Discussion".
In summary, our findings strongly suggest a role for TgCRT in small solute transport that impacts VAC volume in the absence of drug transport, similar to the role proposed for PfCRT [18,19]. The identity of the relevant osmoregulatory transport substrate(s) remains to be determined. However, at least for T. gondii, TgCRT expression is also required for proper segregation of other organelles within the endolysosomal system that, in turn, indirectly facilitates microneme secretion and parasite invasion. The data also indicate that the invasion deficiency exhibited by the Δcrt mutant is likely due to multiple factors, since the recovery of TgSUB1 expression and trimming of microneme proteins in ΔcrtΔcpb did not completely reverse the invasion defects. To the best of our knowledge, this is the first observation of the regulation of the endolysosomal protease expression in apicomplexan parasite by a CRT protein.
Toxoplasma utilizes an endolysosomal system to secrete invasion effectors that disseminate infection. These invasion effectors undergo a series of intracellular proteolytic cleavage and trimming steps to reach their final forms. Therefore, maintenance of the integrity of the endolysosomal system is critical for controlling the secretion of invasion effectors in Toxoplasma. The Vacuolar Compartment (VAC) (also Plant-Like Vacuole or PLV) is an acidic lysosome-like vacuole. Previous work showed that deletion of a cathepsin L-like protease, a major VAC luminal endopeptidase, leads to invasion, replication, and virulence defects [4,11]. Compromised proteolytic activities within these parasites also result in the inefficient degradation of endocytosed host proteins [11]. Liu et al. genetically deleted TgVP1 in T. gondii and observed defective secretion and trafficking of microneme proteins, and reduced invasion and virulence in the mutant [16]. Warring et al. previously reported that a Toxoplasma ortholog of chloroquine resistance transporter (TgCRT) resides in the VAC and that decreased expression of TgCRT leads to swelling of the VAC [15]. However, incomplete reduction of TgCRT expression and lack of systematic dissection of phenotypes in the TgCRT knockdown mutant limit understanding of the molecular mechanism by which a dysfunctional VAC affects the parasite.
Here, we created a TgCRT knockout mutant that completely removes TgCRT from the VAC membrane. The resulting Δcrt strain displays a dramatic increase in VAC size, and the organelle aberrantly overlaps with the adjacent endosome-like compartment (Fig 8). Although a previous study reported that parasites deliver minor amounts of TgCPL to the ELC, which then contributes to maturation of some microneme proteins [4], our data do not reveal abnormal intracellular cleavage or trafficking of microneme proteins in Δcrt parasites. This is likely due to the modest decrease of TgCPL expression in Δcrt. Relatedly, Dogga et al. recently documented that aspartic acid protease 3 (TgASP3) localizes in a post-Golgi compartment and serves as a major maturase for invasion effectors [5].
We also evaluated the processing of microneme proteins by TgROM4 and TgSUB1 in the Δcrt mutant. We found that the microneme proteins were improperly trimmed on the surface of Δcrt parasites by TgSUB1. Patterns of secreted microneme proteins observed for the Δcrt mutant were similar to those for Δsub1 parasites, which led us to examine the expression of TgSUB1 in Δcrt parasites and ESAs. As expected, levels of TgSUB1 were decreased on the surface of Δcrt parasites and in the medium during secretion. Interestingly, the steady-state abundance of TgSUB1 was also significantly decreased in the Δcrt mutant. Surprisingly, we found that the reduction of TgSUB1 was due to a decrease in the transcription level of TgSUB1 in the Δcrt strain, suggesting that the parasites utilize a transcriptional feedback mechanism to regulate TgSUB1.
TgSUB1 is a micronemal GPI-anchored protein. It remains unclear how TgSUB1 becomes activated within Toxoplasma. Previous pulse-chase experiments have revealed that TgSUB1 undergoes maturation in a post-ER compartment, and passes through the endolysosomal system before its arrival at the microneme and subsequent secretion [12]. The propeptide region of TgSUB1 carries targeting information which helps to guide the protein to the micronemes [35]. The propeptide may also function by binding to active sites of mature TgSUB1 to inhibit its proteolytic activity during trafficking. Overlap of the VAC and ELC could bring propeptide-bound TgSUB1 to a protease-abundant environment, where non-specific digestion of the propeptide could then lead to increased digestive activities in the VAC and ultimately result in an increase in osmotic pressure within the hybrid VAC/ELC organelle (Fig 8). In this scenario, the parasites may utilize a feedback mechanism to repress additional expression of TgSUB1 to avoid further VAC swelling. Moreover, we also discovered that the Δcrt parasites had reduced protein and/or transcript levels of several other proteases, including two known VAC proteases, TgCPL and TgCPB. Therefore, the parasites down-regulate a number of endolysosomal-VAC proteases to suppress proteolytic activities in the swollen VAC, presumably to reduce osmotic pressure and thereby control VAC size. Among these proteases, TgSUB1 has been shown to be involved in parasite invasion and virulence defects, but not in replication and egress [3]. Additionally, TgCPL plays a role in parasite invasion by contributing to the maturation of at least two microneme proteins [4]. Therefore, the invasion defects exhibited in the Δcrt mutant could be due to several factors. Given that the Δvp1 mutant showed similar phenotypes as our Δcrt mutant, such as reduced secretion of microneme proteins and decreased invasion, we tested whether the loss of TgVP1 can cause swelling of the VAC. However, TgVP1-deficient parasites showed normal VAC size and TgSUB1 expression (S10 Fig). Also, higher baseline cytosolic calcium and pH levels previously observed in Δvp1 parasites were not seen in Δcrt (S3C and S3D Fig). These data suggest that the invasion phenotypes observed within Δcrt and Δvp1 mutants have different underlying molecular mechanisms.
Altered endolysosomal protease transcript levels in Δcrt parasites suggest that parasites repress transcription factors or enhance transcription repressors to respond to increased VAC size. RNA-Seq analysis did not reveal any significant changes in the AP2-family of transcription factors (S3 Table). In mammalian cells, the transcription factor EB (TFEB) is a master regulator that drives gene expression for autophagy and lysosome biogenesis [42]. A search of the Toxoplasma genome did not identify a TgTFEB ortholog, suggesting that these parasites may adopt an alternative strategy for regulating lysosomal gene expression. Interestingly, our differential gene expression analysis found that the transcript levels of two zinc finger (CCCH) type motif-containing proteins, TGGT1_246200 and TGGT1_226310, were increased and decreased by 2-fold and 3-fold (S1 Table), respectively, in the Δcrt mutant. The CCCH type zinc finger motif-containing protein is known to regulate the stability of mRNA [43]. For example, tristetraprolin inhibits the production of tumor necrosis factor-α in macrophages by destabilizing its mRNA via an interaction with AU-rich elements at the 3’-untranslated region [44]. Further investigation to identify transcription factor(s) and regulator(s) that govern the expression of Toxoplasma lysosomal genes will help elucidate how these parasites regulate the biogenesis and function of the VAC.
In this study, we have determined that TgCRT-deficient parasites have reduced expression of several endolysosomal proteases. We have also found that suppression of proteolytic activities within the swollen VAC decreases the size of the organelle. These findings, along with data verifying that TgCRT is indeed a transporter with function similar to that of PfCRT, support the idea that TgCRT functions to transport essential VAC osmolytes, similar to proposals for PfCRT [18,19,45]. Likely candidate osmolytes include ions and/or amino acids. We suggest that when TgCRT is absent on the membrane of the VAC, protein degradation products (short peptides and amino acids) accumulate within the VAC and increase osmotic pressure, thereby leading to the swollen phenotype. Consistent with this idea, and similar to related observations for P. falciparum treated with cysteine protease inhibitors [46], chemical inhibition of proteolysis via the small inhibitor LHVS dramatically reduces the size of the VAC. For Toxoplasma, LHVS principally targets TgCPL, but also inhibits TgCPB [14]. Therefore, LHVS treatment blocks both of these VAC proteases. Genetic ablation of TgCPL or TgCPB in Δcrt individually resulted in the partial restoration of VAC size. Interestingly, the VAC in ΔcrtΔcpl parasites was reduced to a greater extent than that in ΔcrtΔcpb, suggesting that the deletion of TgCPL reduces proteolytic activity to a higher extent within the VAC than with the loss of TgCPB alone, which is consistent with a previous finding that the maturation of TgCPB is dependent upon the presence of TgCPL [14]. Additionally, the deletion of TgCPB partially restored secretion patterns of microneme proteins and invasion defects. These results also reveal for the first time that TgCPB plays an active role in contributing to proteolysis within the VAC in Toxoplasma parasites.
RNA-Seq analysis identified several other genes with altered transcription levels, suggesting that the parasites may utilize additional strategies to control VAC size. For example, interestingly, levels of aquaporin (TGGT1_215450) transcript were reduced in Δcrt parasites. Previous work showed that this aquaporin is localized to the VAC or PLV [10]. Therefore, it seems likely that Δcrt parasites express less aquaporin to reduce water transport into the VAC/PLV, as an additional tactic to limit VAC swelling. We also found that two putative protein phosphatase 2C (TGGT1_276920 and TGGT1_201520) transcripts are down-regulated in the Δcrt mutant, both of which carry signal peptides indicating endosomal trafficking. TGGT1_276920 and TGGT1_201520 are homologous to PTC3 and PTC1 in S. cerevisiae, respectively. Interestingly, both PTC1 and PTC3 proteins are involved in yeast osmosensory regulation. A mitogen-activated protein kinase pathway is activated when yeast cells experience hyperosmotic conditions. PTC1 and PTC3 negatively regulate this pathway [47,48]. Furthermore, PTC1 was found to control the function and morphology of the yeast vacuole, which further alters its biogenesis [49]. The dramatic change in Toxoplasma VAC volume indicates induced osmotic stress in the Δcrt parasites. The knockout parasites appear to be utilizing a similar mechanism to suppress these protein phosphatases and enhance similar osmoregulatory signaling. We think that similar studies for P. falciparum and other apicomplexan parasites that express CRT orthologs would be informative.
In this study, we also determined the acute virulence of TgCRT-deficient parasites in a murine model. The Δcrt strain did not lead to complete mortality in mice by both subcutaneous and intravenous inoculations. Given the hyper-virulent nature of Type I Toxoplasma strain in a murine model, the Δcrt mutant dramatically lost its acute virulence compared to WT parasites. We also observed that the ΔcrtCRT complement did not result in complete mortality when it was used to infect mice by intravenous inoculation. During the natural route of infection, the parasite infection is spread via the host’s circulation system. Therefore, the intravenous inoculation is a more direct route to measure virulence differences in mice than the subcutaneous inoculation, since it circumvents steps of parasite migration from the infection site to the circulation system. Thus, it is possible that our ΔcrtCRT strain did not fully reverse the acute virulence, albeit significantly recovered the virulence defects compared to Δcrt parasites.
The phenotype of the swollen VAC in the Δcrt strain mirrors the enlarged digestive vacuole in chloroquine (CQ) resistant (CQR) P. falciparum malaria [18]. Peptidomic analysis showed that hemoglobin is not as efficiently degraded within the digestive vacuole (DV) in CQR malaria parasites [45], further suggesting that CQR mutations in PfCRT alter the physiology within the swollen digestive vacuole, thereby compromising DV proteolytic activities. In vitro assays utilizing recombinant PfCRT, reconstituted in proteoliposomes, have revealed that PfCRT may act as a proton gradient dependent, polyspecific nutrient exporter for small solutes including amino acids, oligopeptides, glutathione, and small drugs [21]. These studies also demonstrate that CQR-associated PfCRTs display altered transport efficiency relative to CQ-associated PfCRT. Our study has revealed that TgCRT mediates CQ transport similar to PfCRT. The Δcrt strain appeared more sensitive to CQ relative to WT parasites (S11 Fig), further suggesting that TgCRT is a functional transporter of small solutes across the membrane of the VAC. We suggest that alteration of proteolytic activities in the enlarged VAC of the Δcrt mutant reveals a similar scenario relative to the CQR P. falciparum DV. Given the similarity in components and functionality of the VAC and DV found in Toxoplasma and Plasmodium, this Toxoplasma TgCRT-deficient mutant should prove useful for further studies of the native function of CRT orthologs found within other apicomplexan parasites.
In summary, our findings reveal that the Toxoplasma TgCRT protein is indeed a small molecule transporter that is required for maintaining the normal size and morphology of the VAC/PLV. Unexpectedly, aberrant swelling of the VAC in TgCRT-deficient parasites is also associated with decreased integrity of the parasite’s endolysosomal system, which serves as a conduit for trafficking of invasion effectors. Overlap of the VAC and endosome-like compartment in the TgCRT knockout is also associated with a reduction in transcript and protein levels for several endolysosomal proteases. We found that blocking normal proteolysis within the swollen VAC reduced the size and partially restored the morphology of the organelle. Taken together, these findings suggest that TgCRT mediates the transport of small solutes and that putative accumulation of its substrates increases VAC size. The data show that the integrity of the parasite endolysosomal system is necessary for normal parasite virulence. We suggest that pharmaceutical modulation of the VAC could serve as a novel strategy for managing toxoplasmosis.
This study was performed in compliance with the Public Health Service Policy on Humane Care and Use of Laboratory Animals and Association for the Assessment and Accreditation of Laboratory Animal Care guidelines. The animal protocol was approved by Clemson University’s Institutional Animal Care and Use Committee (Animal Welfare Assurance A3737-01, protocol number AUP2016-012). All efforts were made to minimize discomfort. CO2 overdose was used to euthanize mouse. This method of euthanasia is consistent with the recommendations of the Panel on Euthanasia of the American Veterinary Medical Association.
Morpholine urea-leucyl-homophenyl-vinyl sulfone phenyl (LHVS) was kindly provided by the Bogyo lab at Stanford University. Fluo-4-AM was purchased from Invitrogen (Life Technologies). Ionomycin was purchased from Sigma Aldrich. All oligonucleotide primers used in this study were listed in S2 Table. Other chemicals used in this work were analytical grade and were purchased from VWR unless otherwise indicated.
Toxoplasma gondii parasites were cultured in human foreskin fibroblast (HFF) cells (ATCC, SCRC-1041) or hTERT cells in Dulbecco’s Modified Eagle Medium (DMEM) media supplemented with 10% cosmic calf serum at 37°C with 5% CO2. The parasites were harvested by membrane filtration as described previously [14].
To generate the TgCRT-deficient strain, ~1.5 kilobases (kb) of the 5’- and 3’-UTR of the TgCRT gene, respectively, were PCR-amplified and flanked at both ends of the bleomycin resistance cassette (BLE) to assemble a TgCRT deletion construct. The resulting plasmids were introduced into WT parasites by electroporation. The transfected parasites were selected with 50 μg/ml bleomycin twice, while in their extracellular stage as described previously [14]. Clones of the TgCRT-deficient parasites were isolated by limiting dilution. The correct replacement of TgCRT with the BLE cassette was confirmed by PCR (see S1 Text).
To complement Δcrt parasites, we modified the plasmid pTub-TgCRT-mCherry-3xmyc (a gift from the van Dooren lab), which expresses a C-terminally mCherry-3xmyc epitope-tagged TgCRT under the Toxoplasma tubulin promoter. The plasmid was digested with HpaI and MfeI restriction endonucleases to remove the tubulin promoter and a segment of TgCRT. The remaining DNA fragment served as the backbone for subsequent Gibson assembly to incorporate a PCR amplified ~1 kb region upstream of the Tgku80 gene, the ~1 kb fragment of the TgCRT 5’-UTR region, and the removed partial TgCRT coding sequence to produce the TgCRT complementation plasmid, pCRT-TgCRT-mCherry-3xmyc. The complemented TgCRT is driven by its cognate promoter to maintain physiologic similarity to native TgCRT expression in WT parasites. The 1 kb region located ~6 kb upstream of the TgKu80 gene was used to facilitate a single integration of the TgCRT complementation plasmid into this specific locus by single crossover homologous recombination. The TgCRT complementation construct was digested with SwaI restriction endonuclease enzyme, gel-extracted, purified, and transfected into Δcrt parasites by electroporation.
To introduce NanoLuc luciferase (nLuc) into parasites, we PCR-amplified and assembled the TgTubulin promoter, the coding sequence of the nLuc luciferase, and an HXG selection marker into a nLuc expression construct. The resulting plasmid was transfected into WT, Δcrt, and ΔcrtCRT strains. The transfectants were selected with 25 μg/ml mycophenolic acid and 50 μg/ml xanthine. Stable populations were subjected to limiting dilution to generate individual clones of WT::nLuc, Δcrt::nLuc, and ΔcrtCRT::nLuc and clones were confirmed via luciferase activity.
To generate the ΔcrtΔcpb mutant, the TgCPB gene was replaced with a pyrimethamine resistance cassette using the CRISPR-Cas9 genome editing system [50,51]. The pyrimethamine resistance cassette was PCR-amplified and flanked by 50 bp regions upstream and downstream of the start and stop codons of the TgCPB gene for homologous recombination. A 20 bp region located at the beginning of the coding region of the TgCPB gene was used to design guide RNA and replace the guide RNA targeting TgUPRT gene in the plasmid pSAG1-Cas9::UPRTsgRNA using Q5 site-directed mutagenesis (NEB). The Cas9-GFP and guide RNA constructs were co-transfected into Δcrt parasites with the corresponding repair PCR product. The guide RNA and Cas9 generated a gap within the TgCPB gene to facilitate double crossover homologous recombination. Correct gene replacement was confirmed by PCR. Similar strategies were used to create the ΔcrtΔcpl and ΔcrtΔsub1 mutants. Please refer to the figure legend of S6 Fig for more details.
To epitope-tag TgAMN, we again used CRISPR-Cas9 editing tools to modify the corresponding gene. Guide RNA recognizing the 20 bp region near the TgAMN stop codon was generated using the methods above. The 50-bp homologous regions upstream and downstream of the stop codon of the TgAMN gene were cloned at the 5’- and 3’-ends of the DNA sequence containing the 3xHA epitope tag and the pyrimethamine resistance cassette, respectively, by PCR. The plasmid encoding the guide RNA targeting TgAMN and Cas9-GFP and the PCR product were co-transfected into WT parasites. The stop codon of TgAMN was replaced by the 3xHA epitope tag and pyrimethamine resistance cassette. Stable populations were generated after multiple rounds of pyrimethamine selection and the TgAMN-3xHA fusion protein was confirmed by immunoblotting analysis.
TgSCP was endogenously tagged with a 3xmyc epitope tag via single crossover. An approximately 1 kb region upstream of the TgSCP stop codon was PCR amplified and fused in frame with a 3xmyc epitope to assemble TgSCP-3xmyc. A pyrimethamine resistance cassette was also included, the resulting plasmid was linearized and transfected into WT parasites. The correct tagging was confirmed by immunoblotting.
Threonine 369 was mutated to lysine in the WT TgCRT complementation construct, via site-directed mutagenesis according to the Q5 site-directed mutagenesis procedure (NEB). Linear PCR product was phosphorylated, circularized, and transformed into E. coli. Correct clones were identified by direct DNA sequencing.
T. gondii parasites were allowed to grow in HFF cells for 48 h at 37°C with 5% CO2. Freshly egressed parasites were syringed, filter purified, and harvested in Cytomix buffer (25 mM HEPES, pH 7.6, 120 mM KCl, 10 mM K2HPO4/ KH2PO4, 5 mM MgCl2. 0.15 mM CaCl2, and 2 mM EGTA). Parasites were pelleted at 1,000x g for 10 min, washed once in Cytomix buffer, and resuspended in Cytomix buffer at 2.5 x 107 parasites per ml. 400 μl of parasite suspension was mixed with 20 μg DNA and 2 mM ATP/5 mM reduced glutathione to a final volume of 500 μl. The mixture was electroporated at 2 kV and 50 ohm resistance using the BTX Gemini X2 (Harvard Apparatus). Transfectants were inoculated into a T25 flask pre-seeded with a confluent monolayer of HFF cells. The transfected parasites were allowed to recover prior to drug selection applied 24 h post transfection.
Freshly lysed parasites were used to infect confluent HFF cells pre-seeded in an 8-well chamber slide for 1 hr (pulse-invaded parasites) or 18–24 h (replicated parasites). The extracellular parasites were attached to chamber slides using 0.1% (w/v) poly-L-lysine. Immunofluorescence was performed as described previously [11,14]. Images were viewed and digitally captured using a Leica CCD camera equipped with a DMi8 inverted epifluorescence microscope and processed with Leica LAS X software. For deconvolution microscopy, a series of Z-stack images were captured and processed by the 3-D deconvolution operation embedded in the Leica LAS X software using the following parameters: Total iterations: 10; Refractive index: 1.52; Method: Blind; Remove background: Yes; Rescale intensity: Yes; AutoQuant Deconvolution algorithms licensed from Media Cybernetics Inc. The fluorescence in final images was presented as maximum projection. The colocalization analysis of two proteins of interests was performed by the Coloc2 plugin embedded in the Fuji image processing suite using the following parameters [52]: Threshold regression, Bisection; Algorithms, Li for both channels. Pearson’s correlation coefficient above threshold was recorded and plotted for comparison. The colocalization analysis was derived from 10 parasites per replicate for a total of three replicates. The statistical analysis was calculated by unpaired two-tailed Student’s t-test.
To express PHL2 in the cytoplasm of the parasites, the gene encoding PHL2 was cloned under the Toxoplasma tubulin promoter in a plasmid carrying a chloramphenicol resistance cassette for selection. The resulting plasmids were introduced into WT, Δcrt, and ΔcrtCRT strains by electroporation. After drug selection, the parasite strains expressing PHL2 were cloned out by limiting dilution. Individual clones were confirmed by fluorescence observation via immunofluorescence microscopy. Prior to pH measurement, a calibration curve was generated by measuring the ratio of pHL2 fluorescence excited at 405 nm and 485 nm in the WT strain expressing cytosolic pHL2 according to previously published methods [53]. Briefly, the PHL2-expressing WT parasites were filter-purified, resuspend in PBS at 1x107 parasites per ml. One hundred microliters of parasites were pelleted and resuspended in buffer with different pH values ranging from 6.2 to 7.8. The detergent, Triton X-100, was added into the parasite resuspensions to 0.1% and incubated at room temperature for 10 min to permeabilize parasite cell membranes for PHL2 release. Released PHL2 was excited at 405 and 485 nm and the emitted fluorescence at 528 nm was recorded via a BioTek Synergy H1 fluorescence plate reader. The ratio of the emission intensities at 528 nm (I405/I485) for both excitation wavelengths were plotted, which yields a calibration equation by linear regression analysis. Cytosolic PHL2 expressing WT, Δcrt, and ΔcrtCRT parasites were harvested, resuspended in PBS, and their I405/I485 ratios were measured. The pH of their cytosol was calculated by applying their I405/I485 ratios to the calibration equation. All assays were repeated in three technical replicates per biological replicate in a total of three biological replicates. Data were presented as mean ± SEM.
Freshly egressed parasites were syringed, filter purified, and resuspended at 5 x 108 parasites/ml in D1 medium (DMEM medium supplemented with 1% FBS). One hundred microliters of parasite suspension were transferred to a microfuge tube and incubated at 37°C for 30 min to prepare constitutive ESAs. To isolate induced ESAs, the parasite suspension was incubated in D1 medium supplemented with 1% ethanol for 2 min at 37°C. ESAs were separated from intact parasites by centrifugation at 1,000 x g for 10 min. ESA fractions were transferred to a new microfuge tube, mixed with SDS-PAGE sample loading buffer, and boiled for 5 min for immunoblotting analysis.
Parasite lysates and ESA fractions were prepared in 1x SDS-PAGE sample buffer and boiled for 5 min before resolving on standard SDS-PAGE gels. For immunoblotting, gels were transferred to PVDF membranes by semi-dry protein transfer methods. Blots were blocked with 5% non-fat milk and incubated with primary antibody diluted in 1% non-fat milk. Goat anti-mouse or anti-rabbit IgG antibodies conjugated with horseradish peroxidase were used as the secondary antibody. Immunoblots were developed with SuperSignal WestPico chemiluminescent substrate (Thermo). The chemiluminescence signals were captured using the Azure Imaging System. Bands were quantified by densitometry using LI-COR Image Studio software.
The red-green invasion assay was used to measure the efficiency of parasite invasion. Freshly purified parasites were syringed, filter purified, and resuspended at 5 x 107 parasites/ml in invasion medium (DMEM supplemented with 3% FBS). Two hundred microliters of parasite resuspension were inoculated into each well of an 8-well chamber slide pre-seeded with HFF cells, and parasites were allowed to invade host cells for 30, 60, and 120 min before fixation with 4% formaldehyde for 20 min. Before membrane permeabilization, slides were stained with mouse anti-TgSAG1 monoclonal antibody (1:1,000) for 1 h to label attached parasites. After treatment with 0.1% Triton X-100 for 10 min, the parasites were stained with rabbit polyclonal anti-TgMIC5 antibody (1:1,000) for 1 h to stain both invaded and attached parasites. Subsequently, slides were stained with goat anti-mouse IgG conjugated with Alexa 594 (red) (Invitrogen, 1:1,000) and goat anti-rabbit IgG conjugated with Alexa 488 (green) (Invitrogen, 1:1,000) along with DAPI for nuclear staining. After staining, slides were mounted with anti-fade Mowiol solution and observed by immunofluorescence. Intracellular parasites only showed green fluorescence, whereas extracellular parasites exhibited both red and green fluorescence. Six fields of view from individual invasion experiments were captured by a Leica DMi8 inverted epifluorescence microscope and processed with ImageJ software. The invasion efficiency of each strain was quantified using the following equation ([sum of green parasites] − [sum of red parasites]) ⁄ total host nuclei.
For attachment assay, the HFF cells pre-seeded in 8-well chamber slides were fixed with 2% glutaraldehyde for 10 min at 4°C followed by PBS wash and quench with 1M glycine overnight. The fixed HFF cells were incubated with D3 medium (DMEM supplemented with 3% FBS) for 1 h at 37°C before parasite inoculation. Freshly filter-purified parasites were resuspended at 5x107 parasites per ml in D3 medium (DMEM supplemented with 3% FBS). Two hundred microliters of parasite resuspension (1x107) were added into each well and allowed to attach onto fixed HFF cells for 15 min at 37°C. Unattached parasites were washed off by PBS rinse 10 times. Parasites attached to HFF cells were fixed with 4% paraformaldehyde for 20 min at room temperature and stained with mouse anti-TgSAG1 antibodies for 1 h at room temperature. Goat anti-mouse IgG conjugated with Alexa 594 (red) (Invitrogen, 1:1,000) was used as secondary antibody. Eight view fields in each well for individual strains were captured at 200x magnification and quantified. The assay was repeated in three biological replicates. Final data were combined and plotted using Prism software. The unpaired two-tailed Student’s t-test was used to calculate statistical significance.
Freshly egressed parasites were filter-purified and inoculated into individual wells of an 8-well chamber slide pre-seeded with HFF cells at approximately 1 x 105 cells per well. Non-invaded parasites were washed off at 4 h post-infection. Invaded parasites were allowed to infect host cells for an additional 24 and 36 h before fixation. The infected host cells were stained with monoclonal anti-TgGRA7 (1:1,000) antibody and DAPI to help distinguish individual parasitophorous vacuoles (PVs) and the nuclei of parasites, respectively. Slides were subjected to standard immunofluorescence microscopy for imaging. One hundred parasitophorous vacuoles were enumerated for each strain and plotted as the distribution of different sized PVs. In addition, replication was also expressed as the average number of parasites per PV.
Parasites expressing NanoLuc luciferase were inoculated into a white 96-well tissue culture plate with a flat, solid bottom (Greiner Bio-One) pre-seeded with confluent HFF cells at 1.5 x 103 cells/well. Each strain was inoculated into 4 individual wells to monitor the fold-change of luciferase activity versus time, which is proportional to intracellular growth. At 4 h post-infection, the individual wells were aspirated to remove non-invaded parasites. The first well was treated with 100 μl of lysis buffer containing NanoLuc luciferase substrate and incubated for 10 min, and a luminescence reading was taken by using the BioTek multimode H1 hybrid plate reader. The remainder of the 3 wells were replenished with fresh D10 medium without phenol red for an additional 24, 48, and 72 h. Subsequent luminescence readings were all performed via the methods above. Luminescence readings versus time were normalized against the reading at 4 h post-infection to calculate the fold-change of parasite growth.
A lactate dehydrogenase release assay was used to measure the egress efficiency of parasites. Freshly lysed parasites were filter-purified and resuspended in D10 medium at 5 x 105 parasites/ml. One hundred microliters of parasite suspension were inoculated into each well of a 96-well plate pre-seeded with HFF cells. The parasites were allowed to replicate for 18–24 h, washed, and incubated with 50 μl of Ringer’s buffer (10 mM HEPES, pH 7.2, 3 mM HaH2PO4, 1 mM MgCl2, 2 mM CaCl2, 3 mM KCl, 115 mM NaCl, 10 mM glucose, and 1% FBS) for 20 min. Subsequently, an equal volume of 1 mM Zaprinast dissolved in Ringer’s buffer was added to the wells and incubated for 5 min at 37°C and 5% CO2. Uninfected wells were treated with 50 μl of Ringer’s buffer containing 1% Triton X-100 or normal Ringer’s buffer, serving as positive and negative controls, respectively. The released lactate dehydrogenase was centrifuged at 1,000 x g for 5 min twice to pellet insoluble cell debris. Fifty microliters of supernatant were subjected to the standard lactate dehydrogenase release assay as described previously [54]. The egress efficiency of each strain was calculated using the following equation, ([LDH activity derived from individual parasites]− [LDH activity of negative control]) ⁄ ([LDH activity of positive control]−[LDH activity of negative control]).
Twenty-four hours prior to microscopy, 35 mm MatTek dishes (MatTek Corporation) were incubated with 10% fetal bovine serum (FBS) to provide sufficient protein to form a surface conducive to motility. MatTek dishes were washed once with PBS and loaded with 2 ml of Ringer buffer without Ca2+. MatTek dishes were chilled on ice and parasites were added and allowed to equilibrate for 15 min. To remove parasites that have not attached, dishes were washed twice with 2 mL of Ringer buffer without Ca2+. Dishes were then placed in the General Electric Delta Vision environmental chamber set to 37°C and allowed to equilibrate for 5 min. Time-lapse videos were taken and photographed using an Olympus IX-71 inverted fluorescence microscope with a Photometrix CoolSnapHQ CCD camera driven by Delta Vision software. The exposure duration, gain, laser intensity, and filter settings were identical for all videos taken for quantification. After 30 seconds, 1 μM Ionomycin was added to stimulate motility. Tracings were quantified for two different measurements: A) for circular motility, the total numbers of parasites in the field of view were divided by the total number of parasites completing at least one full circle. Data presented were from 6 independent trials. B) for calculating total distance traveled, ImageJ software with the MTrackJ plugin was used to track and calculate distance. We report the average distance of a minimum of three parasites (out of a maximum of 5) from 5–6 independent biological trials.
Intracellular calcium was determined fluorometrically by loading parasites with Fura 2-AM as described previously [55]. Briefly, parasites were harvested, washed with buffer A (116 mM NaCl, 5.4 mM KCl, 0.8 mM MgSO4, 5.5 mM D-glucose, and 50 mM HEPES, pH 7.4) with glucose (BAG), and filtered through a 5 μm filter. Parasites were resuspended to 1 x 109 parasites/ml in BAG supplemented with 1.5% sucrose and 5 μM Fura 2-AM. Parasite suspensions were incubated for 26 min in a 26°C water bath with mild agitation. Cells were washed twice with BAG to remove extracellular dye and were resuspended to a final density of 1 x 109 cells/ml. A 50 μl aliquot of the cell suspension was added to 2.45 ml of Ringer buffer without calcium in a cuvette placed in an Hitachi F-7000 fluorescence spectrometer (Hitachi). The excitations were set at 340 and 380 nm, and emission at 510 nm. The Fura2 response was calibrated from the ratio of 340/380 nm fluorescence values [55]. To determine the initial calcium levels via the chemical indicator, initial calcium readings from parasites were determined from 4 independent trials by averaging the first 50 seconds of each tracing.
The size of the VAC was quantified based on TgCPL or TgCPB staining (both TgCPL and TgCPB are VAC luminal proteases). Freshly purified parasites were inoculated into pre-seeded HFF chamber slides, allowed to invade host cells for 30 min prior to fixation, stained with polyclonal rabbit anti-TgCPL antibody (1:100) or mouse anti-TgCPB antibody (1:500), and VAC diameter measured by immunofluorescence microscopy. The distance of the widest diagonal of TgCPL or TgCPB staining was used as the diameter of the VAC and was quantified using Leica LAS X software. Measurements for at least 50 individual parasites were performed for each replicate in a total of three replicates. The measurements are presented as mean ± SEM (standard error of mean).
WT and TgCRT-deficient parasites were grown in HFF cells for 2 days. The parasites were syringed, filter-purified, and resuspended in ice-cold PBS buffer. A trizol-based method was used to extracted total RNA by using the Direct-zol RNA MiniPrep Plus kit (Zymo). Purified total RNA was converted to sequencing read libraries using the TruSeq Stranded mRNA sequencing kit (Illumina). The prepared libraries were subjected to 2 x 125 bp paired-end Illumina HiSeq2500 sequencing. Each sample was sequenced to a depth of at least 20 million reads. The sequencing reads per sample were trimmed and mapped to the genome of Toxoplasma GT1 strain (release 34) for gene differential expression profiling by the Clemson University Genomics Computational Lab.
Approximately 500 ng of total RNA was used to measure the steady levels of transcripts for individual genes by using the Luna Universal One-Step RT-PCR kit (NEB). The qPCR assay was performed using the BioRad CFX96 Touch Real-Time PCR detection system. The cycle threshold (CT) values for individual genes were used for double delta CT (ΔΔCT) analysis to calculate their relative abundances to that of WT parasites using the Bio-Rad CFX Maestro software. TgActin was used as the housekeeping gene for normalization.
Six- to eight-week-old, outbred CD-1 mice were infected by subcutaneous or intravenous injection with 100 WT or mutant parasites diluted in PBS. The infected mice were monitored for symptoms daily for a total of 30 days. Mice that appeared moribund were humanely euthanized via CO2 overdose, in compliance with Clemson University IACUC’s approved protocol. The seroconversion of the surviving mice was tested by enzyme-linked immunosorbent assay (ELISA). The surviving mice were allowed to rest for 10 days, prior to subcutaneous injection with a challenge dose of 1000 WT parasites, and were monitored daily for survival for additional 30 days.
TgCRT cDNA was PCR-amplified from pTub-TgCRT-mCherry-3xmyc plasmid using a forward primer that introduced a 5’ KpnI site and S. cerevisiae Kozak sequence, and a reverse primer that omitted the mCherry-3xmyc tag and introduced a 3’ XmaJI site. The PCR amplified DNA was digested with KpnI and XmaJI and subcloned into pYES2-6xHis-BAD-V5 (hexa His, biotin acceptor domain, V5 tags) plasmid behind the GAL1 promoter and in front of the His-BAD-V5 epitope tags to generate the plasmid pYES/TgCRT-hbv. To generate the plasmid pYES/PMA-TgCRT-hbv, DNA encoding TgCRT-hbv was PCR amplified using a forward primer that omitted the first 900 bases of TgCRT and introduced a 5’ SacI site, and a reverse primer that included a 3’ NotI site and His-BAD-V5 tags. The amplified DNA was digested with SacI and NotI and subcloned into a SacI/NotI-digested pYES/PfHB3PMA (from [37]; modified via site-directed mutagenesis to introduce a SacI site at the PMA-PfCRT interface). Mutagenesis reactions were performed using reagents obtained from Agilent (Santa Clara, CA).
Isolation of yeast membranes and detection of proteins by Western blot were completed using previously published methods [22].
Quantitative growth rate analysis was used to calculate CQ transport as previously described in detail elsewhere [22,37,39]. Briefly, growth under each condition was measured in duplicate at an initial cell density of OD600 = 0.1 in 96-well plates placed in a Tecan (Durham, NC) M200Pro or BioTek (Winooski, VT) Epoch2 plate reader. CQ-induced growth delays at 80 mM CQ, pH 6.75 were calculated as the difference in time taken to reach maximal growth rate in PMA-TgCRT non-inducing versus inducing media (see [39]).
Statistical analysis was performed using Prism software (GraphPad version 8). The methods used in different assays were indicated in the figure legends. All statistical significance analysis in this paper used the raw data sets for calculation (refer to S2 Text for raw data sets).
|
10.1371/journal.pntd.0003096 | Co-evolution between an Endosymbiont and Its Nematode Host: Wolbachia Asymmetric Posterior Localization and AP Polarity Establishment | While bacterial symbionts influence a variety of host cellular responses throughout development, there are no documented instances in which symbionts influence early embryogenesis. Here we demonstrate that Wolbachia, an obligate endosymbiont of the parasitic filarial nematodes, is required for proper anterior-posterior polarity establishment in the filarial nematode B. malayi. Characterization of pre- and post-fertilization events in B. malayi reveals that, unlike C. elegans, the centrosomes are maternally derived and produce a cortical-based microtubule organizing center prior to fertilization. We establish that Wolbachia rely on these cortical microtubules and dynein to concentrate at the posterior cortex. Wolbachia also rely on PAR-1 and PAR-3 polarity cues for normal concentration at the posterior cortex. Finally, we demonstrate that Wolbachia depletion results in distinct anterior-posterior polarity defects. These results provide a striking example of endosymbiont-host co-evolution operating on the core initial developmental event of axis determination.
| Filarial nematodes are responsible for a number of neglected tropical diseases. The vast majority of these human parasites harbor the bacterial endosymbiont Wolbachia. Wolbachia are essential for filarial nematode survival and reproduction, and thus are a promising anti-filarial drug target. Understanding the molecular and cellular basis of Wolbachia-nematode interactions will facilitate the development of a new class of drugs that specifically disrupt these interactions. Here we focus on Wolbachia segregation patterns and interactions with the host cytoskeleton during early embryogenesis. Our studies indicate that centrosomes are maternally inherited in filarial nematodes resulting in a posterior microtubule-organizing center of maternal origin, unique to filarial nematodes. This microtubule-organizing center facilitates the concentration of Wolbachia at the posterior pole. We find that the microtubule motor dynein is required for the proper posterior Wolbachia localization. In addition, we demonstrate that Wolbachia rely on polarity signals in the egg for their preferential localization at the posterior pole. Conversely, Wolbachia are required for normal embryonic axis determination and Wolbachia removal leads to distinct anterior-posterior embryonic polarity defects. To our knowledge, this is the first example of a bacterial endosymbiont required for normal host embryogenesis.
| The phylum Nematoda comprises up to 1 million species and is one of the most diverse and successful, with members colonizing all possible ecological niches on earth [1], [2]. Nematodes have an extraordinary ability to adapt to the parasitic life style [3]–[6] and as a result exert profound impacts on agriculture and human health. The Spirurina clade contains only animal parasites, among them the Onchocercidae or filarial nematodes [5]. These thread-like worms are tissue-dwelling parasites, transmitted by arthropods, usually black flies or mosquitoes, to all classes of vertebrates except fish. It is estimated that 150 million people are infected with filarial nematodes, with 1 billion living at risk in tropical areas. Filarial nematodes lead to debilitating diseases such as onchocerciasis (caused by Onchocerca volvulus) and lymphatic filariasis (Brugia malayi, Brugia timori, Wuchereria bancrofti) [7]. A total of eight species of filarial nematodes are responsible for these neglected tropical diseases. With the exception of Loa and certain Mansonella sp., all other human filariae harbor an alpha-proteobacterium of the genus Wolbachia. This symbiosis is restricted to the family of Onchocercidae among nematodes [7], [8]. In addition, Wolbachia are also widespread among arthropods [9] and the bacteria of this genus have been classified into different supergroups, as defined by MultiLocus Sequence Typing [10], [11]. The supergroups C and D represent the majority of Wolbachia in filarial species and are restricted to the Onchocercidae [8].
Wolbachia are required for filarial nematode fertility and survival [12] and we previously showed that removal of either supergroup C or D bacteria by antibiotic therapies against O. volvulus or B. malayi leads to extensive apoptosis [13]. Yet little is known about the actual basis of the mutualistic interaction. Genomic analysis and experimental studies suggest that Wolbachia may contribute to metabolic pathways absent or partially missing in the nematode host, including synthesis of riboflavin, nucleotides and hemes [14]–[16]. However, the recent publication of the Loa genome, a Wolbachia-free human filarial parasite, revealed no metabolic compensation for the lack of mutualistic endosymbionts, suggesting caution in drawing conclusions on the basis of the symbiosis from genomic studies [17].
In the vast majority of filarial species, Wolbachia are present in the hypodermal chords of both male and female adult specimens, and in the female germline [8]. This is achieved through both asymmetric segregation during the mitotic divisions and cell-to-cell migration [18]. Immediately following fertilization, Wolbachia concentrate at the posterior region of the embryo. Wolbachia first localize in the posterior germline precursor lineage by rounds of asymmetric segregation until the 12-cell stage. They then reach a hypodermal lineage, and from this subset of posterior hypodermal cells, the bacteria colonize the whole dorsal and ventral hypodermal syncytia during late larval development, spreading toward the anterior of the worm [18]–[20].
Here we focus on the rapid migration and concentration of Wolbachia at the posterior pole immediately during the oocyte-to-embryo transition in B. malayi as this is a key unexplored initial event determining the distribution of Wolbachia in adult tissues. We used C. elegans, the sole well-studied nematode, as a reference for the oocyte-to-embryo transition in B. malayi. Although phylogenetically distant, the free-living and parasitic species both belong to the secernentean nematodes, and share a very similar embryonic development [21] [22] [23] [1]. To identify host factors involved in Wolbachia asymmetric enrichment after fertilization, we first characterized the cytoskeleton of the B. malayi embryo. As described below, we discovered a posterior microtubule-organizing center (MTOC) in the unfertilized mature oocyte. This is in striking contrast to C. elegans, in which the MTOC originates from the sperm-derived basal body/centrosome and induces cytoskeletal asymmetries essential for proper anterior-posterior polarity establishment [24]. Thus centrosome inheritance and its role in anterior-posterior polarity determination are dramatically different in C. elegans and filarial B. malayi. This maternally-derived B. malayi posterior MTOC facilitates Wolbachia concentration in the posterior of the newly fertilized egg. Using immunofluorescence and recently developed RNA silencing techniques [25], we show that host dynein is required for Wolbachia posterior enrichment in the egg. In addition, Wolbachia posterior localization requires B. m. PAR-1 and PAR-3, the B. malayi orthologs of C. elegans polarity-determining proteins Ce PAR-1 and Ce PAR-3. Finally, we demonstrate that Wolbachia removal results in Anterior-Posterior polarity defects, demonstrating for the first time that Wolbachia plays an essential role in these early embryonic developmental events.
Live specimens were obtained from the NIH/NIAID Filariasis Research Reagent Resource Center (www.filariasiscenter.org). To obtain B. malayi adults devoid of Wolbachia, infected jirds were administered tetracycline at 2.5 mg/ml in drinking water (water changed daily) for a period of six weeks, followed by a one week clearance period. While this treatment is enough to deplete Wolbachia from filarial nematodes, the tetracyclin itself does not affect the host gene expression, including mitochondrial genes, as demonstrated by microarray after treatment of A. viteae, afilarial species devoid of Wolbachia [16]. Untreated infected jirds were maintained in a similar fashion as a control. After the clearance period, adult worms were recovered from the peritoneal cavities into preheated (37°C) culture medium RPMI-1640 supplemented with 100 U/ml penicillin, 100 µg/ml streptomycin, 2 mM L-glutamine, 0.25 µg/ml amphotericin B, and 25 mM HEPES (GIBCO).
The Animal Research and Care Program at UWO follows regulations and guidelines established by the USDA Animal Welfare Act, Public Health Service Policy, and the Association for the Assessment and Accreditation of Laboratory Animal Care (AAALAC). The protocol followed has been approved by UWO IACUC. Protocol 0-03-0026-000252-11-22-11, “Oral Tetracycline Treatment of Mongolian Gerbils (Meriones unguiculatus)”. Approval Date: 11/22/11. Expiration Date: 12/9/14 AAALAC #: 001268.
All the B. malayi genes have been identified as C. elegans orthologs by reciprocal BLAST using the NCBI protein BLAST tool (http://blast.ncbi.nlm.nih.gov.gate1.inist.fr/Blast.cgi)
Two peptides were designed for each of B.m. γ-tubulin and B.m. Zyg-9 and were used together to immunize rabbits:
-B.m. gamma tubulin (gene ID: 6105932 Bm1_55245):
(VRETVQTYRNATKPDFIEIN) and (GSHALEKISDRFPKKLVQTY)
-B.m. zyg-9 (gene ID: 6096160 Bm1_06160):
(MHKSNPLKPPAP)
(RSDRSSSRIGRNTHRSNSVSRDSS)
For Dhc-1 a single peptide was used (gene ID: 6103168 Bm1_41435):
(LGGSPFGPAGTGKTESVKAL)
Peptides were synthesized by the Organic Synthesis group of New England Biolabs with an additional N-terminal cysteine residue to facilitate conjugation to the carrier protein KLH using m-Maleimidobenzoyl-N-hydroxysuccinimide ester (MBS; Pierce, Rockford, IL) [21]. Sera were raised in rabbits by Covance Immunology Services, Denver, PA. Peptides were purified essentially according to a published procedure [26]. Antibodies raised against pericentriolar markers (i.e. B.m. gamma tubulin and B.m. zyg-9) co-localize with MTOCs (cf. Fig. 1). In both Spirurida (i.e. B. malayi) and Rhabditina (i.e. C. elegans), chromosomes are holocentric [27]. In B. malayi, the anti B.m. dhc-1 concentrates along the holocentric chromosomes during metaphase as dynein does in C. elegans [28].
All the silencing experiments were performed as already described [25]. Briefly, B. malayi females were soaked in 1 µM of heterogenous short interfering (hsi)RNA mixtures for 48 hours before egg and embryo collection and fixation. PCR primers used to generate the primary dsRNAs contained T7 promoter sequence followed by two guanine bases at their 5′ ends for transcription by T7 RNA polymerase and enhanced transcription yield.
-Par-1(gene ID: 6100834 Bm1_29690)
forward: 5′- TAA TAC GAC TCA CTA TAG GGG AGA GGA ATC TTG CCA ACG G -3′
reverse: 5′- TAA TAC GAC TCA CTA TAG GGA ACT GCT TGT GCA GAT GCG C -3′
-Par-3(gene ID: 6103110 Bm1_41135)
forward: 5′- TAA TAC GAC TCA CTA TAG GGT TCT GGA TCC CGA TGA TCA G-3′
reverse: 5′- TAA TAC GAC TCA CTA TAG GGT AGA CGT GAT TTC CTA GCG G-3′
-Dhc-1 (gene ID:6103168 Bm1_41435)
forward: 5′- TAA TAC GAC TCA CTA TAG GGA GCA ACT GTC AAG GAA AAG -3′
reverse: 5′- TAA TAC GAC TCA CTA TAG GGA TGG AGA CAA GTC GAT ATC C -3′
Embryos were collected, fixed and stained as already described in detail [25]. Polyclonal anti B.m. Zyg-9, anti B.m. gamma tubulin and anti B.m. Dhc-1 were used at a dilution of 1∶100. Microtubule stainings were performed using the monoclonal DM1α antibody raised against α-tubulin (Cell Signaling Technology, Danvers, MA, USA) at a dilution of 1∶100. Cy5 goat anti-rabbit IgG and Alexa Fluor 488 goat anti–mouse IgG antibodies were used at 1∶150 (Invitrogen). Primary and secondary stainings were both performed overnight either at 4°C or room temperature. Actin stainings were performed using the fluorescent Atto 488 phalloidin (Sigma) at a dilution of 1∶100, added with secondary antibodies. The Wolbachia were visualized with propidium iodide (PI). We previously showed that the PI puncta only correspond to Wolbachia DNA by colocalization with Wolbachia-specific antibodies [19]. For propidium iodide (Molecular Probes) DNA staining, embryos were fixed then incubated overnight at room temperature in PBS+RNAse A (15 mg/mL, Sigma), in rotating tubes overnight. PI incubation itself was done after the secondary antibody wash (1.0 mg/mL solution) by simply shaking the eppendorf for 10 seconds in PBS followed by a 5 minute wash. 30 second centrifugations at 4000 rpm in between steps are enough to pellet all embryos. They were then mounted into Vectashield (Vector Laboratories, Burlingame, CA).
Confocal microscope images were captured on an inverted photoscope (DMIRB; Leica Microsystems, Wetzlar, Germany) equipped with a laser confocal imaging system (TCS SP2; Leica) using an HCX PL APO 1.4 NA 63 oil objective (Leica) at room temperature. 3-D movies were generated using the Volocity 3D Image analysis software (PerkinElmer).
Wolbachia have been shown to rely on host microtubules, kinesin and dynein in insects to properly segregate to the posterior germline pole plasm during oogenesis [29], [30]. To establish whether or not Wolbachia transmission also depends on similar cytoskeletal interactions in filarial nematodes, the microtubule network was characterized during the oocyte-to-embryo transition. To follow the microtubules and pericentriolar material (PCM), anti-B.m. γ-tubulin and anti-B.m. Zyg-9 antibodies were generated (Cf. experimental procedures; Fig. 1).
In the free living nematode C. elegans, as in most animal species, centrosomes are degraded during oogenesis, prior to diakinesis [31], [32]. In inseminated females, the cellularized oocyte follows a meiotic maturation phase, under the control of a sperm major protein (MSP) released from the sperm prior to fertilization [33]. During maturation, the germinal vesicle migrates away from the MSP source, with its associated acentriolar spindle, toward the unpolarized oocyte cortex. Centrosomes have a paternal origin and are inherited upon fertilization. The sperm-supplied centrosome participates to establishment of A-P polarity in the zygote, and the entry point defines the posterior pole of the egg [34].
In contrast to C. elegans, the presence of a microtubule-organizing center (MTOC), located at the opposite pole of the germinal vesicle was detected in unfertilized mature meiosis I oocytes from B. malayi. This polar MTOC is defined by both the presence of PCM components γ-tubulin and Zyg-9 proteins, and its ability to nucleate microtubules (Fig. 1, see also Fig. 2A, Movie S1). However, it disappears after fertilization, by the time pronuclei apposition occurs (Fig. 2(A) to (B)). Upon fertilization, no sperm-associated or paternal nucleus-associated MTOC was ever detected (Fig. 3(A) and (B), Movie S1; n>100). At this stage, microtubules do not nucleate at the surface of the paternal pronucleus, suggesting the absence of a paternally-derived MTOC. Rather, the anti-γ-tubulin antibody revealed numerous cytoplasmic foci (Fig. 2(A)). Some of these foci coalesce around the apposed pronuclei to form the MTOCs while the others are gradually degraded (Fig. 2(B) to (D)). This correlates with the microtubule dynamics at this stage (Fig. 4(C) to (E)). Together, these data demonstrate the presence in B. malayi of a MTOC-associated microtubule cytoskeleton in the mature cellularized oocyte, and suggest a maternal de novo origin of centrosomes in filarial nematodes, in contrast to C. elegans (Fig. 2 (E)).
We next examined Wolbachia dynamics in the mature oocyte and early embryo to better understand how they concentrate at the posterior blastomere during the two cell stage. We first characterized their dynamics in zygotes during the 1st cell cycle (Fig. 4, n>100). Prior to, and soon after fertilization (Fig. 4(A) and (B)), Wolbachia are dispersed in the egg, sometimes showing a preference for the meiotic spindle and the opposite pole [19] (see also Fig. 5). The concentration in the posterior half of the egg starts during pronuclei migration and apposition (Fig. 4C), and is achieved by the beginning of prophase (Fig. 4D). This localization is maintained through mitosis (Fig. 4(D) to (J)) and enables the vast majority of endosymbionts to segregate in the posterior blastomere P1 after cytokinesis (Fig. 4K). This posterior segregation pattern is repeated in the dividing two-cell embryo (Fig. 4L). Thus, Wolbachia are asymmetrically localized very early in the zygote, to become enriched at the posterior end before entry into mitosis.
We established that Wolbachia asymmetrically localize in the egg prior to the first mitosis, and are maintained at the posterior pole during mitosis. To further investigate a possible role of the microtubule cytoskeleton in Wolbachia dynamics, we looked for close association between the endosymbionts and microtubule network (Fig. 5, n>100). We found Wolbachia in the vicinity of microtubules emanating from the polar MTOC after fertilization (Fig. 5(A) and (A′)). Later during mitosis, we found Wolbachia organized along the posterior astral microtubules (Fig. 5(B) and (B′)). These data suggest that the microtubule cytoskeleton may be used by Wolbachia first for concentration, second for maintenance at the posterior pole of the egg.
In Drosophila, Wolbachia rely on plus and minus end directed motor proteins for their concentration at the posterior pole of the Drosophila embryo [29], [30]. Our finding that Wolbachia closely localize to microtubules suggests they may concentrate at the posterior pole through their association with microtubule based motor proteins. The polar MTOC projects microtubule plus-ends inward and it was of interest to ascertain whether or not the Wolbachia may use the host minus-end molecular motor Dynein to segregate to the future posterior pole of the egg. To achieve this, the B.m. Dynein heavy chain 1 (B.m. Dhc-1) was silenced by soaking adult females in hsiRNA for 48 hrs [25], (Fig. 6). We collected a vast majority of multinucleated 1-cell eggs, as a result of chromosome segregation and cytokinesis failure when Dynein was reduced or absent. These highly penetrant phenotypes indicates that the hsiRNA is efficiently knocking down the Dynein levels. In these eggs, Wolbachia were evenly distributed in the cytoplasm (cf. Fig. S1). To circumvent the lack of developmental timing information in these eggs, we focused on zygotes prior to entry into the first mitosis (n = 10). In wild-type eggs, the majority of bacteria are at the posterior pole (n>100). In contrast, upon B.m. dhc-1 hsiRNA treatment, they no longer distribute asymmetrically (Fig. 6(A) and (B)).
To test a putative direct interaction between Wolbachia and Dynein, we raised an antibody against the B.m. dhc-1. Similar to studies in C. elegans [35], the anti-Dynein antibody decorates the condensed chromosomes in the zygote (Fig. 6C arrowhead). Significantly Dynein also colocalizes with posterior localized Wolbachia (Fig. 6(C) and (C′), arrow). This strongly suggests that Wolbachia may use the host Dynein and the polar MTOC for their initial asymmetric enrichment.
In B. malayi, after pronuclei apposition, the polar MTOC is no longer present in the egg. What then keeps Wolbachia in the posterior until the first division takes place? We tested the influence of Anterior-Posterior (A-P) polarity establishment in Wolbachia localization and maintenance.
Establishment of A-P polarity has been extensively studied in zygotes of the free living nematode C. elegans. In this species, symmetry breaking is triggered by sperm entry [34]. A remodeling of the cortical cytoskeleton is associated with a redistribution of the PARs polarity cues, as well as intense cytoplasmic streaming, to form an anterior and a posterior cortical domain by the beginning of mitosis. Subsequently, downstream polarity effectors are required to establish an asymmetric division [36].
To test whether PARs-induced symmetry breaking mechanisms dictate the bacteria asymmetric distribution, the B. malayi orthologs of C. elegans posterior PAR-1 and anterior PAR-3 were identified and silenced by hsiRNA. Due to the relatively low penetrance of the PAR-1 and PAR-3 hsiRNA phenotypes (∼30%, n>100 in both cases), we focused on dividing two cell embryos which showed classic PAR polarity-defect phenotypes: synchronous mitotic divisions and abnormal spindle orientation [37]. In wild-type B. malayi and C. elegans two-cell embryos, the anterior AB blastomere enters mitosis before the posterior P1 blastomere (Fig. 7A). This asynchrony is even more pronounced in B. malayi, where three-cell embryos, composed of AB daughters and dividing P1, are commonly observed. Also in B. malayi, like C. elegans, the posterior P1 spindle rotates by 90° to align along the A-P axis, while the AB spindle remains transverse (Movie S2). As in C. elegans, hsiRNA knockdown of either par1 or par3 disrupts the normal mitotic asynchrony between the two B. malayi blastomeres. In addition, upon B. malayi par-1 hsiRNA, the P1 spindle fails to rotate (Fig. 7A, Movie S3), while upon B. malayi. par-3 hsiRNA treatment, the AB spindle now rotates to align along the long (A-P) axis of the embryo (Fig. 7A). These timing and spindle orientation defects are strikingly similar to those observed in C. elegans [37] and reveal at least partial evolutionary conservation of functions for B. malayi PAR-1 and PAR-3. The presence of these polarity defects correlates with a loss of Wolbachia asymmetric segregation or maintenance at the posterior pole (Fig. 7B). This indicates that the A-P polarity determinants are essential for the stable enrichment of Wolbachia in the posterior P1 blastomere.
By the first mitotic division, Wolbachia are predominantly concentrated in the posterior half of the B. malayi egg. In the C. elegans zygote, the complete establishment of the anterior and posterior cortical domains is already achieved by the beginning of mitosis [38]. As it is likely that A-P polarity set up in B. malayi takes place no later than in C. elegans, it was of interest to determine whether Wolbachia might influence the A-P polarity in the zygote. To investigate this, we analyzed A-P polarity in normal and Wolbachia-depleted two-cell embryos (cf. Experimental Procedures, [13]). This analysis yielded the following phenotypic classes (Fig. 8A): Class I included those with normal division patterns exhibiting mitotic division asynchrony and proper spindle orientation. Class II included those with “posterior polarity” defects exhibiting a failure of P1 spindle rotation and division synchrony, and Class III included those with “anterior polarity” defects exhibiting inappropriate rotation of the AB spindle and division synchrony. The vast majority of wild-type embryos (97%, n = 75) showed class I normal division patterns (Fig. 8(A) and (B)). Embryos devoid of Wolbachia (n = 27) displayed a dramatic loss of normal class I division patterns (48%). The remaining half of embryos lacking Wolbachia displayed either Class II posterior defects (40%, Fig. 8(B) and Movie S4) or Class III anterior (11%) defects. These results reveal that Wolbachia not only rely on A-P polarity cues for their posterior location but also are essential for proper establishment of AP polarity in its filarial nematode host.
Centrosome inheritance is asymmetric in metazoan sexual reproduction. Usually, but not always, centrosomes are degraded in the female germline and provided paternally through the transformation of the sperm-derived basal body. This mechanism of inheritance ensures a tight control of centrosome number and MTOCs in the zygote, [39]. A dramatic exception to the typical pattern of paternal centrosome inheritance occurs in parthenogenetic development of unfertilized eggs in Hymenoperta. In this case, centrosomes and their associated MTOCS are derived exclusively from maternally derived components [40]–[42]. Our studies demonstrate a third unique centrosome/MTOC inheritance pattern in B. malayi. First, the unfertilized mature oocyte contains a maternal-derived MTOC. Second, despite fertilization, centrosomes appear to be produced de novo and to be maternally supplied. Accordingly, no paternally-derived MTOC was observed associated with the paternal chromatin after sperm entry.
Whether or not the maternal MTOC originates from a centrosome remains to be determined, since acentrosomal PCM has been shown to nucleate microtubules in vitro [43]. In any case, this maternal MTOC never interacts with the paternal chromatin and is degraded soon after fertilization. We find that during pronuclei apposition, the PCM component γ-tubulin accumulates around the nuclear envelopes as foci, and this correlates with microtubule enrichment at the nuclear surface. The presence of functional MTOCs capable of microtubule nucleation is only observed after entry of the pronuclei into mitosis. Together, these findings suggest centrosomes are derived exclusively from maternal components and perhaps form de novo in filarial nematodes. New centrosomal markers will be required to identify the origin and composition of the polar MTOC.
These findings also raise important questions regarding the mechanism of symmetry breaking and polarity establishment in filarial nematode embryos. In C. elegans, the paternally supplied centrosome and its associated MTOC play a crucial role in polarity establishment. The sperm derived centrosome/MTOC elicits a dramatic reorganization in the actomyosin cortical network and asymmetric localization of polarity components such as PAR-1 [34], [44]. It is currently unclear how much of a role the maternally-derived MTOC or fertilization plays in symmetry breaking and polarity establishment in the B. malayi embryo. The design of much needed new reagents suitable for filarial species will help us to understand the great variations on fundamental mechanisms between the free living C. elegans and filarial nematode species. This may help us to better understand peculiarities of the parasitic lifestyle, and sources of such evolutionary divergence.
In insects, Wolbachia must navigate the constantly changing cytoskeletal environment of the oocyte in order to concentrate at the posterior pole where the germline will form. Wolbachia rely on host microtubules for their transport through the oocyte. Early in oogenesis they rely on the plus-end motor protein kinesin. Later, the microtubules reorganize and reverse orientation requiring Wolbachia to engage dynein to complete their poleward journey. The studies presented here indicate that Wolbachia in B. malayi are also very likely to rely on microtubules and motor proteins for their asymmetric concentration in the posterior pole of the embryo. Unlike in C. elegans, prior to fertilization B. malayi oocytes possess a robust posteriorly positioned MTOC with microtubules emanating towards the anterior positioned meiotic spindle. Upon fertilization, the Wolbachia associate with microtubules and concentrate at this unusual posteriorly positioned MTOC. We also observe a striking co-localization between Wolbachia and the host dynein heavy chain. Significantly functional RNAi analysis demonstrates that dynein is required for this posterior enrichment. Thus in both insects and filarial nematodes dynein mediated movement is required for the asymmetric posterior positioning of Wolbachia to ensure germline incorporation.
Although the posterior MTOC is established prior to fertilization, fertilization is required for the posterior concentration of Wolbachia. We believe that the maternally supplied posterior MTOC contributes to the initial Wolbachia concentration at the posterior pole. However it appears that maintenance of Wolbachia at the posterior pole requires cytoplasmic rearrangements mediated by fertilization, such as the asymmetric cortical localization of PAR polarity cues, controlling an asymmetric dynein activity at the cortex.
Unlike many intracellular bacteria, Wolbachia have no flagellum, and do not appear to rely on the actin cytoskeleton for intracellular transport. As with Drosophila, Wolbachia in B. malayi concentrate near, and perhaps associate, with microtubules [29]. Upon fertilization in C. elegans, the sperm brings a basal body giving rise to male pronucleus-associated MTOCs, establishing the posterior of the egg [34]. A similar mechanism in filarial nematodes would have explained an early, microtubule-based movement of Wolbachia toward the posterior of the embryo. However the fertilization mechanisms and remodeling of the cytoskeleton during this step appear dramatically different in B. malayi.
Why is fertilization then needed to achieve the asymmetric enrichment, if the polar MTOC is already present in the oocyte? A simple model taking into account the microtubule cytoskeletal peculiarities of the filarial zygote could be envisioned (Fig. 9). A maternal polar MTOC projects microtubules inward, while meiosis is resumed at the opposite pole during fertilization (Fig. 9 I to II). In turn the polar MTOC is degraded and absent by the time pronuclei appose (Fig. 9 III), followed by entry into mitosis and set up of the mitotic spindle (Fig. 9 IV). After fertilization, when the meiotic spindle is no longer present, the bacteria concentration is preferentially displaced toward the MTOC. Cell cycle progression may also alter Wolbachia interaction with the Dynein complex, or its activation, resulting in more engagement on the microtubules [45]. At pronuclei apposition, Wolbachia are in the posterior compartment, most of them associated with the most posterior pronuclear envelope (paternal), but also in the cytoplasm, and in contact with the posterior cortex (Fig. 9 III, i.e. Fig. 2 (B) and (C)). The association with the nuclear membrane correlates with a perinuclear accumulation of γ-tubulin foci (Fig. 2C). Dynein is known to anchor the MTOC to the paternal nuclear envelope in C. elegans [46][35]. This motor may play a role in centrosome biogenesis and recruitment to the nuclear envelope in B. malayi, and may also mediate this Wolbachia localization. Cortical dynein has also been shown to play a crucial role in spindle positioning in C. elegans [47], and Wolbachia cortical posterior localization could be mediated by the dynein itself. Once mitosis is triggered, whether Wolbachia interact with the astral microtubules or the cortex through dynein and/or other host factors, they remain trapped in the posterior compartment until cytokinesis occurs, and eventually segregate into the posterior blastomere (Fig. 9 IV, Fig. 4).
In the C. elegans two-cell stage embryo, symmetry breaking mechanisms similar to those observed in the zygote lead to a polarized P1 [24]. Wolbachia asymmetric pattern of segregation is perfectly repeated when P1 divides (Fig. 4L), confirming the importance of host A-P polarity signals in Wolbachia distribution in the early embryo. No polar MTOC is however required in P1 to achieve the same segregation observed in the zygote P0. It is interesting that Wolbachia has co-evolved to adapt to a microtubule dynamics and architecture unique to fertilization in filarial nematodes. This peculiar de novo centrosome inheritance raises many important questions regarding the filarial oocyte-to-embryo transition.
There are now a number of examples in diverse phyla in which bacteria have a profound influence on metazoan development [48]. For example, mice raised in a germ-free environment, exhibit defects in the enteric nervous system regulating gastrointestinal function [49]. Another striking example of animal bacterial interactions occurs in the Squid- V. fischeri symbiosis. The V. fischeri bacteria are required for proper development and morphology of the light organ of the squid. The bacteria induce very specific changes in cell size, morphology and microvilli formation [50].
Our analysis of the Wolbachia-B. mayali symbiosis provides a unique example in which the bacteria are required for normal host axis formation and embryonic development. B. malayi and C. elegans share similar division patterns during early embryogenesis, with AB dividing first, while in the posterior germline precursor P1, the spindle rotates to align along the long A-P axis. These traits are common among the nematode species so far examined [51]. Without Wolbachia, A-P polarity establishment is compromised in the filarial zygote, as revealed by division timing and spindle orientation defects at the two-cell stage, a hallmark of A-P polarity defects in nematode species.
How do the endosymbionts influence A-P polarity? Since Wolbachia concentrate to the posterior before mitosis in B. malayi, (a stage prior to establishment of A-P cortical domains in C. elegans), it is possible that Wolbachia directly influence localization and/or activation of B. malayi posterior polarity cues (i.e. PARs), or on downstream posterior polarity effectors. Conversely, our experiments silencing B.m par-1 and par-3, result in a failure of Wolbachia to become posteriorly enriched indicating that the PAR proteins are required for proper Wolbachia localization. In Drosophila, Wolbachia also associate with polarity determinants. Wolbachia closely associates with the Gurken polarity complex in the Drosophila oocyte and its titer regulated by Gurken levels. Significantly an overabundance of Wolbachia disrupts Gurken function [52].
The pioneering work of Sander in the 1950's demonstrated that displacing the ball of endosymbionts present in the leaf hopper Euscelis plebejus embryo from the posterior to a more anterior position produced ectopic posterior structures. This demonstrated a close association with posterior patterning determinants [53]. In nematodes Wolbachia not only rely on key host polarity factors for their germline transmission, but have become essential for the proper functioning of these determinants.
At this point, however, we cannot rule out a non cell-autonomous explanation for the effect of Wolbachia-depletion on host A-P polarity. Unlike in C. elegans, B. malayi embryogenesis takes place entirely in the female uterus, where the growth of the embryo is dependent on maternal nutrients acquired from the hypodermis [2], [19], [54]. In addition, the endosymbionts fill the hypodermal tissues, a major site for nutrient storage and metabolism in filarial nematodes, and this bacterial population is also cleared upon antibiotic treatment [13]. Thus, it is then possible that Wolbachia removal from the hypodermis leads to metabolic defects affecting a plethora of signaling pathways, including the embryonic polarity set up. A better understanding of symmetry breaking mechanisms in these parasitic nematodes will help us establish precisely how Wolbachia influence embryonic polarity.
In conclusion, we have shed light on the symbiosis mechanisms underlying Wolbachia transmission in the filarial embryo. They suggest a reciprocal dependence between the host and the symbiont starting as early as in the egg, explaining the success of antifilarial antibiotic therapies targeting Wolbachia, leading to massive embryogenesis defects.
|
10.1371/journal.pntd.0005769 | Yeast-expressed recombinant As16 protects mice against Ascaris suum infection through induction of a Th2-skewed immune response | Ascariasis remains the most common helminth infection in humans. As an alternative or complementary approach to global deworming, a pan-anthelminthic vaccine is under development targeting Ascaris, hookworm, and Trichuris infections. As16 and As14 have previously been described as two genetically related proteins from Ascaris suum that induced protective immunity in mice when formulated with cholera toxin B subunit (CTB) as an adjuvant, but the exact protective mechanism was not well understood.
As16 and As14 were highly expressed as soluble recombinant proteins (rAs16 and rAs14) in Pichia pastoris. The yeast-expressed rAs16 was highly recognized by immune sera from mice infected with A. suum eggs and elicited 99.6% protection against A. suum re-infection. Mice immunized with rAs16 formulated with ISA720 displayed significant larva reduction (36.7%) and stunted larval development against A. suum eggs challenge. The protective immunity was associated with a predominant Th2-type response characterized by high titers of serological IgG1 (IgG1/IgG2a > 2000) and high levels of IL-4 and IL-5 produced by restimulated splenocytes. A similar level of protection was observed in mice immunized with rAs16 formulated with alum (Alhydrogel), known to induce mainly a Th2-type immune response, whereas mice immunized with rAs16 formulated with MPLA or AddaVax, both known to induce a Th1-type biased response, were not significantly protected against A. suum infection. The rAs14 protein was not recognized by A. suum infected mouse sera and mice immunized with rAs14 formulated with ISA720 did not show significant protection against challenge infection, possibly due to the protein’s inaccessibility to the host immune system or a Th1-type response was induced which would counter a protective Th2-type response.
Yeast-expressed rAs16 formulated with ISA720 or alum induced significant protection in mice against A. suum egg challenge that associates with a Th2-skewed immune response, suggesting that rAS16 could be a feasible vaccine candidate against ascariasis.
| Roundworms (Ascaris) infect more than 700 million people living in poverty worldwide and cause malnutrition and physical and mental developmental delays in children. As an alternative or complementary approach to global deworming, a pan-anthelminthic vaccine is under development that targets ascariasis in addition to other human intestinal nematode infections. Towards this goal, two Ascaris suum antigens, As16 and As14, were expressed in Pichia pastoris as recombinant proteins. Mice immunized with rAs16 formulated with ISA720 adjuvant produced significant larva reduction (36.7%) and stunted larval development against A. suum egg challenge. The protection was associated with predominant Th2-type responses characterized by high levels of serological IgG1 (IgG1/IgG2a > 2,000) and Th2 cytokines, IL-4 and IL-5. A similar level of protection was observed in mice immunized with rAs16 formulated with alum that induces mainly a Th2-type immune response, whereas mice immunized with rAs16 formulated with MPLA or AddaVax, both inducing major Th1-type responses, were not significantly protected against A. suum infection. High-yield expression of rAs16 in yeast will allow for large-scale manufacture, and its protective efficacy when formulated with alum suggests its suitability as a vaccine candidate.
| Ascaris lumbricoides, Trichuris trichiura and the hookworm Necator americanus are the three major soil-transmitted helminths (STH) that infect more than one billion poor people in the world and are the leading neglected tropical diseases (NTDs) in terms of disability-adjusted life years (DALYs) [1]. New estimates from the Global Burden of Disease Study 2015 indicate that approximately 761 million people are chronically infected with A. lumbricoides, resulting in 2,700 annual deaths from ascariasis [2, 3]. Global control of STH infections depends on the mass drug administration of anthelminthics such as albendazole or mebendazole targeting children between the ages of 1–14 years [4]. However, the rapid rates of post-treatment re-infection [5], potential drug resistance [6], low treatment coverage for children [7], and low access to clean water [8] compromise the effect of anthelminthics alone as a suitable means to control or eliminate STH infections. Indeed, two systematic reviews have largely failed to confirm the beneficial effects of periodic deworming [9, 10]. Thus, the development of a multivalent pan-anthelminthic vaccine targeting all three major STH infections would be a desirable biotechnology to prevent parasite reinfection and advance efforts for the control and elimination of these diseases [11].To advance such a strategy, two major N. americanus hookworm vaccine antigens are undergoing clinical vaccine tests, but there is a need to simultaneously develop A. lumbricoides and T. trichiura candidate antigens as suitable vaccines to be integrated within the human hookworm vaccine development program [12].
With regard to the development of A. lumbricoides vaccine antigens, the genetically highly homologous pig parasite, Ascaris suum, is commonly used as a model to identify and evaluate vaccine candidates. A. suum and A. lumbricoides are morphologically, immunologically, and genetically very similar [13, 14] and might even be subspecies variants. Indeed, A. suum has been shown to be an important cause of human ascariasis [15]. Similar to its natural host, the pig, mice can be infected with A. suum eggs and larvae will be released into the mouse intestine from which they will migrate to the lungs. However, in mice, these larvae cannot return to the intestine to develop into adult Ascaris worms [16–18]. Nonetheless, the mouse model has proven to be a valuable tool in the identification and evaluation of vaccine candidates against Ascaris infections [19–21]. Indeed, mice infected with A. suum eggs produced significant protection against A. suum egg challenge as judged by the significant reduction in the number of larvae migrating to the lungs or livers [19, 20, 22] and also by reduced lung pathology [23]. Using serum from infected mice or rabbit, several immunodominant antigens have been identified including As16 [20], As14 [19], As24 [24], As37 [21] and As-Enol (enolase) [25], and protective immunity has been induced by immunization with recombinant proteins [19, 20, 26, 27] and with DNA [28].
As14 and As16 were the first two antigens previously identified by the Tsuji lab through immunoscreening of an A. suum cDNA library with serum from Ascaris infected rabbit [19, 20]. They share 47% amino acid sequence identity and similar localization (in larva and adult stages, as well as in excretory/excretory products) [11]. Intranasal immunization with Escherichia coli expressed recombinant As16 and As14 conjugated with the cholera toxin B subunit (CTB) produced significant protection against A. suum infective egg challenge in mice [19, 20]. In addition, rAs16 induced protection in a pig animal model [29], and mice fed with As16-transgenic rice mixed with CTB were also protected against A. suum infection [30]. As14 fused with CTB was also successfully expressed in transgenic rice, but there was no oral immunization and protection reported [31]. Notably though, without CTB as the adjuvant, neither As16 nor As14 were able to induce protective immunity in any model.
Here, we report the production of recombinant As16 and As14 in the yeast Pichia pastoris, a eukaryotic expression system with scalability and without the concern for endotoxin contamination as in E. coli-expressed proteins [32]. Mice immunized with yeast-expressed rAs16 mounted a Th2-biased immune response and showed significant protection in terms of lung larval reduction. However, the same immunization regime for rAs14 did not induce any protection in mice. The immunological mechanisms underlying rAs16-induced protection were evaluated compared to rAs14 that did not induce protection. The results in this study provide a feasible approach to developing a vaccine against ascariasis on the basis of the yeast-expressed rAs16 that can be produced at low cost and formulated with alum, an FDA-approved adjuvant for human use that induces the Th2 immune response necessary to achieve anti-Ascaris immunity.
All animal procedures were conducted in accordance with Baylor College of Medicine Institutional Animal Care and Use Committee (IACUC) approved protocol AN-6297 in compliance with the Animal Welfare Act, PHS Policy, and other Federal statutes and regulations relating to animals and experiments involving animals.
A. suum eggs were originally obtained from an adult female worm collected from an infected pig at a pig slaughter house near Belo Horizonte, Brazil, and maintained in 0.2 N H2SO4 until most of them had developed into the embryonated infective stage (50–250 days). The infective embryonated eggs were shipped to our lab in Houston and used to orally challenge BALB/c mice as previously described [17]. The A. suum larvae hatch in the mouse intestine, and then migrate to the liver and lungs. The number of larvae recovered from mouse lung tissue eight days post infection were used as a biomarker to evaluate vaccine efficacy [17]. Crude extracts of A. suum eggs and lung-stage larvae were prepared by homogenization and sonication, and the insoluble pellet was removed by centrifugation as previously described [33].
Amino acid sequences were aligned using CLUSTAL W and prepared for display using BOXSHADE. The phylogenetic trees were generated for As16 and its homologues from different nematodes using Phylogeny.fr [34] (http://www.phylogeny.fr/index.cgi).
DNA coding for As16 without its signal peptide was codon optimized for expression in yeast and synthesized by GenScript. The DNA coding for As14 without its signal peptide was PCR amplified with As14 specific primers from A. suum larvae cDNA reverse-transcribed from total larval RNA. As16 and As14 coding DNAs were subcloned into the yeast expression vector pPICZαA (ThermoFisher Scientific, Carlsbad). The correct sequences and reading frames of the recombinant plasmids were confirmed by double-stranded DNA sequencing using vector flanking primers, α-factor and 3’-AOX1. The recombinant As16 and As14 (rAs16 and rAs14) with a hexahistidine tag at its C-terminus were expressed in yeast stain P. pastoris X-33 under induction with 0.5% methanol for 48–72 hours and then purified by immobilized metal ion affinity chromatography (IMAC), as described previously [35]. The purity of the recombinant proteins was determined by SDS–PAGE. The protein concentration was measured using BCA (ThermoFisher Scientific, Waltham) and Endotoxin clearance was confirmed using the Charles River Endosafe-PTS system (Charles River, Houston).
Six-week old female BALB/c mice were purchased from Taconic and divided into four groups of 20 animals each. Two vaccine groups were immunized subcutaneously with 50 μg of rAs16 or rAs14 emulsified with the adjuvant Montanide ISA720 (Seppic, Paris, France) in a total volume of 100 μl (antigen/ISA720 = 30/70 v/v). Mice were boosted twice with the same dose on days 21 and 35. The control groups were injected with PBS or PBS+ISA720 using the same regimen. Two weeks after the final vaccination, 5 mice from each group were sacrificed and blood and splenocytes were harvested for immunological tests. The remaining 15 mice from each group were challenged with 2,500 A. suum embryonated eggs in a total volume of 100 μl, administered by oral gavage. Eight days after infection, all infected mice were sacrificed, lungs were harvested, and A. suum lung-stage larvae were collected using a Baermann apparatus, as previously described [17]. Reduction in larval burden was calculated in all groups and the results were compared between the vaccine groups and the PBS and adjuvant control groups.
To improve the protection induced by rAs16 and interpret the immunological mechanism underlying the As16-induced protective immunity, another vaccine trial was performed by formulating 25 μg of rAs16 with either 200 μg of Alhydrogel (Brenntag, Mülheim, Germany), 20 μg of MPLA (InvivoGen, San Diego), or 50 μl of AddaVax (50/50, v/v) (InvivoGen, San Diego), each administered subcutaneously in a total volume of 100 μl per mouse given using the immunization regimen described above. Control groups were given adjuvant only. As a positive control, one group of 20 mice was orally infected three times with 1,000 A. suum embryonated eggs. After these immunizations, all mice were challenged with 2,500 A. suum embryonated eggs.
Sera from all blood samples were isolated and frozen at -20°C. The sera samples were assayed for antigen-specific IgG isotypes (IgG1, IgG2a) by a modified indirect enzyme-linked immunosorbent assay (ELISA). Briefly, individual wells of Nunc-Immuno Maxisorp plates (Thermo Scientific, Waltham) were each coated with 100 μl of rAs16 (3.1 μg/ml) or rAs14 (0.39 μg/ml) in coating buffer (KPL, Milford) overnight at 4°C based on the pretested optimal signal/noise ratio. The coated plates were blocked overnight with 0.1% BSA in PBST (PBS +0.05% Tween-20), then incubated with diluted serum samples, starting at 1:200 in 0.1% BSA in PBST for 2 hours. Horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG1 and IgG2a (Lifespan Biosciences, Seattle) were used as secondary antibodies (1:4,000 in PBST). Sure Blue TMB (KPL, Milford) was added as the substrate. The reaction was stopped by adding 100 μL of 1 M HCl. The absorbance was measured at 450 nm using a spectrophotometer (BioTek, Winooski).
Samples including crude extracts of A. suum lung-stage larvae and eggs and the recombinant proteins were separated by SDS–PAGE, then transferred onto PVDF membrane (ThermoFisher, Waltham). After blocking with 5% (w/v) skim milk powder in PBST, the membrane was incubated with sera from mouse immunized with recombinant proteins or A. suum eggs. HRP-conjugated goat anti-mouse IgG (Invitrogen, Carlsbad) was used as a secondary antibody. The antibody recognized bands were developed by ECL (GE Healthcare, Chicago). Recombinant Tc24 protein, expressed in yeast [36], was used as a negative control.
Spleens were obtained from mice two weeks after the third immunization and the splenocytes were disassociated using a 100 μm cell strainer. The cells were then suspended in complete RMPI medium containing 10% heat inactivated FBS and 1x pen/strep solution. After being centrifuged at 300 x g for 5 min, the cells were resuspended in 2 mL ACK lysis buffer (Thermo Scientific, Waltham) for 5 min. After centrifugation the splenocytes were resuspended in complete RPMI media containing 10% DMSO and stored in liquid nitrogen until use.
For the cytokine stimulation assay, splenocytes were thawed in a 37°C water bath and transferred to 5 mL pre-warmed complete RPMI. Cells were washed once to remove residual DMSO. Splenocytes were seeded in a 96-well U-bottom culture plate (Falcon, Corning) at 1x106 cells per well in 250 μl medium and re-stimulated with either 25 μg/mL rAs14 or rAs16 at 37°C, 5% CO2 for 48 hours. Positive controls were stimulated with 20 ng/mL PMA and 1 μg/mL Ionomycin, and unstimulated negative control cultures were performed concurrently. After 48 hours the cells were pelleted by centrifugation at 300 x g for 5 min and the supernatants were collected for measuring cytokine production. The supernatant samples were tested for levels of IL-2, IL-4, IL-5, IL-10, IL-12(p70), GM-CSF, IFN-γ and TNF-α using a Bio-Plex Pro Mouse Th1/Th2 8-plex kit (Bio-Rad, Hercules). To save material and costs, and to increase the sensitivity of the experiment, the kit was used in combination with DA-Bead plates (Curiox Biosystems, Singapore) as previously described [37]. Samples were run on a Bio-Plex Magpix multiplex reader according to manufacturer's recommendations (Luminex, Austin). Raw Luminex data were analyzed using the Bio-Plex Manager 6.0 software and plotted in GraphPad Prism 6.0. The cutoffs of the cytokine standards were dependent on the lot number of the Bio-Rad kit. To remove individual baseline cytokine values, cytokine values from non-restimulated samples were subtracted from those associated with antigen restimulated samples.
Statistical significance of differences between groups was determined using a Mann-Whitney test using Prism 6. In Fig 5c the groups receiving the antigen + adjuvant were compared to the associated adjuvant alone group, the A. suum egg group, and to the PBS group using a Fisher’s LSD test. Data was presented as means ± standard deviation. For the statistical analysis, p < 0.05 was considered to be statistically significant.
The genes encoding As16 (yeast codon optimized) and As14 (native sequence) without their signal peptides were cloned into the Pichia expression vector pPICZαA. The hexahistidine-tagged rAs16 and rAs14 proteins were expressed in P. pastoris X-33 through induction with 0.5% methanol over 72 hours and then purified by immobilized metal ion affinity chromatography (IMAC). The purified rAs14 and rAs16 proteins were analyzed by SDS–PAGE (Fig 1A). The apparent molecular weight for both proteins was approximately 15.0 kDa, which corresponds well to the sizes of the predicted gene products (14.8 kDa for rAs14 and 15.4 kDa for rAs16).
To determine whether rAs14 or rAs16 were recognized by protective immune sera from mice repeatedly infected with A. suum eggs described below (Protective immunity induced by immunization with rAs16), the infected mouse sera were used for Western blot. We observed that only rAs16 was recognized by the infected mouse sera, while rAs14 was not recognized by the same sera (Fig 1B). Even though the two proteins share 47% identity and 66% similarity in sequence (Fig 1E), there was no obvious immunological cross reaction between them when using rAs16 or rAs14 immunized mouse sera individually (Fig 1C and 1D). The results suggest that As16 antigen was exposed to the immune system during A. suum egg infection and larval migration, whereas As14 antigen may not have been immunologically accessible.
As16 is a 16 kDa nematode specific protein found among several different nematode species (Fig 2). It is present in different developmental stages of A. suum, including larvae and adult worms, but its function remains unknown [20]. As16 is highly conserved in Ascaris spp.; it shares 94% sequence identity with its counterpart in the human parasite A. lumbricoides (Al-Ag2), suggesting the possibility of achieving cross-protection for both Ascaris species if As16 were to be used as a vaccine antigen. Its homologues in filarial worms are ranked among the leading vaccine candidates against human onchocerciasis (Ov-RAL-2) [38] and Brugia malayi filarial infections (Bm-RAL-2) [39]. The As16 homologue from the canine hookworm (Ancylostoma caninum, Ac-16) also protected dogs from blood loss and reduced worm fecundity [40], and the Baylisascaris schroederi homologue, Bs-Ag2, protected giant pandas from infection with that parasite [41].
Immunization with rAs16 and rAs14 formulated with the ISA720 adjuvant elicited significant titers of antigen-specific IgG1 and IgG2a antibodies in mice, with IgG1 as the predominant subclass (Fig 3A). The IgG1/IgG2a ratio after As16 immunization (2662:1) was more than 100-fold higher than the ratio for As14 (206:1) (Fig 3A), suggesting that a predominant Th2-type immune response occurs to both antigens, but in particular to rAs16. Mice given PBS+ISA720 did not show any IgG isotype responses specific to rAs16 or rAs14.
To evaluate the cytokine profiles induced by immunization with rAs16 and rAs14, mice were sacrificed two weeks after the final immunization and their splenocytes were isolated. Cytokine profiles were determined by measuring levels of IL-2, IL-4, IL-5, IL-10 and IFN-γ in supernatants of splenocytes re-stimulated with 25 μg/ml rAs16 or rAs14 for 48 hours. Signal background in blank media was subtracted from re-stimulated samples. Statistically significantly increased levels of the cytokines IL-2, IL-4, IL-5, IL-10 were detected in the supernatants of re-stimulated splenocytes from mice immunized with rAs16 and rAs14; however, IFN-γ was significantly increased only in mice immunized with rAs14, not in the mice immunized with rAs16 (Fig 3B). Splenocytes from the PBS +ISA720 control group did not show any detectable cytokine expression.
All mice were orally challenged with 2,500 A. suum embryonated eggs two weeks after the final immunization. A. suum larvae were collected from the lungs of immunized mice eight days after egg challenge. The lung larva count showed that mice immunized with 50 μg of rAs16 formulated with ISA720 adjuvant showed a 36.7% larva reduction compared to the adjuvant-only control groups, constituting a statistically significant difference (p<0.001) (Fig 4A). In addition, the size of the larvae collected from rAs16 immunized mice was much smaller than those collected from adjuvant control mice, suggesting developmental stunting due to the vaccination (Fig 4B). However, mice immunized with rAs14 did not show any protection in terms of reducing the number of larvae found in the lungs, or affecting the size of the larvae.
To further understand the immunological mechanism underlying the protective immunity induced by rAs16, and to select an adjuvant that performs best in protecting mice from infection, mice were immunized with only half the amount of As16 (25 μg, compared to 50 μg of rAs16 used for formulation with ISA720), formulated with three different adjuvants (Alhydrogel, MPLA and AddaVax). As a positive control, another group of mice was immunized through three trickle infections with 1,000 A. suum eggs. After three immunizations, all mice were challenged with 2,500 A. suum embryonated eggs. Mice immunized with 25 μg of rAs16 formulated with Alhydrogel, an alum adjuvant inducing a predominant Th2-type response, experienced a 38.9% lung larval reduction, which is statistically significant compared to the PBS and adjuvant-only control groups. However, rAs16 formulated with MPLA, a TLR4 agonist inducing a Th1/Th2-mixed type immune response, induced only a 26.1% lung larval reduction that was not statistically different from the control groups. Mice immunized with rAs16 formulated with AddaVax, an oil-in-water based adjuvant similar to the MF59 adjuvant that is licensed for flu vaccines in Europe and known to also induce a Th1/Th2 mixed immune response, were not statistically significantly protected against the A. suum egg challenge. Strikingly, mice immunized with three trickle infections with 1,000 A. suum eggs produced almost sterile immunity (99.6% lung larval reduction) against A. suum challenge (Fig 5A).
Serological antibody measurement revealed that mice immunized with rAs16 formulated with three different adjuvants all produced high titers of IgG1 and IgG2a, with a bias towards IgG1 (IgG1/Ig2a = 120–202) (Fig 5B). Interestingly, mice repeatedly infected with a low-dose of A. suum eggs produced a 99.6% lung larva reduction and also showed a significant increase of anti-As16 specific IgG1 antibodies, but without any accompanying IgG2a response. This suggests native As16 is released and exposed to the host immune system during repeated low-dose A. suum infections and may be involved in the induction of the observed Th2 protective immunity. Cytokines released by splenocytes upon re-stimulation of rAs16 showed that immunization with rAs16 formulated with Alhydrogel induced the release of IL-5, a major cytokine linked to Th2 responses in mice, and some level of IL-12 and GM-CSF. There were no detectable levels of IL-2, IL-4, IL-10, IFN-γ or TNF-α observed in mice immunized with rAs16 + Alhydrogel. Conversely, mice immunized with rAs16 formulated with MPLA and AddaVax showed a significant release of Th1 associated cytokines (IL-12, IFN-γ and TNF-α), Th2 type cytokines (IL-4, IL-5), and IL-10, IL-2 and GM-CSF as well (Fig 5C), even though there was no significant protective immunity observed in these Th1/Th2 adjuvant groups.
Western blot analysis with sera from mice immunized with rAs16 and rAs14 demonstrated that anti-As16 mouse sera recognized a band at ~14 kDa in the soluble extracts of lung larvae of A. suum, but not in the extracts of A. suum eggs. For As14, mouse anti-As14 sera specifically recognized a band of about 13 kDa in lung larval extracts, but not in A. suum eggs, indicating native As16 and As14 are expressed only at the larval stage, not in the eggs of A. suum (Fig 6). As16 and As14 are likely expressed in A. suum larvae after hatching and during the migration to the lungs. However, since infected mice only produced antibodies to As16 but not to As14 (Fig 1), it appears that only As16 is exposed to the immune system during larval migration.
A. suum larval stage antigens induce significant protection through the reduction of worm larvae migrating to the lungs and thus subsequently reduce lung pathology [17, 22]. It was confirmed in this study that almost sterile protection with up to 99.6% lung larval reduction upon A. suum egg challenge was observed in mice infected three times with 1,000 A. suum eggs, indicating A. suum larvae produce some antigens that induce protective immunity during the migration of the L3 larvae to the lungs. Therefore, these larval antigens recognized by the immune serum from low-dose Ascaris egg-infected mice constitute potential vaccine antigen candidates. In previous studies, As16 and As14 had been identified as two such possible targets; immunization with the E. coli-expressed recombinant rAs16 or rAs14 proteins induced protective immunity in mice when adjuvanted with the cholera toxin B subunit (CTB) and administered through the intranasal route [19, 20, 29]. Indeed, in this study we identified that the protective immune sera from mice infected with A. suum eggs strongly recognized rAs16, confirming that As16 is an antigen which exposes to the host immune system and is immunogenic during A. suum infection, and the immune response to As16 may contribute to the protective immunity. Simultaneously, we confirmed that As16 is expressed in larvae, not in eggs. However, we did not observe that As14 was recognized by the A. suum-infected mouse immune sera and accordingly the immunization of As14 didn’t elicit any protective immunity against A. suum infection in this study, indicating that exposure of antigen is required for inducing protective immunity.
To develop a vaccine that can be used cost-effectively in endemic areas, we expressed As16 and As14 as recombinant proteins in the yeast P. pastoris X-33, a process that can be easily scaled up and manufactured without concerns for endotoxin contamination from proteins expressed in E. coli [32]. The purified rAs16 and rAs14 were used to test their vaccine efficacy in mice through subcutaneous administration when formulated with adjuvant ISA720. We found that mice immunized with yeast-expressed rAs16 produced a 36.7% lung larva reduction which is statistically significant compared to the adjuvant-only control. This level of protection was reproducible in another experiment with rAs16 formulated with Alhydrogel. In addition to the reduced number of larvae migrating to the lung, we also observed developmental stunting in those larvae that migrated to the lungs, suggesting immune responses induced by rAs16 damage the larvae’s viability and/or impair the development of survived larvae. We notice that As16 formulated with ISA720 and alum induced less lung larval reduction (36.7–38.9%) in this study compared to previous 58% induced by intranasal administration of As16 conjugated CBT as adjuvant in the previous study [20]. It is possibly due to the different adjuvants, immunization route or the post-translational modification of rAs16 expressed in E. coli and P. pastoris system that may induce different immune responses. The intranasal immunization may induce mucosal immunity in respiratory ducts that may contribute to the better protection against larva migrated to lungs [20], but there is no evidence that intranasal immunization could induce mucosal immunity in the gut in which the adult Ascaris worms parasitize. CTB is a nontoxic portion of cholera toxin, a toxin that causes massive watery diarrhea. CTB has the capacity not only to bind to monosialotetrahexosylganglioside on epithelia or antigen presenting cells to induce immune response as an adjuvant, but also to evoke a regulatory response that causes safety concern [42]. Therefore CTB is not widely used as an adjuvant for a human vaccine test. The effect of CBT as an adjuvant on inducing intestinal immunity against gastrointestinal pathogens through oral immunization is not well determined. The mice fed with rice transgenic with As16-CTB fusion were not able to induce enough immune response against A. suum infection unless CT was added [30]. Here we report mice subcutaneously immunized with As16 formulated with ISA720 and alum. The latter is a commonly used adjuvant approved by FDA for human use at low cost, and produced significant larval reduction in lungs upon A. suum infection, making As16 feasible and practical to be used for human trial.
The mechanism of protection induced by rAs16 remains under investigation. Possibly A. suum larva-secreted As16 is critical for the survival and development of the A. suum larvae in mammalian hosts. Therefore, the neutralization of this antigen through specific antibodies or other components of the immune responses may weaken the viability of the migrating larvae and block them from reaching the lungs, a critical step in migration back to intestine to develop to adult worms [22]. It has been demonstrated that anti-As16 antibodies inhibited the molting and survival of infective L3 when co-incubated together in vitro [29]. More specifically, immune-mediated protection appears to be associated with high levels of antigen-specific IgG1 and Th1/2 type cytokine secretion including elevated IL-4, IL-10 and INF-γ in previous studies [20, 29]. We observed that immunization with rAs16 formulated with ISA720 induced much higher antibody titers of IgG1 than IgG2a (2662:1) and elevated IL-4 and IL-5 but without any detectable IFN-γ response, indicating that the protection induced by immunization with rAs16 formulated with ISA720 in this study is associated with predominant Th2-type response rather than Th1-type response. IL-2 and IL-10 were also observed to be induced upon the immunization of As16 formulated with ISA720. Even though IL-2 is considered a Th1-biased cytokine, it also plays a central role in Th2 differentiation [43]. IL-10 has been essentially re-classified as a Treg-associated cytokine [44], but it is not clear if IL-10 is involved in the protective immunity since IL-10 was not induced in mice immunized with As16 formulated with Alhydrogel that produced similar protection. The protection of mice immunized with As16 formulated with Alhydrogel was associated with high level of IgG1 and IL-5, further confirming the Th2-type immune responses contribute mainly to the protective immunity of As16 against A. suum infection, This finding is consistent with other studies showing that a Th2-type response is critical in achieving protective immunity against helminthic infections [45–47].
As the second vaccine antigen candidate investigated in this study, As14 shares 47% sequence identical with As16 and produced protection in immunized mice when formulated with CTB in our previous study [19], we did not see any protection with immunization with As14 formulated with ISA720 adjuvant in this study. We did confirm expression of the antigen in A. suum larvae as well as strong immunogenicity when immunized in mice, but As14 was not recognized by immune sera from A. suum infected mice. Our findings suggest that As14 might not be exposed to the host immune system during infection so that immune response could not access it, or it might trigger a Th1-type immune response characterized with high level of IFN-γ that is not related to protection. Another possibility for not seeing protection with As14 immunization plus ISA720 in this study is that As14 may only induce protection when formulated with CBT and administered through the intranasal route that induces mucosal immunization [19].
Immunization with rAs16 alone, without an adjuvant, is not sufficient to yield protective immunity [20, 29], indicating that an adjuvant is necessary for boosting the rAs16-associated immune response or inducing a certain type of immune response that relates to protection. Since CBT may not be a suitable adjuvant for use in humans, we have tried some other adjuvants that are used commonly in vaccine efficacy tests. ISA720 is a water-in-oil emulsion that has been used in clinical trials for malaria vaccines and other infections [48, 49]. In this study, immunization with rAs16 emulsified with ISA720 induced a strong Th2-type immune response that was associated with lung larva reduction in immunized mice. To further interpret the protective immunological mechanism underlying the protection induced by As16, three other adjuvants (Alhydrogel, MPLA and AddaVax) were formulated with rAs16. Alhydrogel, an alum-based adjuvant approved by the FDA for use in humans, is known to stimulate a Th2-type response [50, 51]. MPLA (Monophosphoryl Lipid A) is a potent activator of TLR4 that boosts a combined Th1/Th2 response [52]. AddaVax is a squalene-based oil-in-water nano-emulsion adjuvant similar to MF59 that has been licensed in Europe for flu vaccines with the ability to induce a balanced Th1/Th2-type response [53, 54]. Our studies revealed that mice immunized with 25 μg of rAs16 formulated with Alhydrogel produced a similar level of lung larva reduction (38.9%) as observed with 50 μg of rAs16 formulated with ISA720 (36.7%); accompanied by high levels of IgG1 and IL-5, but without IFN-γ or TNF-α production. Immunization with rAs16 formulated with MPLA or AddaVax induced a mixed Th1/2-type responses, with high levels of both Th1 associated cytokines (IL-12, IFN-γ and TNF-α) and Th2 type cytokines (IL-4, IL-5, IL-10) and IL-2; however, this, at best, only resulted in a 26.1% larva reduction, which was not statistically different from the controls. Together, these results suggest that a Th1-type response might be counterproductive in terms of achieving protective immunity and might even interfere with the desired Th2-type response. Some evidence showed that IL-12 and IFNγ suppressed Th2 responses that caused chronic helminth and plasmodium co-infection [55]. In Schistosoma infections, Th1-type response did not contribute to protective immunity, but caused serious inflammatory pathology in the liver [56]. Other studies have found that the Th2-polarized protection induced by some nematode infections actually downregulated the Th1-type response [47, 57].
Interestingly, mice immunized with ISA720 and Alhydrogel formulated rAs16 that showed protection against A. suum egg challenge were all found to have high levels of IL-5. IL-5 plays a major role in the regulation of eosinophil formation, maturation, recruitment and survival [58]; eosinophils have been identified as important contributors in the control of helminth infections in mammalian hosts [59, 60]. IL-5 knockout mice, for instance, were incapable of augmenting blood and tissue eosinophil levels and were diminished in their ability to kill Strongyloides stercoralis larvae through innate and adaptive immune responses [61]. Thus, the induction of IL-5 in the immunized mice may recruit and activate eosinophils towards the migrating larvae and damage the parasite through antigen-dependent cellular cytotoxicity (ADCC) or through releasing toxic granules [59]. In addition to IL-5, granulocyte-macrophage colony-stimulating factor (GM-CSF) was also elevated in protected mice. In helminth infections, GM-CSF is known as a potent stimulator of granulocytes, namely eosinophils, basophils, and dendritic cells [62], all serving as first responders to the parasites [47]. It has been demonstrated that activated murine eosinophils can serve as specific antigen-presenting cells after infection with the cestode Mesocestoides corti [63], prime naïve T cells, and aid in the maturation of antigen-specific T-cells, for example upon S. stercoralis infection [64]. However, it cannot be concluded whether IL-5 or GM-CSF are involved in the protection since mice immunized with As16 formulated with MPLA and AddaVax also showed similar levels of IL-5 and GM-CSF but did not elicit significant protection. The real roles of IL-5 and GM-CSF in the As16 immunity may be more complicated and may associate with other factors.
We confirmed in this study that As16 is a good target for vaccine development against ascariasis. However, the As16-induced protection is not complete. Compared to the sterile immunity induced by A. suum infection, the partial protective immunity induced by As16 may reflect that As16 is just one of the larva-secreted protective antigens secreted by infected larvae, or trickling release of As16 from naturally infected larva may increase the immunogenicity and protection of As16. Non-sterile immunity or low protection rate is a common problem for vaccine development against helminth infections [65, 66], possibly due to the complexity of the life cycle of parasites, different antigens expressed by different stages, and immune-evasion strategies developed by helminths. However, the severity of helminth-caused diseases usually depends on worm burden, since low infection usually is asymptomatic [65]. Therefore, reducing the worm burden by vaccination, even not sterile, may benefit in reducing the seriousness of disease. To further increase the level of protection elicited by As16, its combination with other vaccine candidates such as As24 [26], As37 [21], and As-enolase [11, 25] is currently being investigated in a mouse model. The individual or combination of antigens will eventually be evaluated in a pig model, the permissive host of A. suum in which adult worms can be developed. As16 and other vaccine candidates are expected to be more effective at reducing worm burden in a pig model since most of these antigens are expressed on the both stages of larva and adult worm.
The ultimate goal is to develop a multivalent vaccine or vaccine combination with other STH vaccines such as Na-GST-1 and Na-APR-1, which are currently under clinical trial for preventing hookworm infection [12], to prevent infections or re-infections of more than one STH, as a complementary strategy for MDA to control STH endemic. The combination of different antigens from different STHs may increase protective immunity due to cross-protection since STHs are genetically related and share some sequence homology [11]. The individuals in endemic areas are usually repeatedly infected with Ascaris and other STHs. The existing immune responses to the infections may confound the vaccine efficacy; therefore the vaccine immunogenicity and efficacy should be tested in a re-infection animal model. Another concern for As16 as a surface-associated antigen of larval is IgE response during natural infection. Immunization of As16 and other larval antigens in individuals with existing IgE may induce allergic response as hookworm larva-secreted Na-ASP-2 does [67]. Therefore, immunoscreening for the IgE response in the endemic population is needed before the decision is made to use As16 for a vaccine trial in endemic area.
As16 (A. suum, BAC66614.1); As14 (A. suum, BAB67769.1); Ce16 (Caenorhabditis elegans, NP_495640.1); Dv16 (Dictyocaulus viviparus, KJH51207.1); Acey16 (Ancylostoma ceylanicum, EPB72254.1); Ac16 (A. caninum, ABD98404.1); Ad16 (A. duodenale, KIH68079.1); Sv-SXP (Strongylus vulgaris, AGF90534.1); Na-SAA2 (Necator americanus, XP_013290850.1); Al-Ag1 (A. lumbricoides, ACJ03764.1); Bs-Ag1 (Baylisascaris schroederi, ACJ03761.1); Av-RAL2 (Acanthocheilonema viteae, AAB53809.1); Ll-SXP (Loa loa, XP_003142836.1); Ov-RAL2 (Onchocerca volvulus, P36991.1); WB14 (Wuchereria bancrofti, AAC17637.1); Bm-RAL-2 (Brugia malayi, XP_001900036.1); Bs-Ag2 (B. schroederi, ACJ03762.1); Al-Ag2 (A. lumbricoides, ADB45852.1); Hc16 (Haemonchus contortus, CDJ91573.1); Asim16 (Anisakis simplex, BAF43534), Tc16 (Toxocara canis, KHN84076.1) and As-RAL-2 (Anisakis simplex, BAF75709.1).
|
10.1371/journal.pcbi.1006076 | Cox-nnet: An artificial neural network method for prognosis prediction of high-throughput omics data | Artificial neural networks (ANN) are computing architectures with many interconnections of simple neural-inspired computing elements, and have been applied to biomedical fields such as imaging analysis and diagnosis. We have developed a new ANN framework called Cox-nnet to predict patient prognosis from high throughput transcriptomics data. In 10 TCGA RNA-Seq data sets, Cox-nnet achieves the same or better predictive accuracy compared to other methods, including Cox-proportional hazards regression (with LASSO, ridge, and mimimax concave penalty), Random Forests Survival and CoxBoost. Cox-nnet also reveals richer biological information, at both the pathway and gene levels. The outputs from the hidden layer node provide an alternative approach for survival-sensitive dimension reduction. In summary, we have developed a new method for accurate and efficient prognosis prediction on high throughput data, with functional biological insights. The source code is freely available at https://github.com/lanagarmire/cox-nnet.
| The increasing application of high-througput transcriptomics data to predict patient prognosis demands modern computational methods. With the re-gaining popularity of artificial neural networks, we asked if a refined neural network model could be used to predict patient survival, as an alternative to the conventional methods, such as Cox proportional hazards (Cox-PH) methods with LASSO or ridge penalization. To this end, we have developed a neural network extension of the Cox regression model, called Cox-nnet. It is optimized for survival prediction from high throughput gene expression data, with comparable or better performance than other conventional methods. More importantly, Cox-nnet reveals much richer biological information, at both the pathway and gene levels, by analyzing features represented in the hidden layer nodes in Cox-nnet. Additionally, we propose to use hidden node features as a new approach for dimension reduction during survival data analysis.
| With the wide application of genomics technologies, gene expression data of patients are often used as inputs to predict patients’ survival. Computationally, survival prediction is usually framed as a regression problem to model patients’ survival time (or other event time). The most common method is the Cox-PH model, a semi-parametric proportional hazards model, where the covariates of the models explain the relative risks of the patients, termed hazard ratios [1]. Given the large amount of input features in gene expression data, penalization methods such as LASSO (L1 norm), ridge (L2 norm) and MCP [2] regularizations are often used to help select representative feature in Cox-PH models. A modification of Cox-PH model is CoxBoost [3]. It is an iterative “gradient boosting” method, where the parameters are separated into individual partitions. The partition that leads to the largest improvement in the penalized partial log likelihood is selected and in subsequent iterations, the model selects another block and refits those parameters by maximizing the penalized partial log likelihood [3]. Random Forests Survival (RF-S) is a tree-based, non-linear, ensemble method [4], rather than a proportional hazards model. For each tree in the forest, data is bootstrapped, and nodes are split by maximizing the log-rank statistic. The cumulative hazard function (CHF) is estimated in each tree and a patient’s CHF is calculated as an average over all the trees in the ensemble.
Besides these methods above, Artificial Neural networks (ANNs), a type of model that is based on the idea of neurons in processing information, could be trained to predict survival as well. Developed in 1943, ANNs were used to model non-linear behavior [5]. In an ANN, hidden units, termed as neurons or nodes, may be activated or deactivated, depending on the input signals, based their own linear weight and bias parameters. The data are fed forward through the network, and for each hidden unit these weight and bias parameters are learned through back propagation along the gradient of the loss function. In recent years, ANNs have caught renewed attention to solve problems in genomics field [6, 7], thanks to increased parallel computing power and the promise of deep learning [8]. For example, Alipanhi et al. used deep learning in order to better predict the bind of RNA and DNA to proteins [9]. Ciresan et al. used convolutional neural networks to detect cell mitosis in histological breast cancer images [10]. However, relative to these new areas, survival prediction using ANN has been lagging behind.
The first ANN model to predict survival was done by Faraggi and Simon, who used four clinical input parameters to model prostate cancer survival[11]. However, their simple model was not suitable for high throughput input data, where tens of thousands of features are present per patient. Subsequently, other authors attempted to implement ANN methods to predict patient survival. One study applied ANNs to high dimensional survival data by simplifying the regression as a binary classification problem [12, 13], and another study fit continuous variables of survival time to discrete variables through binning [12, 13]. These approaches potentially led to loss of accuracy in prediction. Another study used time as an additional input in order to predict patient survival or censoring status [14], which will overfit when the survival and censoring are correlated. Thus far, an ANN model based on proportional hazards to analyze high throughput data in the genomics era is lacking.
To address all the issues of ANN based predictions as mentioned earlier, we have developed a new software package, named Cox-nnet. We use a two layer neural network: one hidden layer and the output layer. Rather than approximating survival as a classification problem, we used the output layer to perform Cox regression based on the activation levels of the hidden layer. Cox-nnet also computes feature importance scores, so that the relative importance of specific genes to prognosis outcome can be assessed. More importantly, the hidden layer node structure in the ANN can be analyzed to reveal more useful information regarding relevant genes and pathways, compared to other methods in the study. A similar idea for classification (rather than survival analysis) was recently explored in dimension reduction of single cell RNA-Seq data, in which a set of input genes with high weights to the hidden nodes of the neural network, in single cell RNA-Seq was analyzed using GO analysis [15]. Overall, Cox-nnet is a desirable survival analysis method with both excellent predictive power and ability to elucidate biological functions related to prognosis.
The neural network model used in this paper is shown in Fig 1 and an overview of modules in the Cox-nnet package is shown in S1 Fig. The current ANN architecture is composed of the input, one fully connected hidden layer (143 nodes) and an output “proportional hazards” layer. Cox-nnet performs cross-validation (CV) to find the optimal regularization parameter. Due to the large number of parameters, overfitting is a potential problem in ANNs, particularly for small datasets. Thus for regularization, we experimented with a range of regularization methods, including ridge, dropout [16], and the combination of ridge and dropout (see details in Methods). We found that dropout regularization offered overall the best model (S2A and S2B Fig). Furthermore, we compared Cox-nnet structures within no hidden layer (a standard Cox-PH model), one hidden layer (143 nodes) and two hidden layers (143 nodes in both layers) under dropout regularization (S3 Fig). We found that a single hidden layer Cox-nnet performed slightly better than those with no hidden layer (standard Cox-PH) or two hidden layers (S3A and S3B Fig). Thus, we used the single hidden-layer Cox-nnet with dropout regularization (average dropout rate = 7.75 +/- 0.042), for comparison with other survival methods in all following analyses.
Many other functions are implemented to improve the usability of the package (S1 Fig). Among them, the optimizers for adapting the learning rate include momentum gradient descent [17] and Nesterov accelerated gradient [18]. A comparison of these descent methods is shown in S4A Fig. We chose Nesterov accelerated gradient search method for this report. Other parameterization details of Cox-nnet are described in the Methods section.
We compared four methods, including Cox-nnet, Cox-PH (including Ridge, LASSO and MCP penalizations), CoxBoost and RF-S on 10 datasets from The Cancer Genome Atlas (TCGA). These datasets were selected for having at least 50 death events (S1 Table). For each dataset, we trained the model on 80% of the randomly selected samples and determined the regularization parameter using 5-fold CV on the training set. We evaluated the performance on the remaining 20% holdout test set. We replicated this evaluation 10 times in order to assess the average distribution of each method.
We used four accuracy metrics to evaluate the performance of the model. The first one is C-IPCW (inverse probability of censoring weighted) [19]. This metric aims to overcome the inaccuracy of the unweighted concordance index when censoring time is correlated with the patient’s hazard score. The second metric is Harrell’s concordance index (C-harrel) [20], which is an unweighted concordance index that evaluates the relative ordering of the samples, comparing the prognostic index (i.e., log hazard ratio) of each patient with the survival times. The third metric is the log-ranked p-value from Kaplan-Meier survival curves of two different survival risk groups. This is done by using the median Prognosis Index (PI), the output of Cox-nnet, to dichotomize the patients into high risk and low risk groups, similar to our earlier reports [21, 22]. A log-ranked p-value is then computed to differentiate the Kaplan-Meier survival curves from these two groups. It is worth noticing that the dichotomization of patients ignores the differences within each dichotomized group, thus may lead to less accuracy compared to C-index and IPCW metrics. Finally, the Integrated Brier Score (Brier) was also calculated. This score calculates the squared error between the predicted survival probability and the actual survival of patients at each time point [22–24].
The comparison of C-IPCW among the four methods over the 10 TCGA datasets is shown in Fig 2. Based on the C-IPCW score, Cox-nnet has better overall rankings than other methods (Fig 2B), but the improvement over Cox-PH is lacking statistically significance in most cases (Fig 2A). Note among the three penalization methods applied to Cox-PH, ridge penalization has the best overall accuracy (S5 Fig), and thus Cox-PH with ridge penalization is chosen to compare with the other methods. However, when using C-harrel (S6 Fig) and the log-rank p-value metrics (S7 Fig), Cox-nnet had significantly improved performance compared to all other methods. Based on the Brier score metric, Cox-nnet had significantly higher predictive accuracy compared to RF-S (S8 Fig). Overall, the other non-linear method (RF-S), an ensemble-based method consistently ranks worse than Cox-nnet and Cox-PH (Fig 2, S6, S7 and S8 Figs).
To explore the biological relevance of the hidden nodes of Cox-nnet, we used the TCGA Kidney Renal Cell Carcinoma (KIRC) dataset as an example. We first extracted the contribution of each hidden node to the PI score for each patient (Fig 3A). The contribution was calculated as the output value of each hidden node weighted by the corresponding coefficient at the Cox regression output layer. As expected, the value of the hidden nodes strongly correlated to the PI score. However, there is still significant heterogeneity among the nodes, suggesting that individual nodes may reflect different biological processes.
We hypothesize that the top (most variable) nodes may serve as surrogate features to discriminate patient survival. To explore this idea, we selected the top 20 nodes with the highest variances and presented the patients PI scores using t-SNE (Fig 3B). t-SNE is a non-linear dimensionality reduction method that embeds high-dimensional datasets into a low dimensional space (usually two or three dimensions). This method has been widely used to visualize data with large number of features, by enhancing the separation among samples[25]. The hidden nodes represent a dimension reduction of the original data and they clearly discriminate samples by their PI scores, as shown by the t-SNE plot (Fig 3B, left). As comparison, we performed t-SNE using all differentially expressed genes of patients with low prognostic index and high prognostic index (Fig 3B, right). The t-SNE plots demonstrates that the nodes in Cox-nnet effectively capture the survival information. Therefore, the top node PI scores can be used as features for dimension reduction in survival analysis.
To further explore the biological relevance of the top 20 hidden nodes, we conducted Gene Set Enrichment Analysis (GSEA) [26] using KEGG pathways [27], as described in the Methods section. Briefly, we calculated significantly enriched pathways using Pearson’s correlation between the log transformed gene expression input and the output score of each node across all patients in the KIRC dataset (Fig 3C and S2 Table). We compared these enriched pathways to those from GSEA of the Cox-PH (ridge) model (S3 Table), the competing model with the second best prognosis prediction. A total of 110 (out of 187) significantly enriched pathways (S2 Table) were identified in at least one node, including seven pathways enriched in all 20 nodes that were not found by the Cox-PH method (Table 1). In contrast, Cox-PH only identified 30 significantly enriched pathways using the same significance threshold. We also used the genes values from CoxBoost and RF-S, however they did not produce any significantly enriched pathways. Among the seven pathways enriched in all 20 nodes from Cox-nnet, the p53 signaling pathway stands out as an important biologically relevant pathway (S9 Fig), since it was shown to be highly prognostic of patient survival in kidney cancer [28].
Next, we estimated the predictive accuracies of the leading edge genes [27] enriched in the GSEA from Cox-nnet vs. those enriched in Cox-PH model. Leading edge genes are those genes in the pathway of interest that contribute positively to the enrichment score in GSEA. We used the C-IPCW of each leading edge gene, obtained from single-variable analysis (Fig 4). Collectively, leading edge genes from Cox-nnet have significantly higher C-IPCW scores (p = 1.253e-05) than those from Cox-PH, suggesting that Cox-nnet has selected more informative features. In order to visualize these gene level and pathway level differences between Cox-nnet and Cox-PH, we reconstructed a bipartite graph between leading edge genes for Cox-nnet or feature genes (for Cox-PH) and their corresponding enriched pathways (Fig 5). Besides the p53 pathway mentioned earlier that is specific to Cox-nnet, several other pathways, such as insulin signaling pathway, endocytosis and adherens junction, also have many more genes enriched in Cox-nnet. Among these genes specific to Cox-nnet, many have been previously reported to relevant to renal carcinoma development and prognosis, such as CASP9[27], TGFBR2[30], KDR (VEGFR)[31]. These results suggest that Cox-nnet model reveals richer biological information than Cox-PH.
Additionally, we compared the partial derivative of the hidden nodes (rather than the Cox-nnet output), with respect to the input genes. We first calculated the gradient for each patient and calculated the average partial derivatives and replicated the GSEA analysis as for the previous analysis. However, we found that fewer pathways are significant, and are less relevant to cancer using this approach.
To further examine the importance of each gene relative to the survival outcome, we calculated the average partial derivative of the output of the model (i.e., the log hazard ratio) with respect to each input gene value across all patients. As demonstrated by the leading edge genes in seven common pathways of all nodes in Cox-nnet, the feature importance scores from Cox-nnet appear to be more biologically insightful compared to the feature importance values from the Cox-PH model (S9 Fig). For example, the feature importance for the BAI1 gene in the p53 pathway is much higher in the Cox-nnet model compared to the Cox-PH model. Corresponding to our finding, the BAI gene family was found to be involved in several types of cancers including renal cancer [32]. BAI1 acts as an inhibitor to angiogenesis and is transcriptionally regulated by p53 [33–36]. Its expression level was significantly decreased in tumor vs. normal kidney tissue, and was even lower in advanced stage renal carcinoma[37]. Mice kidney cancer models treated with BAI1 showed slower tumor growth and proliferation [36]. Additionally, the MAPK1 gene (also known as ERK2), annotated in two pathways identified by Cox-nnet (the Adherens Junction and Insulin Signaling pathway), has a much higher feature importance score in Cox-nnet compared to Cox-PH. MAPK1 is one of the key kinases in intra-cellular transduction, and was found constitutively activated in renal cell carcinoma [38]. Drugs inhibiting the MAPK cascade have been targeted for development[39]. We list the top 20 genes from each method in S4 Table.
We have implemented Cox-nnet, a new ANN method, to predict patient survival from high throughput omics data. Cox-nnet is an alternative to the standard Cox-PH regression, enabling automatic discovery of biological features at both the pathway and gene levels. The hidden nodes in the Cox-nnet model have distinct activation patterns, and can serve as surrogate features for survival-sensitive dimension reduction. More significantly enriched KEGG pathways that correlate with top nodes in Cox-nnet are identified, as compared to those from the Cox-PH model, suggesting that Cox-nnet reveals more relevant biological information. We show how a critical pathway for renal cancer development, p53 pathway is identified only by cox-nnet but not Cox-PH model in TCGA KIRC dataset. Other pathways, including insulin signaling pathway, endocytosis and adherens junction, have many more genes enriched by Cox-nnet. Moreover, leading edge genes obtained from these KEGG pathways identified as enriched by Cox-nnet (which are a fraction of the gene features considered by the model) have collectively higher associations with survival. Enrichment analysis on the top genes from Random Forest and CoxBoost did not produce any significant pathways. As a promising new predictive method for prognosis, the current Cox-nnet implementation has some limitations. Its architecture is relatively simple, including one or two hidden layers and an output Cox regression layer. It is possible to incorporate other more sophisticated architecture into the model, such as including more layers of neurons or more sophisticated hidden layers. However, deeper ANN is not necessarily more beneficial (S3 Fig), when compared to the regularization methods. This suggests that ANN may overfit the small size of the genomics data tested. New variations of neural networks, such as convolutional neural network approach or a recurrent network approach as those reported showed good performance in processing imaging or other types of positional data [40], and they could be used as input to a proportional hazards output layer. Additionally, it is possible to embed a priori biological pathway information into the network architecture, e.g., by connecting genes in a pathway to a common node in the next hidden layer of neurons [15]. In the future, we plan to further analyze how different neural network architectures affect the performance of Cox-nnet and compare the biological insights from the various models.
We analyzed 10 TCGA datasets which were combined into a pan-cancer dataset. The TCGA datasets included the following cancer types: Bladder Urothelial Carcinoma (BLCA), Breast invasive carcinoma (BRCA), Head and Neck squamous cell carcinoma (HNSC), Kidney renal clear cell carcinoma (KIRC), Brain Lower Grade Glioma (LGG), Liver hepatocellular carcinoma (LIHC), Lung adenocarcinoma (LUAD), Lung squamous cell carcinoma (LUSC), Ovarian serous cystadenocarcinoma (OV) and Stomach adenocarcinoma (STAD). RNA-Seq expression and clinical data were downloaded from the Broad Institute GDAC [41]. Overall survival time and censoring information were extracted from the clinical follow-up data. Raw count data were normalized using the DESeq2 R package [17] and then log-transformed. Datasets were selected from TCGA based on the following criteria: > 300 samples with both RNASeq and survival data and > 50 survival events. In total, 5031 patient samples were used (see S1 Table for a patient tabulation by individual dataset).
Cox-nnet is an extension to the Cox-PH model. Individual hazard, an instantaneous measure of the likelihood of an event, is estimated based on a set of features [17]. The hazard function is:
h(t|xi)=h0(t)expθi
(1)
θi=xiTβ
(2)
Where θi is the log hazard ratio for patient i. This model uses partial log-likelihood as the cost function
pl(β)=∑C(i)=1(θi−log∑ti≥tjexp(θj))
(3)
Since gene expression data have tens of thousands in initial features, penalization methods are usually implemented along with Cox-PH. We experimented with 3 penalization methods, namely LASSO (L1 norm), ridge (L2 norm), and mimimax concave penalty (MCP). MCP attempts to moderate the biased large penalty for large coefficients in LASSO [17] (S5 Fig). MCP reduces the regularization for large coefficients and plateaus at a value selected through cross-validation. LASSO and ridge regularization were performed using the Glmnet R package [42] and the MCP regularization was performed with the Ncvreg R package [43].
CoxBoost, is an iterative “gradient boosting” method modified from the Cox-PH model [44]. In CoxBoost, parameters are separated into individual partitions, and the partition that leads to the largest improvement in the penalized partial log likelihood is selected for that iteration. In subsequent boosting iteration, the model selects another block and refits those parameters by maximizing the penalized likelihood function. In this method, the number of boosting iterations is used as the complexity parameter in CoxBoost and optimized via cross-validation (CV).
Random Forests Survival (RF-S) is a tree-based, non-linear, ensemble method [1], rather than a proportional hazards model. For each tree in the forest, samples are boostrapped, and at each node in a tree, features are boostrapped, and the node is split by selecting the feature that maximizes the log-rank statistic. At the leaf nodes, the cumulative hazard function (CHF) is estimated and a patient’s CHF is calculated as an average over all the trees in the ensemble.
Cox-nnet is a neural network whose output layer is a cox regression. In a Cox-nnet model, xi in Eq (2) is replaced by the output of the hidden layer, and the linear predictor is:
θi=G(Wxi+b)Tβ
(4)
Where W is the coefficient weight matrix between the input and hidden layer with the size H x J, b is the bias term for each hidden node and G is the activation function (applied element-wise on a vector).
In this manuscript, the tanh activation function is used:
G(z)=exp(z)−exp(−z)exp(z)+exp(−z)
(5)
Subsequently, the partial log likelihood in Eq (3) is extended by a ridge regularization term:
Cost(β,W)=pl(β,W)+λ(‖β‖2+‖W‖2)
(6)
In addition to ridge regularization, we also employ dropout regularization (2). This approach has been shown to reduce overfitting and improve performance over other regularization schemes[45]. In dropout, a hyperparameter p is the probability of retaining nodes during each iteration of training. I.e., the activation of each node is set to zero with probability 1-p. The optimal parameter is determined through CV on the training set, using C-index as the performance metric [2]. A complete description of the hyperparameters and optimization approaches for each method is shown in S5 Table. Briefly, we used 5-fold cross-validation and grid search to optimize parameters in Cox-nnet.
In neural networks, because of the large amount of parameters and hyperparameters, overfitting is a potential problem. In Cox-nnet, we experimented with three regularization approaches given previous guidelines: ridge, dropout and combination of ridge and dropout. Ridge regularization is one of the most common methods to reduce overfitting, recommended by Demuth et al. [3]. In this scheme, the L2 norm of all the weights are added to the cost function of the model, leading to a “weight decay” term in the gradient. Dropout is a recent regularization method for networks, inspired by Bayesian analysis on weighted averages of different network architectures to improve the model performance [4]. In dropout networks, each training iteration uses different network architecture; nodes are randomly deactivated from the network during training based on a probability hyperparameter between 0 and 1. Instead of entire models being reweighted, the output of each node is reweighted during evaluation. This method was previously shown to perform better than other regularization methods, such as ridge regularization [17]. Our results on Cox-nnet confirmed this earlier conclusion. Also similar to a previous study, we found that additional complexity of combining dropout and ridge regularization does not yield better performance [17].
We implement Cox-nnet in Python with Theano package, a package for automatic differentiation that is widely used for designing neural networks. Cox-nnet is trained through back propagation. The partial log likelihood is usually written as two summations (one nested within the other) conditioned on survival time and censoring status (Eq 3). Because the partial log likelihood is usually written as a nested summation, one may write a program to calculate the partial log likelihood using nested loops. However, these types of operations are very slow in Theano, whereas matrix operations are very fast. Therefore, it is imperative to convert the usual form of the partial log likelihood into a mathematically equivalent form using matrix multiplication. First, we define an indicator matrix R with elements:
Rij={1,ti≤tj0,ti>tj
(7)
We also define an indicator vector C with elements given by the censoring of each patient. An operation using R replaces the conditional sum over ti ≥ tj, and an operation using C replaces the conditional sum over C(i) = 1 in Eq 4. For the models trained in this manuscript, the number of iterations was fixed at 1e4. The learning rate was initialized at 0.1, and decayed exponentially by a factor of 0.9 if the loss did not decrease. The number of hidden nodes in the hidden layer is chosen to be the square root of the number of input nodes, following the “pyramid” rule of thumb [21]. The optimization strategy used was Nesterov accelerated gradient [41]. For the two hidden layer neural network, we used the same number of hidden nodes as the single hidden layer model in both the first and second hidden layers.
Many functions are implemented to make it easier to train and evaluate models for survival analysis, including CVSearch, CVProfile, CrossValidation, and TrainCoxMlp (S1 Fig). CVSearch, CVProfile, CrossValidation are methods that perform CV to find the optimal regularization parameter. TrainCoxMlp performs optimization of coefficients on the regularized partial log likelihood function.
The source code of cox-nnet can be found at: https://github.com/lanagarmire/cox-nnet, and can be installed through the Python Package Index (PyPI). Documentation of package can be found at http://garmiregroup.org/cox-nnet/docs/. An example of analyzing Cox-nnet through R can be found at: http://garmiregroup.org/cox-nnet/docs/examples/#interfacing-and-analysis-with-r
Cox-nnet can be run on multiple threads or a Graphics Processing Unit (GPU), an advantage of the Theano framework. We measured the running time on the KIRC dataset and re-measured it 5 times. The computational time for Cox-nnet included model compilation time and cross validation time to optimize dropout.
To evaluate the performance of all methods, we resampled the data 10 times. In each resampling iteration, we trained each model on 80% of the samples for each dataset (chosen randomly) and evaluated the performance on the 20% holdout test set. The output of Cox-PH, Cox-nnet and CoxBoost are the log hazard ratios (i.e., Prognosis Index, or PI) for each patient. The hazard ratio describes the relative risk of a patient compared to a non-parametric baseline. In contrast, the output of RF-S is an estimation of the survival time for each patient. We use C-index, IPCW [17], log-ranked p-value and Brier score to measure the accuracy performance of each model. We also compare running time of each model (S4B Fig).
C-index: is a measure of how well the model prediction corresponds to the ranking of the survival data [17]. It is calculated for censored survival data, which evaluates a value between 0 and 1, with 0.5 equivalent to a random process. The C-index can be computed as a summation over all events in the dataset, where patients with a higher survival time and lower log hazard ratios (and conversely patients with a lower survival time but higher log hazard ratios) are considered concordant.
C-IPCW (Inverse Probability of Censoring Weighting): it is a method that takes into account the censoring probability in the concordance index, by weighting pairs of patients in the calculation based on the inverse of their individual probabilities to be censored [17]. In this manuscript, we use the Kaplan-Meier estimate of censorship.
log-ranked p-value: a PI cutoff threshold is used to dichotomize the patients in the data set into higher and lower risk groups, similar to our earlier report [46]. A log-ranked p-value is then computed to differentiate the Kaplan-Meier survival curves between the higher vs. lower risk groups. In this report, we used the median log hazard ratio as the cutoff threshold.
Brier-score: the Brier-score is the mean square error of the difference between the probability of an event and the event value (1 or 0). The integrated Brier score was used as the performance metric, which is the Brier score averaged over the time interval of the dataset.
Running-time: Cox-nnet can be run on multiple threads or a Graphics Processing Unit (GPU), an advantage of the Theano framework. We measured the running time on the KIRC dataset, and re-measured it 5 times. The computational time for Cox-nnet included model compilation time and cross validation time to optimize dropout.
For computing the importance of a feature in Cox-nnet, we use a method of partial derivatives [19]. For each patient, we compute the partial derivatives of each linear output of the model (e.g., the log hazard ratio) with respect to the input. The average of the partial derivatives over each input across all patient samples is calculated as the feature score. As comparison, we also compute the partial derivatives of each hidden layer node with respect to the inputs.
T-distributed stochastic neighbor embedding (t-SNE) is a non-linear dimensionality reduction method that is commonly used for visualizing high-dimensional data [20]. We selected the top 20 nodes of the Cox-nnet model with the highest variances, and clustered the patient samples using t-SNE. To do this, we used the tsne package in R [47]. As comparison, we also plot t-SNE based on the top 33% and bottom 33% of patients as determined by the prognostic index. A total of 8467 and 5805 genes were deemed significantly up and down regulated respectively in the KIRC dataset, using DESeq2 for differential expression analysis.
We performed GSEA on the correlation of normalized log gene expression to the node output (for Cox-nnet) or the model output (for Cox-PH, Random Forest and CoxBoost), across all patients in the KIRC dataset. Using the fgsea package in R [48], we calculated statistical significance of the KEGG pathways by performing 10,000 permutations, followed by multiple hypothesis testing with Benjamini Hochberg adjustment.
To test for statistical significance between the performance of Cox-nnet and other methods, we use the “multcomp” package in R to perform multiple linear hypothesis testing, with the method as the factor of interest, and the iteration number (used to control the random sampling seed in each 80%/20% split) as a covariate. We then perform multiple comparisons that compare Cox-nnet with the other methods, and adjust for multiple hypothesis testing with the Benjamini Hochberg procedure [20]. We apply this statistical approach to all performance metrics (C-harrell, C-IPCW, log-rank, and brier score).
overfitting in Cox-nnet may come from patient censoring. To investigate this, we ran a simulation RNA-Seq dataset. We used the ssizeRNA package in R to generate simulated RNA-Seq data counts in R [23, 24]. We generated four sub-groups of 200 patients each (a total of 800 patient samples) with 1000 genes, among which 20% of the genes are differentially expressed for each group. The prognosis index for patients in each group was randomly generated based on expression of 100 randomly selected genes, and the survival times were sampled based on the Weibull survival distribution. Censoring times were chosen from the exponential distribution with rate = 0.05. We randomly generated this dataset 100 times and estimated the performance metrics on 20% holdout test-sets. We compared C-index and IPCW metrics with censoring to uncensored C-index (S10 Fig). Neural network-based Cox-nnet and tree-based Random Forest survival do not differ significantly from Cox-PH models. This may be due to the simplification in the simulated data. For example, the simulated gene-expression does not have common biological functions and embedded co-linearability as shown in TCGA data.
|
10.1371/journal.pgen.1007594 | Seven-transmembrane receptor protein RgsP and cell wall-binding protein RgsM promote unipolar growth in Rhizobiales | Members of the Rhizobiales (class of α-proteobacteria) display zonal peptidoglycan cell wall growth at one cell pole, contrasting with the dispersed mode of cell wall growth along the sidewalls of many other rod-shaped bacteria. Here we show that the seven-transmembrane receptor (7TMR) protein RgsP (SMc00074), together with the putative membrane-anchored peptidoglycan metallopeptidase RgsM (SMc02432), have key roles in unipolar peptidoglycan formation during growth and at mid-cell during cell division in Sinorhizobium meliloti. RgsP is composed of a periplasmic globular 7TMR-DISMED2 domain, a membrane-spanning region, and cytoplasmic PAS, GGDEF and EAL domains. The EAL domain confers phosphodiesterase activity towards the second messenger cyclic di-GMP, a key regulatory player in the transition between bacterial lifestyles. RgsP and RgsM localize to sites of zonal cell wall synthesis at the new cell pole and cell divison site, suggesting a role in cell wall biogenesis. The two proteins are essential for cell wall biogenesis and cell growth. Cells depleted of RgsP or RgsM had an altered muropeptide composition and RgsM binds to peptidoglycan. RgsP and RgsM orthologs are functional when interchanged between α-rhizobial species pointing to a conserved mechanism for cell wall biogenesis/remodeling within the Rhizobiales. Overall, our findings suggest that RgsP and RgsM contribute to the regulation of unipolar cell wall biogenesis in α-rhizobia.
| Bacteria face the challenge of maintaining their peptidoglycan cell wall integrity during growth and division. The enzymes involved in cell wall biogenesis are tightly regulated and targeting of these enzymes by β-lactams cause cell death. In contrast to the well-characterized mode of dispersed cell wall formation in many rod-shaped bacteria, the mechanisms controlling polar cell wall formation in α-proteobacteria are largely unknown. Seven-transmembrane receptors (7TMRs) are widespread in eukaryotes and prokaryotes but only few of them have been functionally studied in bacteria. Here we suggest that a 7TMR-DISM protein cooperates with a putative peptidoglycan-hydrolyzing protein, to facilitate unipolar cell wall formation and cell division in the Rhizobiales. This 7TMR-DISM protein also contributes to degradation of the second messenger cyclic di-GMP.
| Rod-shaped bacteria have evolved diverse modes of cell growth. In Bacillus subtilis, Escherichia coli and Caulobacter crescentus, cells grow by elongating along the lateral sidewall, incorporating peptidoglycan (PG) cell wall material in a dispersed pattern along the sidewall [1]. Other bacteria elongate by zonal growth in which PG is incorporated at one or both cell poles [2]. Unipolar growth is observed in α-proteobacterial Rhizobiales, such as Agrobacterium tumefaciens and Sinorhizobium meliloti [3], while Gram-positive Actinomycetales including Mycobacterium tuberculosis grow bipolarly [4]. Later in the cell cycle, all species switch PG synthesis completely or partially to mid-cell for cell division [1]. During PG synthesis nascent material is incorporated into the pre-existing PG structure (sacculus), by the activity of hydrolases which selectively cleave bonds in the stress-bearing sacculus [5]. Because the integrity of the PG sacculus is essential for maintaining cell shape and resisting turgor [6], cell growth and PG synthesis require precise spatial and temporal regulation of the incorporation of new material into the PG sacculus [7]. Rod-shaped bacteria with a dispersed mode of PG incorporation along the lateral cell wall utilize the cytoplasmic actin-like MreB protein to direct PG synthesis. Dynamic filaments or patches of MreB are believed to serve as platforms for the intracellular and extracellular PG synthesis machineries [8,9]. By contrast, most rod-shaped bacteria with polar growth do not contain MreB homologs [10] and it is currently unknown how polar cell wall elongation is regulated in these bacteria.
Bis-(3′-5′)-cyclic dimeric guanosine monophosphate (c-di-GMP) has a central role in the regulation of motility, adherence, biofilm formation and virulence and in several bacterial species, also for promoting cell cycle progression, growth and development [11–15]. c-di-GMP is synthesized by diguanylate cyclases (DGCs) with a conserved GGDEF domain and degraded by phosphodiesterases (PDEs) with either an EAL domain or a HD-GYP domain [16]. In the α-proteobacterium C. crescentus, the spatial organization of proteins involved in c-di-GMP metabolism contributes to cell polarity and cell cycle progression [14], and we showed previously that strong overproduction of this second messenger inhibited growth and resulted in cell filamentation in S. meliloti [17].
Seven-transmembrane receptors (7TMRs) form the largest, most ubiquitous and most versatile family of membrane receptors. In eukaryotes, they are involved in signaling via interaction with cytoplasmic G-proteins [18]. A distinct class of bacterial 7TMRs consists of so-called 7TMR-DISMs, which stands for 7TMR with diverse intracellular signaling modules [19]. 7TMR-DISM proteins contain seven transmembrane α-helices (7TMR-DISM_7TM domain) fused to various cytoplasmic signaling and/or extracytoplasmic 7TMR-DISMED1 or 7TMR-DISMED2 domains [19]. These extracytoplasmic domains have been predicted to bind ligands such as carbohydrates [19]. Common cytoplasmic signaling modules in 7TMR-DISM proteins include histidine protein kinase domains, GGDEF and EAL domains involved in c-di-GMP homeostasis and sensory Per-Arnt-Sim (PAS) domains. PAS domains are known to sense small molecules, ions, gases, light or redox state [20]. With up to 14 paralogs per genome, for example in the spirochete Leptospira interrogans, 7TMR-DISMs are widely distributed in both Gram-negative and Gram-positive bacteria [19]. However, only a few of these proteins have been functionally characterized. These include the Pseudomonas aeruginosa 7TMR-DISM histidine kinases RetS and LadS, which are involved in regulation of biofilm formation, and the GGDEF domain containing protein NicD, which is involved in biofilm dispersal [21,22].
The 7TMR-DISM protein SMc00074 (renamed RgsP for rhizobial growth and septation c-di-GMP phosphodiesterase) was previously shown to be essential in S. meliloti [17,23,24]. Here, we provide evidence that RgsP is important for PG synthesis in certain α-rhizobial species. We identify SMc02432 (a putative membrane-anchored periplasmic PG metallopeptidase, renamed RgsM for rhizobial growth and septation metallopeptidase) as a RgsP interaction partner and show that both are required for unipolar cell growth in S. meliloti and related Rhizobiales. RgsP was also found to be an active PDE, substantially contributing to c-di-GMP homeostasis and hence possibly connecting c-di-GMP signaling with spatiotemporal control of PG synthesis.
RgsP is composed of 7TMR-DISMED2, 7TMR-DISM_7TM, PAS, GGDEF and EAL domains. Previous systematic mutagenesis of c-di-GMP-related genes identified rgsP (SMc00074) as a potentially essential gene in S. meliloti Rm2011 [17]. C-terminally 3×FLAG-tagged RgsP accumulated in growing cells and was only detected at very low levels in stationary phase cells (S1A Fig), suggesting a role during cell growth. To further study the function of this protein, we constructed a RgsP depletion strain (Rm2011 rgsPdpl) by placing the native rgsP gene under the control of an IPTG-inducible promoter. Using the same promoter, we confirmed depletion of a C-terminally 3×FLAG-tagged RgsP variant in the absence of IPTG (S1B Fig).
Growth of Rm2011 rgsPdpl was strongly dependent on IPTG (Fig 1A). Cells cultured in the presence of IPTG showed wild type-like growth and morphology, whereas RgsP-depleted cells lost the rod shape and had an irregular, uneven cell surface (Fig 1B and 1C; S2 and S3A Figs). 4.7% of these cells were stained by the dead cell indicator propidium iodide (S4A Fig). This suggests that the physiological effect of RgsP depletion probably is mostly bacteriostatic after 24 h of incubation in absence of IPTG. To estimate the capability of RgsP-depleted cells to resume growth, we analyzed colony formation and microscopically monitored growth on TY medium containing IPTG. In cultures of Rm2011 rgsPdpl grown without added IPTG for 12 or 24 h, the number of colony forming units (CFU ml-1 OD600-1) was reduced by 67.8% or 97.4%, respectively, compared to IPTG-supplemented cultures (S4B Fig). Following the fate of RgsP-depleted cells (previously grown in the absence of IPTG for 24 h), 25.8% were able to recover and to give rise to microcolonies on TY agarose pads with IPTG. In contrast, only 3.0% of the cells were able to divide at least once during the 12 h observation period on pads lacking IPTG (S4C Fig). Thus, RgsP depletion predominantly resulted in growth arrest, which was relieved upon IPTG addition in a minor fraction of the cells only. However, when interpreting these results, it has to be taken into account that the cells grown under depletion conditions may contain some residual amounts of RgsP.
Zones of PG synthesis were visualized by HADA (7-hydroxycoumarin-3-carboxylic acid-D-alanine)-labeling [25]. In IPTG-supplemented cultures with induced rgsP expression, 98% of the cells were stained at one of the cell poles and/or septal region. By contrast, only 33% of the RgsP-depleted cells incorporated HADA (Fig 1D; S5A Fig). Moreover, upon RgsP depletion, the proportion of pre-divisional cell doublets with visible septum constriction increased to 42% compared to 13% in RgsP-replete cultures (Fig 1D; S5B Fig). These results suggest that polar cell wall synthesis and late stages of cell division were impaired in the absence of RgsP.
To determine the subcellular localization of RgsP, we replaced rgsP with rgsP-egfp at its native genomic location. In exponentially growing S. meliloti cells, RgsP-EGFP co-localized with HADA-labeled sites of PG synthesis at one of the cell poles and the division site (Fig 2A and 2B). Time-lapse microscopy showed that RgsP-EGFP remained at the pole during the entire phase of cell elongation until it relocated to the mid-cell region (Fig 2C). We next localized RgsP-EGFP simultaneously with ParB-mCherry. ParB binds to the parS sites close to the chromosomal origin of replication [26]. Early in the cell cycle, a single fluorescent ParB-mCherry focus localized at the old cell pole and RgsP-EGFP localized at the opposite, new pole (Fig 2C). Later, a second ParB-mCherry focus moved to the new pole. Following a period of co-localization (~90 minutes) of RgsP-EGFP and ParB-mCherry at the new pole, the polar RgsP-EGFP signal disappeared and appeared at mid-cell. Thus, RgsP-EGFP localizes at the new cell pole early in the cell cycle and later at the cell division site. Importantly, these are the sites of PG synthesis.
To relate the temporal pattern of RgsP relocation to the mid-cell to known pre-divisional processes, we co-localized RgsP with the early divisome component FtsZ [7]. mCherry-FtsZ showed diffuse fluorescence in growing cells, and localized to the mid-cell in pre-divisional cells (Fig 2A). Septal localization of RgsP-EGFP always coincided with the presence of a mCherry-FtsZ focus at mid-cell, but not vice versa (Fig 2A and 2B), suggesting that RgsP accumulates at the septal site later than FtsZ. Time-lapse microscopy showed localization of RgsP-EGFP at mid-cell about 24 minutes after occurrence of the mCherry-FtsZ focus. This RgsP-EGFP relocalization was immediately followed by the onset of septum constriction, which was completed by cell division about 24 minutes later (Fig 2D). This implies that RgsP may also have a function in septation.
The dynamics of RgsP relocation to the mid-cell region was analyzed at a higher time resolution relative to PleD. We found previously that the DGC PleD localizes to the new cell pole within 20 minutes before cell division [17]. Simultaneous tracking of PleD-EGFP and RgsP-mCherry showed that the accumulation of PleD-EGFP at the growing cell pole temporally correlated with the RgsP-mCherry signal fading away at the pole and relocating to the division site (S6 Fig), indicating mutually exclusive localization of RgsP and PleD at the new pole.
To dissect the functionality of RgsP relative to its complex domain organization, we generated RgsP variants, each with or without C-terminally fused EGFP (Fig 3A). The gene variants were ectopically expressed either from the native promoter P*rgsP (S7A Fig) on the single-copy plasmid pABCS2-mob (native level of 3×FLAG-tagged RgsPwt (S1C Fig)) or from a constitutive synthetic promoter (Psyn) on the low-copy plasmid pR_egfp (elevated level of 3×FLAG-tagged RgsPwt (S1D Fig)) in the RgsP-depleted strain Rm2011 rgsPdpl. All RgsP variants were assayed for their ability to complement the cell growth and morphology defects of the RgsP-depleted strain, and for subcellular localization of EGFP fusion proteins. Ectopically expressed rgsPwt or rgsPwt-egfp restored growth and cell morphology of RgsP-depleted cells (Fig 3B and 3C), and RgsPwt-EGFP localized similarly when expressed from the ectopic and native gene locus (Figs 2 and 3D).
Point mutations targeting the conserved inhibitory site (I site) and the DGC active site motifs in the GGDEF domain (RxxD to AxxA and GGDQF to GAAAF) and the PDE active site in the EAL domain (EAL to AAL), or removal of the EAL domain did not significantly affect protein localization or complementation of the cell growth and morphology defects (Fig 3B, 3C and 3D). RgsP and RgsP-EGFP variants lacking the GGDEF or both GGDEF and EAL domains were unable to fully complement growth and morphology defects unless they were expressed at higher levels from Psyn (Fig 3B and 3C). Nevertheless, the EGFP-tagged versions showed normal cellular localization (Fig 3D). When gene expression was driven by P*rgsP on pABCS2-mob, levels of corresponding 3×FLAG-tagged wild type protein and variants were similar, except for RgsPΔGGDEF, which was detected at 67% of the corresponding wild type protein level (S1C Fig). This implies that RgsP lacking the EAL domain fully supports cell growth, whereas lack of the GGDEF domain impairs protein stability or function.
To investigate the possible role of c-di-GMP in RgsP function, we tested complementation of RgsP depletion by rgsP and rgsP-egfp variants in the c-di-GMP-deficient strain Rm2011 ΔXVI, which lacks all genes predicted to encode active DGCs and in which the level of c-di-GMP was below the detection limit [17]. The complementation properties of each of these RgsP versions were indistinguishable in strains with or without c-di-GMP, and RgsP-EGFP showed normal localization in Rm2011 ΔXVI (S8 Fig). This suggests that c-di-GMP is not required for RgsP localization and the essential function of this protein in cell growth.
Deletion of the N-terminal domains 7TMR-DISMED2, 7TMR-DISM_7TM or PAS abolished the ability of RgsP to complement RgsP depletion phenotypes, irrespective of the expression vector (Fig 3B and 3C). When gene expression was driven by P*rgsP on pABCS2-mob (or Psyn on pR_egfp), 3×FLAG-tagged RgsPΔPAS, RgsPΔPASΔGGDEFΔEAL, RgsPΔ7TMR-DISMED2, and RgsPΔ7TMR-DISMED2Δ7TMR-DISM_7TM variants accumulated to 58% (736%), 30% (592%), 6% (236%) and 9% (19%), respectively, of the corresponding wild type protein levels (S1C and S1D Fig). Furthermore, the fluorescence signal of these RgsP variants, tagged with EGFP, did not show the characteristic polar and septal localization (Fig 3D).
Taken together, these data suggest that the N-terminal part of RgsP determines the essentiality of this protein, whereas the EAL domain is dispensable for the growth-promoting function of RgsP.
Since the GGDEF and EAL domains may have a regulatory role, we analyzed the enzymatic activities and ability to bind c-di-GMP in vitro. A purified His6-tagged variant of RgsP containing the PAS, GGDEF and EAL domains (His6-RgsPPAS-GGDEF-EAL) hydrolyzed [α-32P]-c-di-GMP, whereas no cleavage product was detected with the PDE active site mutant variant His6-RgsPPAS-GGDEF-EAL-AAL (S9A Fig). In a DGC activity assay with [α-32P]-GTP as a substrate, His6-RgsPPAS-GGDEF-EAL-AAL did not synthesize c-di-GMP, in contrast to C. crescentus DgcA used as a positive control (S9B Fig). This is in agreement with the degenerate active site GGDQF in the RgsP GGDEF domain and our previous in vitro DGC activity assay with a RgsP fragment comprising the PAS, GGDEF and EAL (active site intact) domains [17]. Since an intact I site RxxD is present in the GGDEF domain of RgsP, we assayed for the ability of RgsP to bind c-di-GMP in a differential radial capillary action of ligand assay (DRaCALA) using a His6-tagged RgsP variant containing only the PAS and GGDEF domains (His6-RgsPPAS-GGDEF). In this assay, the positive control His6-DmxB from Myxococcus xanthus produced the characteristic DRaCALA pattern, whereas His6-RgsPPAS-GGDEF was not able to prevent the diffusion of [α-32P]-c-di-GMP (S9C Fig). Thus, RgsPPAS-GGDEF-EAL has c-di-GMP PDE activity, presumably catalysed by the EAL domain, but the GGDEF domain does not bind or synthesize c-di-GMP.
To evaluate c-di-GMP PDE activity of RgsP in vivo, we determined the c-di-GMP content of RgsP-depleted Rm2011 rgsPdpl complemented with rgsPwt, rgsPAAL and rgsPGAAAF expressed from P*rgsP. The c-di-GMP content of cells expressing rgsPwt and rgsPGAAAF was similar, whereas expression of the PDE active site mutant variant rgsPAAL resulted in a two-fold increase in c-di-GMP content (S9D Fig). This data is consistent with the in vitro data and provides evidence that RgsP substantially contributes to c-di-GMP degradation in vivo.
To identify protein interaction partners of RgsP, we performed co-immunoprecipitation (Co-IP) experiments with a C-terminally 3×FLAG-tagged variant of RgsP, encoded by rgsP-3×flag replacing the native rgsP at its chromosomal location. 27 RgsP interaction partner candidates were identified (S1 Table). Among these, the hypothetical transmembrane protein RgsM (SMc02432) was most abundant and identified with the highest number of unique peptides. Co-IP with RgsM-3×FLAG resulted in identification of RgsP, further supporting an interaction between the two proteins (S2 Table).
Analysis of RgsM amino acid sequence with transmembrane topology and signal peptide prediction tool PHOBIUS [27] suggested cytoplasmic localization of the amino acids 1–31, a short hydrophobic transmembrane α-helical region (amino acids 32–57) and periplasmic localization of the remaining C-terminal portion (Fig 4A). Amino acids 508–606 of RgsM represent a conserved peptidase M23 domain (pfam01551) often referred to as a LytM domain. This domain is predicted to have PG endopeptidase activity and is characteristic for zinc-dependent metallopeptidases. The LytM domain of RgsM contains a conserved HxxxD motif (S3 Table), which is required for zinc ion coordination and hydrolysis of glycine-glycine bonds in the peptides of staphylococcal PG by Staphylococcus aureus LytM [28–30]. The remaining RgsM amino acid sequence did not provide any hint about its possible function. To verify the predicted membrane topology of RgsM, we fused this protein to a truncated E. coli alkaline phosphatase PhoA, which is only active in the periplasm, and is missing its own signal peptide. The phosphatase detection assay revealed that PhoA, following RgsM1-66, indeed localized to the periplasm in both S. meliloti and E. coli. This strongly suggests the periplasmic localization of RgsM60-646 (S10 Fig).
Interaction between RgsP and RgsM was verified by a bacterial two-hybrid assay [31]. In this assay, putative interaction partners, fused to Bordetella pertussis adenylate cyclase fragments T18 and T25, are produced in an E. coli adenylate cyclase-deficient strain. Interaction between the two fusion proteins results in reconstitution of a functional enzyme, which activates expression of the lacZ reporter gene. Simultaneous production of T18-RgsM and RgsP∆GGDEF∆EAL-T25 fusion proteins, as well as of T18-RgsM and T25-RgsM, resulted in increased β-galactosidase activity, indicating protein-protein interactions (Fig 4B). Thus, RgsM was able to interact with RgsP and to homodimerize in a heterologous host. This is consistent with the Co-IP data that suggested RgsP-RgsM interaction in S. meliloti.
Attempts to generate a rgsM knockout mutant failed, suggesting that rgsM is essential. Similar to the growth phase-dependent accumulation of RgsP, C-terminally 3×FLAG-tagged RgsM was only detected in growing cells but not in stationary phase cells (S1A Fig). Next, we constructed a RgsM depletion strain Rm2011 rgsMdpl by placing the chromosomal rgsM gene under the control of an IPTG-inducible promoter. Depletion of a C-terminally 3×FLAG-tagged RgsM variant was confirmed using the same genetic setup (S1B Fig). The growth of Rm2011 rgsMdpl was significantly impaired in the absence of IPTG (Fig 5A). Strikingly similar to cells depleted of RgsP, RgsM-depleted cells lost the wild type rod shape (Figs 5B and 1B; S2 Fig). Electron microscopy revealed regions of low electron density in RgsM-depleted cells (Fig 5C; S3B Fig). 7.1% of these cells were stained by propidium iodide, indicating both lethal and bacteriostatic effects of RgsM depletion, similar to RgsP depletion (S4A Fig). Following depletion of RgsM for 12 or 24 h, the number of colony forming units was reduced by 89.4% or 99.5%, respectively, relative to cultures pre-incubated under non-depletion conditions (S4B Fig). We microscopically monitored growth of cells, previously grown under depletion conditions for 24 h. 10.9% of the previously RgsM-depleted cells gave rise to microcolonies on medium supplemented with IPTG, whereas 99.1% of these cells did not divide on medium lacking IPTG during the 12 h observation period (S4C Fig).
Labeling of RgsM-depleted cells with HADA revealed incorporation of new PG in only a minor part of pre-divisional cells at the septal region, whereas Rm2011 rgsMdpl grown in IPTG-containing medium displayed a wild type-like HADA staining (Fig 5D; S5A Fig). Similar to RgsP depletion, RgsM depletion resulted in an increase in the proportion of pre-divisional cell doublets (with visible septum constriction) to 40% (S5B Fig). Thus, like RgsP, RgsM is required for cell growth and division and PG biosynthesis.
To analyze RgsM localization relative to RgsP, we constructed a strain with rgsP-mCherry and mVenus-rgsM replacing rgsP and rgsM at their native genomic locations. mVenus-RgsM accumulated at one cell pole or at mid-cell and showed co-localization with RgsP-mCherry throughout the cell cycle (Fig 6A and 6B). This indicates that RgsM, like RgsP, localizes to sites of zonal cell wall synthesis. Localization of mVenus-RgsM was dependent on RgsP, since RgsP depletion resulted in diffuse mVenus-RgsM fluorescence (Fig 6C). Likewise, in RgsM-depleted cells, only diffuse RgsP-mCherry signal was observed (Fig 6D), indicating that polar and septal localization of RgsP and RgsM is mutually dependent. Taken together, these data imply a functional relation between RgsP and RgsM.
To further characterize the role of RgsM, we analyzed effects of rgsM overexpression. When grown in TY medium, RgsM-overproducing cells were indistinguishable from the empty vector control (S11 Fig). Contrastingly, in LB medium they grew very poorly and appeared enlarged and spherical (Fig 7A and 7B), with dramatically enlarged periplasm and inner membrane invaginations (Fig 7C; S3C Fig). Overexpression of rgsM in LB impaired PG biosynthesis, as judged from a very weak dispersed HADA staining, only visible after adjusting HADA fluorescence signal intensity, in contrast to cells carrying an empty vector, which showed polar and septal HADA signals (Fig 7D; S5A and S11C Figs). Noteworthy, even in the absence of RgsM overexpression, the rod cell shape differed slightly in TY and LB, with broader and shorter cells grown in LB (S11B Fig). Since the most prominent difference between the composition of TY and LB media is the content of NaCl (86 mM in LB, none in TY) and CaCl2 (2.7 mM in TY, none in LB), we analyzed the effects of these salts on the growth of RgsM-overproducing cells. Increasing the NaCl concentration to 300 mM in LB alleviated the rgsM overexpression-associated morphology and growth defects, and addition of CaCl2 to 2.7 mM resulted in wild type-like growth (S12 Fig). Replacement of Zn2+ with Ca2+ in RgsM may inactivate the protein and thereby mitigate the effect of rgsM overexpression. However, this explanation for the effect of CaCl2 is unlikely since the presence of Ca2+ in the growth medium did not negatively affect growth of S. meliloti, as RgsM depeletion did. Thus, the growth defect resulting from rgsM overexpression in LB may be a combined effect of an artificial increase in RgsM abundance and outer membrane destabilization in the absence of calcium [32]. This was probably counteracted by elevated medium osmolarity, which is supposed to reduce turgor.
To analyze the putative RgsM metallopeptidase active site H510xxxD, we constructed the RgsMH510A variant. The H510A mutation mitigated the strong cell morphology defect caused by RgsM overproduction in LB. Notably, both in TY and LB, growth of the rgsMH510A overexpressing cells was inhibited and morphology was altered similar to RgsM-depleted cells (S11 Fig), suggesting a possible dominant negative effect of RgsMH510A overproduction on the native RgsM function. Accumulation of overproduced RgsMwt and RgsMH510A in Rm2011 was confirmed by detecting the corresponding 3×FLAG-tagged variants (S1E Fig).
Overexpression of rgsM, but not rgsMH510A, compromised the cell envelope of E. coli and resulted in cell lysis in LB lacking NaCl. This was detected using the β-galactosidase substrate chlorophenol red-β-D-galactopyranoside (CPRG) as an indicator of cell lysis and increased membrane permeability [33] and by microscopy (Fig 7E; S13A Fig). As in S. meliloti, overproduction of wild type RgsM and RgsMH510A inhibited growth of E. coli in LB medium (S13B Fig). Both corresponding 3×FLAG-tagged RgsM variants accumulated in E. coli (S1E Fig).
To investigate the role of RgsP and RgsM in cell wall biogenesis, we determined the muropeptide composition of PG isolated from S. meliloti Rm2011 depleted and non-depleted of RgsP or RgsM. Muropeptide profiles of the depletion strains grown in TY under non-depletion conditions were similar to those obtained for the Rm2011 wild type (S14A Fig). Depletion of RgsP or RgsM resulted in alterations in the relative abundance of specific muropeptides. In both strains, the uncross-linked pentapeptides and the 3-3-cross-linked TetraTri dimers accumulated, whereas the 4-3-cross-linked TetraTetra dimers and TetraTetraTetra trimers were less abundant than in non-depleted cells (Fig 8A; S14 and S15 Figs; S4 Table).
Rm2011 cells, overproducing RgsM in LB, accumulated more Penta and TetraPenta muropeptides compared to cells harboring the empty vector control, whereas overproduction of RgsMH510A resulted in similar but less pronounced changes in the muropeptide profiles (S16A Fig). The accumulation of pentapeptides upon either depletion or overproduction of RgsP and RgsM points to altered incorporation and/or processing of new material into the growing PG sacculus, and the increase in 3–3 cross-links suggests increased LD-transpeptidase activity. Both might be indirect effects in cells with impaired PG growth, in response to stress. Muropeptides of E. coli S17-1 overproducing RgsM or RgsMH510A and grown in LB remained unchanged suggesting that the phenotypic changes of E. coli cells were independent of a putative PG hydrolase activity of RgsM (S16B Fig).
To gain further mechanistic insights into RgsP and RgsM functions in S. meliloti, purified His6-tagged variants of both proteins were assayed for PG binding in vitro. Whereas His6-RgsM57-646 was recovered in the insoluble fraction after incubation with PG, His6-RgsP7TMR-DISMED2 mostly remained in the soluble fraction (Fig 8B). This suggested that RgsM bound tightly to PG and that this was not the case for the periplasmic domain of RgsP (RgsP7TMR-DISMED2).
PG hydrolase activity of His6-RgsM57-646 alone or in combination with His6-RgsP7TMR-DISMED2 was assayed with S. meliloti PG as substrate. No evident changes in the muropeptide profiles were observed, indicating that His6-RgsM57-646 did not hydrolyze PG in vitro (S17A Fig). Since we observed a prominent proteolysis product of RgsM in Western blot analysis (S1E Fig), we asked if this might represent the active form of the enzyme. Therefore, N-terminally truncated version of His6-RgsM (amino acids 260–646) was analyzed. Although this fragment retained the ability to bind PG, it did not show PG hydrolase activity in vitro (S17B and S17C Fig). It cannot be excluded that the assay conditions were not optimal or that an additional factor is required to activate RgsM. Overall, the effects of RgsM and RgsP depletion or overproduction on the muropeptide profiles and strong PG binding by RgsM further support involvement of both proteins in PG biogenesis.
Comparative protein sequence analysis using BLASTP revealed conservation of RgsP and RgsM in Rhizobiales, Rhodobacterales, and in a single γ-proteobacterial species, whereas no homologs were detected in the remaining eubacterial phyla (S3 Table). Thus, we asked whether the homologous proteins from other species were also functionally conserved. Rhizobium etli and A. tumefaciens rgsP homologs RHE_CH00976 (rgsPRe) and Atu0784 (rgsPAt) were translationally fused to egfp at their native genomic locations. These tagged proteins displayed polar and septal localization corresponding to the HADA-staining zones in R. etli and A. tumefaciens, similar to the labeled RgsP variant in S. meliloti (S18A Fig; Fig 2). We also tested the ability of A. tumefaciens and R. etli rgsP and rgsM homologs, ectopically expressed from a taurine-inducible promoter, to complement S. meliloti RgsP and RgsM depletion strains. Without promoter induction ('leaky expression'), rgsPRe, rgsPAt and rgsMRe fully complemented the growth and morphology defects of the respective S. meliloti depletion strains, similar to the S. meliloti homologs, whereas induction with taurine was required for complementation of RgsM depletion with rgsMAt (S18B and S18C Fig). Similar subcellular localization of RgsP homologs in S. meliloti, R. etli and A. tumefaciens, and cross-complementation of rgsP and rgsM between these species provide evidence for functional conservation of both proteins in the Rhizobiales.
Members of the Rhizobiales, including S. meliloti, show unipolar cell growth [3]. However, the molecular mechanisms governing polar growth of the PG sacculus are largely unknown in these bacteria. In this study, we provide evidence suggesting important roles for RgsP (a member of the seven-transmembrane receptor family 7TMR-DISM) and its interaction partner RgsM (a putative PG metallopeptidase) in polar cell wall growth.
Proteins associated with the polar PG growth zones in A. tumefaciens include the division scaffold proteins FtsZ and FtsA, PG synthase PBP1a and LD-transpeptidase Atu0845 [34,35], which are directly or indirectly involved in PG biosynthesis. RgsP does not show homology to known cell division or PG biosynthesis proteins; however, it localizes to sites of zonal cell wall synthesis in S. meliloti, R. etli and A. tumefaciens. This feature is likely to be conserved in α-rhizobia containing a RgsP homolog. The phenotypes caused by RgsP depletion in S. meliloti − i.e. growth inhibition, altered cell shape, reduced incorporation of HADA and accumulation of penta-muropeptides in the PG sacculus − imply a regulatory role of RgsP in PG biosynthesis. Although known protein components of the cell elongation and division machineries have not been detected in the RgsP and RgsM pull-down assays, indirect, transient or low affinity interactions of RgsP or RgsM with such proteins may have escaped this analysis.
In E. coli, pentapeptides are specific for newly synthesized PG and are quickly processed as PG matures [36] by trans-, endo- and carboxypeptidase reactions [7]. In agreement with a previous report [3], we did not detect pronounced pentapeptide peaks in the S. meliloti wild type muropeptide profile. Perhaps in the wild type these pentapeptides were quickly processed but accumulated in RgsP-depleted cells due to impaired PG maturation. In E. coli, lack of the PG carboxypeptidase PBP5 results in an increase in the PG pentapeptide content; an effect which is further augmented by eliminating endopeptidases PBP4 and PBP7 [37]. Along with accumulation of monomeric pentapeptides, tetrapeptide dimer and trimer levels decreased upon RgsP depletion, indicating impaired PG cross-linking or enhanced hydrolysis of the peptide bridges. The cell shape can be influenced by the degree of PG cross-linking [38–44]. Hence, the impaired growth and altered cell shape of RgsP-depleted cells might have resulted from dysregulation of PG remodeling.
Cell wall biogenesis by necessity involves incorporation of new PG material into the existing PG sacculus, and this requires local hydrolysis of peptide bridges by PG endopeptidases. These include proteins with a M23 metallopeptidase (or LytM) domain. In E. coli and Vibrio cholerae, single genetic knockdowns of LytM domain proteins were not lethal because of genetic redundancy [45,46]. By contrast, the LytM domain protein RgsM is essential in S. meliloti, which is in agreement with a previous report [24]. The A. tumefaciens RgsM and RgsP orthologues were also shown to be essential [47]. Although we were not able to detect RgsM PG hydrolase activity in vitro, several findings are in agreement with an enzymatic activity in vivo. The RgsM LytM domain contains conserved histidine residues, which have been found to be essential for PG hydrolase activity [28,30,48]. Moreover, rgsM overexpression in S. meliloti destabilized the cell envelope and caused perturbations in the muropeptide composition. However, expression of rgsM in E. coli resulted in cell lysis but did not cause changes in the muropeptide profile. Structural studies of S. aureus LytM and Neisseria meningitidis LytM domain protein NMB0315 suggested the full-length proteins to adopt conformations interfering with enzymatic function, implying activation by proteolysis or protein-protein interactions in vivo [28,49]. Thus, we speculate that in vivo, RgsM may hydrolyze PG once activated by a yet-unknown factor. Alternatively, RgsM may have a regulatory role similar to E. coli and C. crescentus LytM domain proteins, which activate amidases to cleave septal PG [50–54]. Some LytM domain proteins also contain PG-binding domains [53,55]. PG binding by RgsM, demonstrated in vitro, is in agreement with an enzymatic or regulatory role of this protein.
Based on the mutual pull-down of RgsP and RgsM, an interaction in a bacterial two-hydrid assay and similar phenotypes of RgsM- and RgsP-depleted cells, we suggest that both proteins are involved in the same regulatory pathway. We narrowed down the essential part of RgsP to the N-terminal section including the 7TMR-DISM and PAS domains, which both may have a sensory function. In bacteria, 7TMR-DISMs are best understood in P. aeruginosa. In this bacterium, ligand binding to 7TMR-DISMED2 domains and their homodimerization were described as regulatory cues [21,22,56,57]. We speculate that RgsP either homodimerizes or, once triggered by an unknown cue, interacts with RgsM, to modulate RgsM dimerization and activity. This model is in agreement with the similar phenotypes of RgsP- and RgsM-depleted cells and the dominant effect of RgsMH510A overproduction in S. meliloti. This enzymatically inactive protein variant may compete with native RgsM for the interaction sites on RgsP or other interacting proteins.
Although abundance of RgsP lacking the PAS domain was only moderately reduced, the localization and growth-promoting function of this protein was dramatically affected. A role of PAS domains or PAS-like motifs in polar protein localization was previously reported for example in C. crescentus. These include the C. crescentus single PAS domain protein MopJ, which directly interacts with the polarly localized histidine protein kinases DivJ and CckA, which are both involved in cell cycle regulation [58,59]. We speculate that the PAS domain may be important for recruitment of RgsP to PG biosynthesis sites. The nature of the signal, perceived by the RgsP PAS domain, and its regulatory output remain to be investigated.
RgsP PDE activity substantially contributes to c-di-GMP turnover in S. meliloti. To our knowledge, RgsP is the only c-di-GMP PDE which is polarly localized in S. meliloti. Polarly localized c-di-GMP PDEs in C. crescentus and P. aeruginosa contribute to heterogeneity in the cellular c-di-GMP content [60–62]. Lack of the RgsP PDE activity did not cause apparent phenotypic defects. Yet, we cannot exclude that RgsP PDE activity has a regulatory function in S. meliloti, which may be linked to its localization. A regulatory function may have escaped our phenotypic analyses or may not have been detected because of compensation by any of the twelve additional c-di-GMP PDEs encoded by the S. meliloti genome [17].
Whereas the RgsPSm GGDEF domain seems to be inactive because of a degenerate active site [17], RgsPAt contains an intact GGDEF motif, suggesting enzymatic activity of this protein. Since RgsPAt complemented depletion of RgsPSm in S. meliloti, the DGC activity of RgsP is unlikely to be relevant for its growth-promoting function. However, the GGDEF domain may modulate PDE activity of the EAL domain, as for example in case of C. crescentus PdeA or P. aeruginosa BifA and RmcA [61,63,64].
Overall, our study suggests that compared to well-studied γ-proteobacterial models, α-rhizobia utilize different sets of proteins for PG metabolism during cell elongation and division, and for relocating the PG growth machinery from a pole to the cell division site. Identification of further enzymes involved in PG synthesis and remodeling, and understanding the regulatory roles of RgsP and RgsM will lead to the clarification of specific mechanisms for polar PG biogenesis in α-rhizobia.
Bacterial strains and plasmids used in this study are shown in S5 Table. S. meliloti was grown at 30°C in tryptone-yeast extract (TY) medium [65], LB medium [66], modified MOPS-buffered minimal medium [67], and nutrient-depleted 30% minimal medium (nitrogen, carbon, and phosphate sources reduced to 30%). R. etli was grown in TY medium and A. tumefaciens was grown in LB medium at 30°C. E. coli was grown in LB medium at 37°C. For S. meliloti, antibiotics were used at the following concentrations (mg/l; liquid/solid medium): kanamycin, 100/200, gentamicin, 15/40, tetracycline, 5/10, spectinomycin, 200/200 and streptomycin, 600/600. For E. coli the following concentrations were used: kanamycin, 50/50, gentamicin, 5/8, tetracyclin, 5/10, spectinomycin 100/100, and ampicillin 100/100. For A. tumefaciens kanamycin was used at 100/200 and for R. etli at 30/30. Unless otherwise specified, the inducers isopropyl β-D-1-thiogalactopyranoside (IPTG) and taurine were added at 0.5 and 20 mM, respectively. Chlorophenol red-β-D-galactopyranoside (CPRG), an almost membrane-impermeable β-galactosidase substrate previously used to detect compromised E. coli cell envelope and lysis by purple staining of agar cultures [33], was added at 20 μg/ml. Alkaline phosphatase substrate 5-bromo-4-chloro-3-indolylphosphate (BCIP) was used at 50 μg/ml and β-galactosidase substrate 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-Gal) was used at 40 μg/ml.
Growth assays were performed using 100 μl cultures in flat-bottom 96-well plates (Greiner), grown at 30 oC with shaking at 1,200 rpm. Three to six culture replicates were analyzed per strain. Optical density (OD600) was recorded using Infinite M200 PRO fluorescence reader (Tecan). For growth assays involving protein depletion, cultures with or without IPTG were inoculated with 0.15 μl of stationary TY preculture with IPTG and grown for 24 h. Relative growth was calculated as a ratio of OD600 of cells grown without IPTG and OD600 of cultures supplemented with 0.5 mM IPTG. For growth assays, involving taurine- or IPTG-induced gene overexpression, the cultures containing or not containing taurine or IPTG were inoculated with 0.15 μl of stationary TY preculture and the growth was recorded at the indicated time points.
TY medium (5 g/l tryptone, 3 g/l yeast extract, 0.4 g CaCl2×2H2O). LB medium (10 g/l tryptone, 5 g/l yeast extract, 5 g/l NaCl). MOPS-buffered minimal medium (MM) (10 g/l MOPS, 10 g/l mannitol, 3.55 g/l sodium glutamate, 0.246 g/l MgSO4×7H2O, 0.25 mM CaCl2, 2 mM K2HPO4, 10 mg/l FeCl3×6H2O, 1 mg/l biotin, 3 mg/l H3BO3, 2.23 mg/l MnSO4×H2O, 0.287 mg/l ZnSO4×7H2O, 0.125 mg/l CuSO4×5H2O, 0.065 mg/l CoCl2×6H2O, 0.12 mg/l NaMoO4×2H2O, pH 7.2).
Constructs used in this work were generated using standard cloning techniques and are listed in S5 Table. The primers used are listed in S6 Table. All constructs were verified by sequencing. Plasmids were transferred to S. meliloti by E. coli S17-1-mediated conjugation as previously described [68]. Electroporation was used to introduce plasmids to R. etli and A. tumefaciens following the protocol previously described [69].
To generate chromosomally integrated constructs encoding RgsP or RgsM with C-terminally fused enhanced green fluorescent protein (EGFP) or triple FLAG-tag (3×FLAG), the 700 to 800 bp 3' portion of the gene excluding the stop codon was cloned into suicide plasmids pK18mob2-egfp or pG18mob-3×flag yielding translational fusions of the C-terminal portion of the protein coding sequence to the corresponding tag. Integration of these gene fusion constructs into the S. meliloti genome by homologous recombination resulted in a replacement of the native gene copy with egfp- or 3xflag-tagged gene copy at the corresponding native chromosomal location.
To construct markerless translational fusions of rgsP and rgsM to egfp or mCherry at the native chromosomal location, the 700–800 bp 3' portion of the gene fused to egfp or mCherry was cloned into suicide plasmid pK18mobsacB [70] together with 700–800 bp of adjacent downstream genomic region. The resulting constructs were introduced into S. meliloti and transconjugants were subjected to sucrose selection to obtain double recombinants that have lost the integrated vector [70]. Correct positions of chromosomally encoded gene fusions were verified by PCR.
To obtain RgsP and RgsM depletion strains, plasmids designed to uncouple the native promoter from the coding sequence and to place the coding sequence under the control of IPTG-inducible promoters were constructed. The rgsP gene is most likely co-transcribed with the preceding rimJ gene, encoding a probable ribosomal-protein-alanine acetyltransferase (S7A Fig). To uncouple transcription of rimJ and rgsP, and to place rgsP under the control of an IPTG-inducible promoter, the lac-T5 tandem promoter sequence was inserted between rgsP and rimJ without altering the rimJ open reading frame. To this end, a DNA fragment containing the IPTG-inducible T5 promoter and a Shine-Dalgarno sequence followed by the partial 5' rgsP coding sequence starting from the start codon (586 bp) was PCR-amplified from rgsP expression plasmid pWBT-SMc00074 [17]. This fragment was inserted into pK18mob2 [70] downstream of the IPTG-inducible lac promoter, thus generating a lac-T5 tandem promoter. In the case of RgsM depletion, the partial 5' rgsM coding sequence starting from the start codon (406 bp) was PCR-amplified from S. meliloti genomic DNA and inserted into pK18mob2 downstream of the IPTG-inducible lac promoter. Integration of these constructs into the genome placed the full-length protein coding sequence under the control of the corresponding IPTG-inducible promoter. Adjacent open reading frames were not affected by integration of these constructs. Conditional RgsP and RgsM depletion was then achieved by constitutively expressing the lacI repressor gene from vector pSRKGm in the strains Rm2011 rgsPdpl and Rm2011 rgsMdpl, respectively.
Gene overexpression constructs were generated by insertion of the corresponding coding sequences downstream of either the IPTG-inducible lac promoter in medium-copy vectors pSRKKm and pSRKGm [71], the IPTG-inducible T5 promoter in medium-copy vector pWBT, the taurine-inducible tauA promoter in low-copy vector pR-Ptau, or a constitutive synthetic promoter (Psyn) in low-copy vector pR_egfp [72]. In the case of pR_egfp, a 114 bp region upstream of the rgsP coding region was included.
To generate single-copy ectopic rgsP or rgsP-egfp expression constructs, we first determined the native rgsP promoter and its activity levels. At its native chromosomal locus, rgsP is most likely in an operon with rimJ. Therefore, a 300 bp genomic region upstream of rgsP as well as a 944 bp region including the putative rimJ promoter and the whole rimJ coding sequence (S7A Fig) were tested for promoter activity using fusions to egfp (S7B Fig). Since higher levels of egfp expression were observed in the case of the 944 bp fragment, we used this DNA region as native rgsP promoter. To exclude possible non-desirable effects of an additional rimJ copy, a nonsense mutation was introduced 64 nucleotides downstream of the rimJ start codon yielding promoter construct P*rgsP (S7A Fig). The rgsP, rgsP-egfp or rgsP-3×flag coding sequences were inserted downstream of the P*rgsP sequence in single-copy plasmid pABC2S-mob [73]. For generation of amino acid substitutions or protein variants lacking specific domains splicing by overlap extension PCR was applied.
The rgsP promoter-egfp fusions were generated by insertion of the upstream non-coding region (either long (944 bp) or short (300 bp)) and the three first codons of rgsP into replicative medium-copy number plasmid pSRKKm-egfp [17]. Constructs for purification of His6-tagged proteins were generated by insertion of the coding sequence excluding the start codon into expression vector pWH844 [74]. Fusions of rgsM to phoA were assembled from the full-length or partial rgsM coding sequence and the E. coli phoA coding sequence missing the first 26 codons.
For promoter-egfp activity assays, TY overnight cultures were diluted 1:500 in 100 μl of TY medium or 30% MM and grown in 96-well plates at 30°C with shaking at 1,200 rpm. EGFP fluorescence (excitation 488 ± 9 nm; emission 522 ± 20 nm, gain 82) and OD600 were determined using the Infinite 200 Pro multimode reader (Tecan) and calculated as relative fluorescence units (RFU), which represent fluorescence values divided by OD600. Background EGFP fluorescence was determined using a control strain harboring pSRKKm-egfp. Fluorescence of three independent transconjugants was measured as biological replicates.
Microscopy of bacteria on 1% agarose pads was performed using the Nikon microscope Eclipse Ti-E equipped with a differential interference contrast (DIC) CFI Apochromat TIRF oil objective (100x; numerical aperture of 1.49) and a phase-contrast Plan Apo l oil objective (100x; numerical aperture, 1.45) with the AHF HC filter sets F36-513 DAPI (excitation band pass [ex bp] 387/11 nm, beam splitter [bs] 409 nm, and emission [em] bp 447/60 nm), F36-504 mCherry (ex bp 562/40 nm, bs 593 nm, and em bp 624/40 nm), F36-525 EGFP (ex bp 472/30 nm, bs 495 nm, and em bp 520/35 nm) and F36-528 YFP (ex bp 500/24 nm, bs 520 nm, and em bp 542/27 nm). Images were acquired with an Andor iXon3 885 electron-multiplying charge-coupled device (EMCCD) camera. For time-lapse analysis, MM 1% agarose pads were used, and images were acquired every 2, 4 or 15 min at 30°C. IPTG was added to the medium at 0.2 or 0.5 mM for microscopy of cells harboring pSRKGm-parB-mCherry or pSRKGm-mCherry-ftsZ, respectively.
Treatment of S. meliloti, R. etli and A. tumefaciens cells with fluorescently-labeled D-amino acid HADA was performed as previously described [75]. Briefly, cells were grown for 24 h in liquid medium in glass tubes to an OD600 of 0.4–0.6. 80 μl of the cultures were then mixed with 0.25 μl 100 mM HADA solution and incubated for 2.5 min (A. tumefaciens), 3 min (S. meliloti) or 3.5 min (R. etli) at 30 oC with shaking at 800 rpm. After addition of 186 μl 100% ethanol and 5–20 min incubation at room temperature (RT), cells were washed three times with 0.9% NaCl and subsequently placed onto 1% agarose pads.
Cell viability was assessed by DNA staining with the fluorescent intercalating agent propidium iodide. 100 μl of S. meliloti liquid cultures (OD600 0.4–0.8) were mixed with 1 μl of 2 mg/ml propidium iodide stock solution and incubated for 5 min at room temperature. Cells were washed three times with 0.9% NaCl and subsequently placed onto 1% agarose pads.
Concentrated S. meliloti cell suspensions were high pressure frozen (HPF Compact 02, Wohlwend, CH) and freeze-substituted (AFS2, Leica, Wetzlar, Germany) in a medium based on acetone, containing 0.25% osmium tetroxide, 0.2% uranyl acetate and 5% ddH2O according to the following protocol: -90°C for 20 h, from -90°C to -60°C in 1 h, -60°C for 8 h, -60°C to -30°C in 1 h, -30°C for 8 h, -30°C to 0°C in 1 h, 0°C for 3 h. Still at 0°C, samples were washed three times with acetone before a 1:1 mixture of Epon 812 substitute resin (Fluka, Buchs, CH) and acetone was applied at room temperature for 2 h. The 1:1 mixture was substituted with pure resin to impregnate the samples overnight. After another substitution with fresh Epon, samples were polymerized at 60°C for 2 days. The sample containing polymerized Epon blocks were then trimmed with razor blades and cut to 50 nm ultrathin sections using an ultramicrotome (UC7, Leica, Wetzlar, Germany) and a diamond knife (Diatome, Biel, Switzerland). Sections were applied onto 100 mesh copper grids coated with pioloform. For additional contrast, mounted sections were post-stained with 2% uranyl acetate for 20–30 min and subsequently with lead citrate for another 1–2 min. The sections were finally analyzed and imaged using a JEM-2100 transmission electron microscope (JEOL, Tokyo, Japan) equipped with a 2k x 2k F214 fast-scan CCD camera (TVIPS, Gauting, Germany).
Strain Rm2011 rgsPdpl ectopically expressing rgsP variants from P*rgsP was grown in triplicates in liquid TY medium without IPTG and harvested in the exponential growth phase 24 h after inoculation. Quantification of intracellular c-di-GMP was performed as previously described [76]. Briefly, cells were collected by centrifugation and nucleotides were extracted three times with acetonitrile/methanol/water (2:2:1), dried and subjected to liquid chromatography-tandem mass spectrometry. c-di-GMP was normalized to total protein, determined using Bradford reagent (Bio-Rad).
Heterologous protein expression and purification was performed as previously described [17]. E. coli BL21(DE3) harboring expression plasmids were grown in LB medium in flasks to OD600 of 0.5–0.6 and protein expression was induced with 0.4 mM IPTG overnight at RT. Cells were lysed using French press (pressure 1,000 lb/in2) and the lysates were centrifuged for 60 min at 24,000 ×g and 4°C. Cleared lysates were applied to His SpinTrap columns (GE Healthcare) following the manufacturer’s instructions and eluted with 0.5 M imidazole. Purity of isolated proteins was assessed by SDS-PAGE and Coomassie brilliant blue staining. Protein concentration was determined using Bradford reagent (Bio-Rad).
[α-32P]-labeled c-di-GMP was synthesized using purified Caulobacter crescentus His6-DgcA at 10 μM from GTP and [α-32P]-GTP (0.1 μCi/μl) at 1 mM in the reaction buffer (50 mM Tris-HCl, 300 mM NaCl, 10 mM MgCl2, pH 8.0), overnight at 30°C. The reaction was then treated with 5 units of calf intestine alkaline phosphatase (Fermentas) for 1 h at 22°C to hydrolyze unreacted GTP and stopped by incubation for 10 min at 95°C. The precipitated proteins were removed by centrifugation (10 min, 20,000 ×g, 22°C) and the supernatant was used for the PDE activity and the c-di-GMP binding assays.
c-di-GMP binding was determined using a differential radial capillary action of ligand assay (DRaCALA) with [α-32P]-labeled c-di-GMP, as previously described [77]. This assay is based on the ability of dry nitrocellulose to prevent diffusion of bound protein-ligand complexes and thereby separate them from free ligand. Reaction mixtures (50 μl) containing [α-32P]-labeled c-di-GMP and 20 μM of indicated protein in the binding buffer (10 mM Tris, 100 mM NaCl, 5 mM MgCl2, pH 8.0) were incubated for 10 min at RT. 10 μl of this reaction mixture was spotted onto nitrocellulose membrane and allowed to dry prior to exposing a phosphor-imaging screen (Molecular Dynamics). Data were collected using a STORM 840 scanner.
DGC and PDE activities were determined as previously described [78,79]. Reaction mixtures (40 μl) containing purified proteins at 10 μM, in reaction buffer (50 mM Tris-HCl, 300 mM NaCl, 10 mM MgCl2, pH 8.0) were first pre-incubated for 5 min at 30°C. DGC reactions were initiated by adding GTP/[α-32P]-GTP (0.1 μCi/μl) to 1 mM, incubated at 30°C for the indicated periods of time and stopped by adding an equal volume of 0.5 M EDTA. PDE reactions were initiated by adding [α-32P]-labeled c-di-GMP and stopped by adding an equal volume of 0.5 M EDTA after indicated time periods. 2 μl of the PDE or DGC reaction mixtures were spotted on polyethyleneimine-cellulose TLC chromatography plates, developed in 2:3 (v/v) 4 M (NH4)2SO4/1.5 M KH2PO4 (pH 3.65). Plates were dried prior to exposing a phosphor-imaging screen (Molecular Dynamics). Data were collected and analyzed using a STORM 840 scanner (Amersham Biosciences).
Co-IP and protein identification by mass spectrometry was performed as previously described including small modifications [80]. Cultures of Rm2011 rgsP-3×flag, Rm2011 rgsM-3×flag and control strain Rm2011 harboring the empty vector pWBT were grown in TY medium in flasks to an OD600 of 0.6 and cross-linked with 0.1% formaldehyde for 15 min at RT. Reaction was quenched by adding glycine at a final concentration of 0.35 M. Cells were washed, resuspended in lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 2 mM phenylmethylsulfonyl-fluorid [PMSF], pH 7.4) and lysed using a French press (pressure 1,000 lb/in2). Cleared lysates obtained after ultracentrifugation (100,000 ×g, 1 h, 4°C) were incubated with anti-FLAG M2 affinity gel (FLAG Immunoprecipitation Kit, Sigma) overnight at 4°C on a rolling shaker. Bound proteins were eluted with 3×FLAG peptide solution.
For mass-spectrometry analysis, proteins were digested by Sequencing Grade Modified Trypsin (Promega) at 37°C overnight. The mass spectrometric analysis was performed using an Orbitrap Velos Pro mass spectrometer (Thermo Fisher Scientific). An Ultimate nanoRSLC-HPLC system (Dionex), equipped with a custom 20 cm x 75 μm C18 RP column filled with 1.7 μm beads was connected online to the mass spectrometer through a Proxeon nanospray source. 1-15 μl of the tryptic digest were injected onto a C18 pre-concentration column. Automated trapping and desalting of the sample was performed at a flow rate of 6 μl/min using water/0.05% formic acid as solvent. Separation of the tryptic peptides was achieved with the following gradient of water/0.05% formic acid (solvent A) and 80% acetonitrile/0.045% formic acid (solvent B) at a flow rate of 300 nl/min: holding 4% B for 5 min, followed by a linear gradient to 45% B within 30 min and linear increase to 95% solvent B in additional 5 min. The column was connected to a stainless steel nanoemitter (Proxeon, Denmark) and the eluent was sprayed directly towards the heated capillary of the mass spectrometer using a potential of 2,300 V. A survey scan with a resolution of 60,000 within the Orbitrap mass analyzer was combined with at least three data-dependent MS/MS scans with dynamic exclusion for 30 s either using CID with the linear ion-trap or using HCD combined with orbitrap detection at a resolution of 7,500. Data analysis was performed using Proteome Discoverer (Thermo Fisher Scientific) with SEQUEST and MASCOT (version 2.2; Matrix science) search engines using either SwissProt or NCBI databases. The combined transmembrane topology and signal peptide prediction online tool PHOBIUS [27] was used to predict regions with high hydrophobicity within candidate protein interactions partners of RgsP and RgsM.
Bacterial two-hybrid analysis was performed as previously described [31]. The adenylate cyclase-deficient strain E. coli BTH101 was co-transformed with plasmids carrying rgsP∆GGDEF∆EAL or rgsM translationally fused to T25 and T18 fragments of Bordetella pertussis adenylate cyclase. Transformant colonies were grown in 100 μl LB supplemented with antibiotics at 30°C for 6 hours with shaking 1,200 rpm. 10 μl of each culture was spotted onto LB agar plates containing kanamycin, ampicillin, X-Gal and IPTG. Plates were imaged after 30 h of incubation at 30°C.
β-galactosidase activity was determined as previously described with small modifications [52]. Cells grown on LB agar containing IPTG were resuspended in 1 ml of Z-buffer (60 mM Na2HPO4, 40 mM NaH2PO4, 10 mM KCl, 1 mM MgSO4, pH 7.0), in triplicates and the OD600 was recorded. Cell permeabilization was facilitated by addition of 50 μl of chlorophorm and 50 μl of 0.05% SDS. The aqueous phase was mixed with an equal volume of Z-buffer containing 50 mM β-mercaptoethanol, and ortho-nitrophenyl-β-D-galactopyranoside (ONPG) was added to the final concentration of 0.5 mg/ml. A420 was recorded using the Infinite M200 PRO fluorescence reader (Tecan). Miller Units (MU) were calculated as follows: MU = 1000*A420/(t*V*OD600). t represents the time in min and V the volume in ml.
Western blot analysis was performed as previously described [68]. Briefly, S. meliloti and E. coli strains expressing 3×flag-tagged rgsP or rgsM variants were grown in glass tubes supplemented with corresponding antibiotics. Unless otherwise specified, cells were grown in TY medium without IPTG and collected at an OD600 of 0.4–0.8 24 h after inoculation. Cells were adjusted to an OD600 of 1, 10 μl of lysed cells were loaded to SDS-PAGE gel and separated proteins were transferred to PVDF membrane (Thermo Fisher Scientific). RgsP protein variants were detected using anti-FLAG M2-Peroxidase (HRP) antibody (Sigma-Aldrich). Membranes were incubated with Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific) and imaged using the luminescence image analyzer LAS-4000 (Fujifilm).
Isolation of PG sacculi was achieved following a published protocol [81]. Briefly, S. meliloti strains were grown in 400 ml TY and LB media in flasks for 24 h at 30°C until OD600 reached 0.4–0.8. E. coli strains were grown in 400 ml LB for 4 h at 37°C until OD600 reached 0.5–0.9. Cells were harvested by centrifugation, resuspended in PBS buffer, added dropwise to boiling 8% SDS solution and stirred for 30 min. PG sacculi were pelleted by ultracentrifugation in a Beckman Coulter Optima at 440,000 ×g, at ambient temperature for 1 h and resuspended in water. Centrifugation and resuspension steps were repeated until the sacculi were free of SDS as verified by a published assay [82]. The sacculi were then incubated with 100 μg/ml of α-amylase in reaction buffer (10 mM Tris-HCl, 10 mM NaCl, pH 7.0) for 2 h at 37°C, followed by incubation with 200 μg/ml pronase E for 1 h at 60°C, to remove high molecular weight glycogen and proteins, respectively. The enzymes were removed by incubation in 4% SDS solution for 15 min at 80°C. Sacculi were washed free of SDS as described above and resuspended in 0.02% sodium azide.
Purified sacculi were digested with the muramidase cellosyl (Hoechst, Frankfurt, Germany) at 37°C overnight. The reaction was terminated by boiling the sample at 100°C for 10 min. The samples were centrifuged (15,000 ×g for 10 min) and the soluble muropeptides were recovered from the supernatant, reduced with NaBH4 and separated by HPLC as described for E. coli [81]. Muropeptides were then separated on a C18 reversed-phase column (ProntoSIL) using an Agilent 1220 infinity HPLC system. Separation was carried out over a 180 min linear gradient from buffer A (50 mM sodium phosphate, pH 4.31) to buffer B (75 mM sodium phosphate, pH 4.95, 15% methanol). Muropetides were detected by their absorbance at 205 nm. Muropeptides of interest corresponding to major peaks were manually collected and analyzed by tandem mass spectrometry (MS/MS) as previously described [83].
Purified PG (~100 μg) from S. meliloti strain Rm2011 was centrifuged at 15,000 ×g, 4°C for 14 min and resuspended in binding buffer (10 mM Tris-maleate, 10 mM MgCl2, 50 mM NaCl, pH 6.8). 10 μg of protein of interest was incubated with or without PG in a final volume of 100 μl, and incubated at 4°C for 30 min. Samples were centrifuged as described above and the supernatant was collected (supernatant fraction) whilst the pellet was resuspended in 200 μl of binding buffer. Another centrifugation step as described above was carried out (wash fraction). Bound proteins were released from PG by incubation with 100 μl of 2% SDS, at 4°C for 1 h before being collected by a final centrifugation step as done at earlier steps. The proteins present in the different fractions were analyzed by SDS-PAGE.
Proteins (2 μM final concentration) were incubated with 10 μg of sacculi from S. meliloti, overnight at 37°C, in a volume of 100 μl (containing 10 mM HEPES-NaOH, 1 mM ZnCl2, 150 mM NaCl, 0.05% Triton X-100, pH 7.5). Proteins were inactivated by boiling for 10 min. Cellosyl was added (1 μM final concentration) and the samples were incubated overnight again at 37°C. Samples were boiled for 10 min and the soluble muropeptides were collected by centrifugation at 15,000 ×g for 10 min, and taking the supernatant fraction. Muropeptides were reduced with NaBH4 and analyzed by HPLC as described above.
|
10.1371/journal.pbio.2002985 | A population genetic interpretation of GWAS findings for human quantitative traits | Human genome-wide association studies (GWASs) are revealing the genetic architecture of anthropomorphic and biomedical traits, i.e., the frequencies and effect sizes of variants that contribute to heritable variation in a trait. To interpret these findings, we need to understand how genetic architecture is shaped by basic population genetics processes—notably, by mutation, natural selection, and genetic drift. Because many quantitative traits are subject to stabilizing selection and because genetic variation that affects one trait often affects many others, we model the genetic architecture of a focal trait that arises under stabilizing selection in a multidimensional trait space. We solve the model for the phenotypic distribution and allelic dynamics at steady state and derive robust, closed-form solutions for summary statistics of the genetic architecture. Our results provide a simple interpretation for missing heritability and why it varies among traits. They predict that the distribution of variances contributed by loci identified in GWASs is well approximated by a simple functional form that depends on a single parameter: the expected contribution to genetic variance of a strongly selected site affecting the trait. We test this prediction against the results of GWASs for height and body mass index (BMI) and find that it fits the data well, allowing us to make inferences about the degree of pleiotropy and mutational target size for these traits. Our findings help to explain why the GWAS for height explains more of the heritable variance than the similarly sized GWAS for BMI and to predict the increase in explained heritability with study sample size. Considering the demographic history of European populations, in which these GWASs were performed, we further find that most of the associations they identified likely involve mutations that arose shortly before or during the Out-of-Africa bottleneck at sites with selection coefficients around s = 10−3.
| One of the central goals of evolutionary genetics is to understand the processes that give rise to phenotypic variation in humans and other taxa. Genome-wide association studies (GWASs) in humans provide an unprecedented opportunity in that regard, revealing the genetic basis of variation in numerous traits. However, exploiting this opportunity requires models that relate genetic and population genetic processes with the discoveries emerging from GWASs. We present such a model and show that it can help explain the results of GWASs for height and body mass index. More generally, our results offer a simple interpretation of the findings emerging from GWASs and suggest how they relate to the evolutionary and genetic forces that give rise to phenotypic variation.
| Much of the phenotypic variation in human populations, including variation in morphological, life history, and biomedical traits, is “complex” or “quantitative”, in the sense that heritable variation in the trait is largely due to small contributions from many genetic variants segregating in the population [1,2]. Quantitative traits have been studied since the birth of biometrics over a century ago [1–3], but only in the past decade have technological advances made it possible to systematically dissect their genetic basis [4–6]. Notably, since 2007, genome-wide association studies (GWASs) in humans have led to the identification of many thousands of variants reproducibly associated with hundreds of quantitative traits, including susceptibility to a wide variety of diseases [4]. While still ongoing, these studies already provide important insights into the genetic architecture of quantitative traits, i.e., the number of variants that contribute to heritable variation, as well as their frequencies and effect sizes.
Perhaps the most striking observation to emerge from these studies is that, despite the large sample size of many GWASs, all variants significantly associated with a given trait typically account for less (often much less) than 25% of the narrow sense heritability ([4,7,8], but see [9]). (Henceforth, we use “heritability” to refer to narrow sense heritability.) While many factors have been hypothesized to contribute to the “missing heritability” [7,8,10–14], the most straightforward explanation and the emerging consensus is that much of the heritable variation derives from variants with frequencies that are too low or effect sizes that are too small for current studies to detect. Comparisons among traits also suggest that there are substantial differences in architectures. For example, recent meta-analysis GWASs uncovered 7 times as many variants for height (697) as for body mass index (97), and together, the variants for height account for more than 4 times the heritable variance than the variants for body mass index do (approximately 20% versus approximately 3%–5%, respectively), despite comparable sample sizes [15,16].
These first glimpses underscore the need for theory that relates the findings emerging from GWASs with the evolutionary processes that shape genetic architectures. Such theory would help to interpret the “missing heritability” [17–20] and to explain why architecture differs among traits. It may also allow us to use GWAS findings to make inferences about underlying evolutionary parameters, helping to answer enduring questions about the processes that maintain phenotypic variation in quantitative traits [5,21].
Development of such theory can be guided by empirical observations and first-principles considerations. New mutations affecting a trait arise at a rate that depends on its “mutational target size” (i.e., the number of sites at which a mutation would affect the trait). Once they arise, the trajectories of variants through the population are determined by the interplay between genetic drift, demographic processes, and natural selection acting on them. These processes determine the number and frequencies of segregating variants underlying variation in the trait. The genetic architecture further depends on the relationship between the selection on variants and their effects on the trait. Notably, selection on variants depends not only on their effect on the focal trait but also on their pleiotropic effects on other traits. We therefore expect both direct and pleiotropic selection to shape the joint distribution of allele frequencies and effect sizes.
Multiple lines of evidence suggest that many quantitative traits are subject to stabilizing selection, i.e., selection favoring an intermediate trait value [5,22–27]. For instance, a decline in fitness components (e.g., viability and fecundity) is observed with displacement from mean values for a variety of traits in human populations [28–30], in other species in the wild [31,32], and in experimental manipulations [31,33]. While less is known about complex diseases, they may often reflect large deviations of an underlying continuous trait from an optimal value [1], with these continuous traits subject to directional (purifying) selection in some cases and to stabilizing selection in others. What remains unclear is the extent to which stabilizing selection is acting directly on variation in a given trait or is “apparent”, i.e., results from pleiotropic effects of this variation on other traits.
Other lines of evidence suggest that pleiotropy is pervasive. For one, theoretical considerations about the variance in fitness in natural populations and its accompanying genetic load suggest that only a moderate number of independent traits can be effectively selected on at once [34]. Thus, the aforementioned relationships between the value of a focal trait and fitness are likely heavily affected by the pleiotropic effects of genetic variation on other traits [25,34–36]. Second, many of the variants detected in human GWASs have been found to be associated with more than one trait [37–41]. For example, a recent analysis of GWASs revealed that variants that delay the age of menarche in women tend to delay the age of voice drop in men, decrease body mass index, increase adult height, and decrease risk of male pattern baldness [37]. More generally, the extent of pleiotropy revealed by GWASs appears to be increasing rapidly with improvements in power and methodology [37,42–45]. These considerations and others [45,46] point to the general importance of pleiotropic selection on quantitative genetic variation.
The discoveries emerging from human GWASs further suggest that genetic variance is dominated by additive contributions from numerous variants with small effect sizes. Dominance and epistasis may be common among newly arising mutations of large effect (e.g., [47–51]), but both theory and data suggest that they play a minor role in shaping quantitative genetic variation within populations (e.g., [9,52–56]). Indeed, for many traits, most or all of the heritability explained in GWASs arises from the additive contribution of variants with squared effect sizes that are substantially smaller than the total genetic variance (e.g., [15,16,57,58]). Moreover, statistical quantifications of the total genetic variance tagged by genotyping (i.e., not only due to the genome-wide significant associations) suggest that such contributions may account for most of the heritable variance in many traits (e.g., [9,59–61]). Finally, considerable efforts to detect epistatic interactions in human GWASs have, by and large, come up empty-handed [9,56,62], with few counterexamples, mostly involving variants in the major histocompatibility complex region ([53,56,63,64], but see [65]). Thus, while the discovery of epistatic interactions may be somewhat limited by statistical power [56], theory and current evidence suggest that nonadditive interactions play a minor role in shaping human quantitative genetic variation. Motivated by these considerations, we model how direct and pleiotropic stabilizing selection shape the genetic architecture of continuous, quantitative traits by considering additive variants with small effects and assuming that together they account for most of the heritable variance.
To date, there has been relatively little theoretical work relating population genetics processes with the results emerging from GWASs. Moreover, the few existing models have reached divergent predictions about genetic architecture, largely because they make different assumptions about the effects of pleiotropy. Focusing on disease susceptibility, Pritchard [19] considered the “purely pleiotropic” extreme, in which selection on variants is independent of their effect on the trait being considered. In this case, we expect the largest contribution to genetic variance in a trait to come from mutations that have large effect sizes but are also weakly selected or neutral, allowing them to ascend to relatively high frequencies. Other studies considered the opposite extreme, in which selection on variants stems entirely from their effect on the trait under consideration [26,66–70], and have shown that the greatest contribution to genetic variance would arise from strongly selected mutations [67,68] (we return to this case below).
In practice, we expect most traits to fall somewhere in between these extremes. While there are compelling reasons to believe that quantitative genetic variation is highly pleiotropic, the effects of variants on different traits are likely to be correlated. Thus, even if a given trait is not subject to selection, variants that have a large effect on it will also tend to have larger effects on traits that are under selection (e.g., by causing large perturbation to pathways that affect multiple traits [36,45]). Motivated by such considerations, Eyre-Walker (2010) [20], Keightley and Hill (1990) [18], and Caballero et al. (2015) [71] considered models in which the correlation between the strength of selection on an allele and its effect size can vary between the purely pleiotropic and direct selection extremes. These models diverge in their predictions about architecture, however. Assuming, as seems plausible, an intermediate correlation between the strength of selection and effect size, Eyre-Walker finds that genetic variance should be dominated by strongly selected mutations [20], whereas Keightley and Hill and Caballero et al. conclude that the greatest contribution should arise from weakly selected ones [18,71]. Their conclusions differ because of how they chose to model the relationship between selection and effect size, a choice based largely on mathematical convenience. We approach this problem by explicitly modeling stabilizing selection on multiple traits, thereby learning, rather than assuming, the relationship between selection and effect sizes.
We model stabilizing selection in a multidimensional phenotype space, akin to Fisher’s geometric model [72]. An individual’s phenotype is a vector in an n-dimensional Euclidian space, in which each dimension corresponds to a continuous quantitative trait. We focus on the architecture of one of these traits (say, the first dimension), where the total number of traits parameterizes pleiotropy. Fitness is assumed to decline with distance from the optimal phenotype positioned at the origin, thereby introducing stabilizing selection. Specifically, we assume that absolute fitness takes the form
W(r→)=exp(−r22w2),
(1)
where r→ is the (n-dimensional) phenotype, r=‖r→‖ is the distance from the origin, and w parameterizes the strength of stabilizing selection. However, we later show that the specific form of the fitness function does not matter. Moreover, the additive environmental contribution to the phenotype can be absorbed into w ([73]; Section 1.1 in S1 Text); we therefore consider only the genetic contribution.
The genetic contribution to the phenotype follows from the multidimensional additive model [74]. Specifically, we assume that the number of genomic sites affecting the phenotype (the target size) is very large, L ≫ 1, and that allelic effects on the phenotype at these sites are vectors in the n-dimensional trait space. An individual’s phenotype then follows from adding up the effects of her or his alleles, i.e.,
r→=∑l=1L(a→l+a→l′),
(2)
where a→l and a→l′ are the phenotypic effects of the parents’ alleles at site l.
The population dynamics follows from the standard model of a diploid, panmictic population of constant size N, with nonoverlapping generations. In each generation, parents are randomly chosen to reproduce with probabilities proportional to their fitness (i.e., Wright-Fisher sampling with viability selection), followed by mutation, free recombination (i.e., no linkage), and Mendelian segregation. We further assume that the mutation rate per site, u, and the population size are sufficiently small such that no more than 2 alleles segregate at any time at each site (i.e., that θ = 4Nu ≪ 1) and therefore an infinite sites approximation applies. The number of mutations per gamete per generation therefore follows a Poisson distribution with mean U = Lu; based on biological considerations (see Sections 4.1 and 4.2 in S1 Text), we also assume that 1 ≫ U ≫ 1/2N. The size of mutations in the n-dimensional trait space, a(=‖a→‖), is drawn from some distribution, assuming only that a2 ≪ w2. We later show that this requirement is equivalent to the standard assumption about selection coefficients satisfying s ≪ 1 (also see Section 4.3 in S1 Text). The directions of mutations are assumed to be isotropic, i.e., uniformly distributed on the hypersphere in n-dimensions defined by their size, although we later show that our results are robust to relaxing this assumption as well.
In the first 3 sections, we develop the tools that we later use to study genetic architecture. We start by considering the equilibrium distribution of phenotypes in the population and generalizing previous results for the case with a single trait [26,66,67,70]. Under biologically sensible conditions, this distribution is well approximated by a tight multivariate normal centered at the optimum. Namely, the distribution of n-dimensional phenotypes, r→, in the population, is well approximated by the probability density function:
f(r→)=1(2πσ2)n/2exp(−r22σ2),
(3)
where σ2 is the genetic variance of the phenotypic distances from the optimum (see Eq A25 in S1 Text for closed form); and under plausible assumptions about the rate and size of mutations (i.e., when 1 ≫ U ≫ 1/2N and a2 ≪ w2), it satisfies σ2 ≪ w2, implying small variance in fitness in the population (Section 4.2 in S1 Text). Intuitively, the phenotypic distribution is normal because it derives from additive and (approximately) independently and identically distributed contributions from many segregating sites. Moreover, the population mean remains extremely close to the optimum because stabilizing selection becomes increasingly stronger with the displacement from it and because any displacement is rapidly offset by minor changes to allele frequencies at many segregating sites.
With phenotypes close to the optimum, only the curvature of the fitness function at the optimum (i.e., the multidimensional second derivative) affects the selection acting on individuals. In addition, it is always possible to choose an orthonormal coordinate system centered at the optimum, in which the trait under consideration varies along the first coordinate and a unit change in other traits (along other coordinates) near the optimum has the same effect on fitness. These considerations suggest that the equilibrium behavior is insensitive to our choice of fitness function around the optimum. Moreover, in S1 Text (Section 5), we show that the rapid offset of perturbations of the population mean from the optimum (by minor changes to allele frequencies at numerous sites) lends robustness to the equilibrium dynamics with respect to the presence of major loci, moderate changes in the optimal phenotype over time, and moderate asymmetries in the mutational distribution.
Next, we consider the dynamic at a segregating site and generalize previous results for the case with a single trait [68–70]. This dynamic can be described in terms of the first 2 moments of change in allele frequency in a single generation (see, e.g., [75]). To calculate these moments for an allele with phenotypic effect a→ and frequency q (=1-p), we note that the phenotypic distribution can be well approximated as a sum of the expected contribution of the allele to the phenotype, 2qa→, and the distribution of contributions to the phenotype from all other sites, R→. From Eq 3, it then follows that the distribution of background contributions is well approximated by probability density:
f(R→|a→,q)=1(2πσ2)n/2exp(−(R→+2qa→)22σ2).
(4)
By averaging the fitness of the 3 genotypes at the focal site over the distribution of genetic backgrounds, we find that the first moment is well approximated by
E(Δq)≈a2w2 pq(q−12 ),
(5)
assuming that a2 and σ2 ≪ w2 (Section 4 in S1 Text). By the same token, we find that
V(Δq)≈pq2N,
(6)
which is the standard second moment with genetic drift.
The functional form of the first moment is equivalent to that of the standard viability selection model with underdominance. This result is a hallmark of stabilizing selection on (additive) quantitative traits: with the population mean at the optimum, the dynamics at different sites are decoupled, and selection at a given site acts to reduce its contribution to the phenotypic variance (2a2pq), thereby pushing rare alleles to loss. Comparison with the standard viability selection model shows that the selection coefficient in our model is s = a2/w2, or S = 2Ns = 2Na2/w2 in scaled units. In other words, the selection acting on an allele is proportional to its size squared in the n-dimensional trait space (where w translates effect size into units of fitness).
The statistical relationship between the strength of selection acting on mutations and their effect on a given trait follows from the aforementioned geometric interpretation of selection. Specifically, all mutations with a given selection coefficient, s, lie on a hypersphere in n-dimensions with radius a=ws, and any given mutation satisfies
s=1w2a2=1w2∑i=1nai2,
(7)
where ai is the allele’s effect on the i-th trait (Fig 1A). Our assumption that mutation is isotropic then implies that the probability density of mutations on the hypersphere is uniform.
The distribution of effect sizes on a focal trait, a1, corresponding to a given selection coefficient, s, follows. Given that mutation is symmetric in any given trait, E(a1|s) = 0, and given that it is symmetric among traits,
E(a12|s)=a2 /n=(w2/n )s.
(8)
More generally, the probability density corresponding to an effect size a1 is proportional to the volume of the (n − 2)–dimensional cross section of the hypersphere with projection a1 (Fig 1A). For a single trait, this implies that a1 = ±a with probability ½, and for n > 1, it implies the probability density
φn(a1|a)=Γ(n/2)/Γ((n−1)/2)n/212π(a2/n)(1−1na12(a2/n))n−32
(9)
(Section 1.2 in S1 Text). Intriguingly, when the number of traits n increases, this density approaches a normal distribution, i.e.,
a1a2/n~N(0,1),
(10)
implying that the distribution of effect sizes given the selection coefficient becomes
a1~N(0,(w2/n)s).
(11)
This limit is already well approximated for a moderate number of traits (e.g., n = 10; Fig 1B).
The limit behavior also holds when we relax the assumption of isotropic mutation. This generalization is important because, having chosen a parameterization of traits in which the fitness function near the optimum is isotropic, we can no longer assume that the distribution of mutations is also isotropic [76]. Specifically, mutations might tend to have larger effects on some traits than on others, and their effects on different traits might be correlated. In Section 5.4 in S1 Text, we show that the limit distribution (Eq 11) also holds for anisotropic mutation (excluding pathological cases). To this end, we introduce the concept of an effective number of traits, ne, which can take any real value ≥1 and is defined as the number of equivalent traits required to generate the same relationship between the strength of selection on mutations and their expected effects on the trait under consideration (i.e., replacing n in Eq 11). The robustness of our model, along with mounting evidence that genetic variation is highly pleiotropic (see “Introduction”), suggests that the limit form may apply quite generally. In that regard, we note that even in this limit, the strength of selection on mutations and their effects on the focal trait are correlated, implying that the kind of “purely pleiotropic” extreme postulated in previous works cannot arise [18–20].
We can now derive closed forms for summary statistics of the genetic architecture (see Section 2.3 in S1 Text). For mutations with a given selection coefficient, the frequency distribution follows from the diffusion approximation based on the first 2 moments of change in allele frequency (Eqs 5 and 6; [75]), and the distribution of effect sizes follows from the geometric considerations of the previous section. Conditional on the selection coefficient, these distributions are independent, and therefore, the joint distribution of frequency and effect size equals their product. Summaries of architecture can be expressed as expectations over the joint distribution of frequencies and effect sizes for a given selection coefficient and then weighted according to the distribution of selection coefficients. While we know little about the distribution of selection coefficients of mutations affecting quantitative traits, we can draw general conclusions from examining how summaries of architecture depend on the strength of selection.
We focus on the distribution of additive genetic variances among sites, a central feature of architecture that is key to connecting our model with GWAS results. We start by considering how selection affects the expected contribution of a site to additive genetic variance in a focal trait. We include monomorphic sites in the expectation, such that the expected total variance is given by the product of the expectation per-site and the population mutation rate, 2NU. Under the infinite sites assumption, sites are monomorphic or biallelic, and their expected contribution to variance is
E(2a12pq|S)=E(a12|S)E(2pq|S)=w22NnSE(2pq|S)
(12)
(expressed in terms of the scaled selection coefficient S). Thus, the degree of pleiotropy only affects the expectation through a multiplicative constant.
This multiplicative factor would have a discernable effect in generalizations of our model in which the degree of pleiotropy varies among sites. For example, if the degree of pleiotropy of one set of sites was k and of another set was l > k, and both sets were subject to the same strength of selection, then the expected contribution to genetic variance of sites in the first set would be l/k times greater than in the second (from Eq 12). While such generalizations may prove interesting in the future, here we focus on the model in which the degree of pleiotropy is constant. In this case, the multiplicative factor introduced by pleiotropy is not identifiable from data, because even if we could measure genetic variance in units of fitness (e.g., rather than in units of the total phenotypic variance), we still would not be able to distinguish between the effects of w and n on the genetic variance per site. We therefore focus on the effect of selection on the relative contribution to variance, which is insensitive to the degree of pleiotropy in our model.
The effect of selection on the relative contribution to genetic variance was described by Keightley and Hill (in the one-dimensional case [68]) and is depicted in Fig 2A. When selection is strong (roughly corresponding to S > 30), its effect on allele frequency (which scales with 1/S) is canceled out by its relationship with the effect size (Eq 8), yielding a constant contribution to genetic variance per site, vS = 2w2/nN, regardless of the selection coefficient (Section 3.1 in S1 Text; Fig 2A and Fig A1b in S1 Text). Henceforth, we measure genetic variance in units of vS. When selection is effectively neutral (roughly corresponding to S < 1) and thus too weak to affect allele frequency, the expected contribution of a site to genetic variance scales with the effect size and equals ½S (·vs) and therefore is lower than under strong selection (Section 3.1 in S1 Text; Fig 2A and Fig A1a in S1 Text). In between these selection regimes, selection effects on allele frequency are more complex and are influenced by underdominance (Section 3.1 in S1 Text). As the selection coefficient increases, the expected contribution to variance reaches vS at S ≈ 3 and continues to increase until it reaches a maximal contribution that is approximately 30% greater at S ≈ 10 (Fig 2A), after which it slowly declines to the asymptotic value of vS (Fig 2A and Fig A1b in S1 Text). Henceforth, we refer to this selection regime as intermediate (not to be confused with the nearly neutral range, which is much narrower and does not include selection coefficients with S > 10). These results suggest that effectively neutral sites should contribute much less to genetic variance than intermediate and strongly selected ones [67,68].
While intermediate and strongly selected sites contribute similarly to variance, their minor allele frequencies (MAFs) can differ markedly (Fig 2B). As an illustration, segregating sites with MAF > 0.1 account for approximately 72% and approximately 49% of the additive genetic variance for intermediate selection coefficients of S = 3 and 10, respectively, when almost no segregating sites would be found at such high MAF for a strong selection coefficient of S = 100 (Fig 2B). Thus, within the wide range of selection coefficients characterized as intermediate and strong, genetic variance arises from sites segregating at a wide range of MAFs ranging from common to exceedingly rare.
Next, we consider how genetic variance is distributed among sites with a given selection coefficient. We focus on the distribution among segregating sites (including monomorphic effects would just add a point mass at 0). This distribution is especially relevant to interpreting the results of GWASs, because, to a first approximation, a study will detect only sites with contributions to variance exceeding a certain threshold, v(=2a12pq), which decreases as the study size increases (see “Discussion”). We therefore depict the distribution in terms of the proportion of genetic variance, G(v), arising from sites whose contribution to genetic variance exceeds a threshold v.
We begin with the case without pleiotropy (n = 1), in which selection on an allele determines its effect size (Fig 3A). When selection is strong (S > 30), the proportion of genetic variance exceeding a threshold v is also insensitive to the selection coefficient and takes a simple form, with
G(v)=exp(−2v)
(13)
(Fig 3A; Section 3.2 in S1 Text). In contrast, in the effectively neutral range (S < 1),
G(v)=1−v/vmax,
(14)
where the dependency on the selection coefficient enters through vmax=18S, which is the maximal contribution to variance and corresponds to an allele frequency of ½ (Fig A4a; Section 3.2 in S1 Text). In the intermediate selection regime, G(v) is also intermediate and takes a more elaborate functional form (Section 3.2 in S1 Text). These results suggest how genetic variance would be distributed among sites given any distribution of selection coefficients (Fig 3A): starting from sites that contribute the most, the distribution would at first be dominated by strongly selected sites, and then the intermediate selected sites would begin to contribute, whereas effectively neutral sites would enter only for v<18S≪1.
Pleiotropy causes sites with a given selection coefficient to have a distribution of effect sizes on the focal trait, thereby increasing the contribution to genetic variance of some sites and decreasing it for others. In Section 3.2 of S1 Text, we show that increasing the degree of pleiotropy, n, increases the proportion of genetic variance, G(v), for any threshold, v, regardless of the distribution of selection coefficients (Fig A5 in S1 Text). When variation in a trait is sufficiently pleiotropic for the distribution of effect sizes to attain the limit form (Eq 11)
G(v)=(1+2v)exp(−2v)
(15)
for strongly selected sites and
G(v)=exp(−4v/S)
(16)
for effectively neutral ones (Fig 3B and Fig A4b in S1 Text; Section 3.2 in S1 Text). The intermediate selection range is split between these behaviors: on the weaker end, roughly corresponding to S < 5, G(v) is similar to the effectively neutral case (Fig A4b and Section 3.2 in S1 Text); and on the stronger end, roughly corresponding to S > 5, G(v) is similar to the case of strong selection, with measurable differences only when v ≫ vs (inset in Fig 3B and Section 3.2 in S1 Text). We would therefore expect that as the sample size of a GWAS increases and the threshold contribution to variance decreases, intermediate and strongly selected sites (more precisely, sites with S > 5) will be discovered first, and effectively neutral sites will be discovered much later. In S1 Text (Section 3.2 and Fig A3 in S1 Text), we also derive corollaries for the distribution of numbers of segregating sites that make a given contribution to genetic variance.
In humans, GWASs for many traits display a similar behavior: when sample sizes are small, the studies discover almost nothing, but once they exceed a threshold sample size, both the number of associations discovered and the heritability explained begin to increase rapidly [4,77]. Intriguingly though, both the threshold study size and rate of increase vary among traits. These observations raise several questions: How is the threshold study size determined? How should the number of associations and explained heritability increase with study size once this threshold is exceeded? With an order of magnitude increase in study sizes into the millions imminent, how much more of the genetic variance in traits should we expect to explain? The theory that we developed provides tentative answers to these questions.
To relate the theory to GWASs, we must first account for the power to detect loci that contribute to quantitative genetic variation. In studies of continuous traits, the power can be approximated by a step function, where loci that contribute more than a threshold value v* to additive genetic variance will be detected and those that contribute less will not (Section 6.1 in S1 Text). The threshold depends on the study size, m, and on the total phenotypic variance in the trait, VP, where v* ∝ VP/m (Section 6.1 in S1 Text; [77]); conversely, the study size m needed to detect loci with contributions above v* is proportional to VP/v*. Given a trait and study size, the number of associations discovered and heritability explained then follow from our predictions for the distribution of variances among sites.
When genetic variation in a trait is sufficiently pleiotropic, our results suggest that the first loci to be discovered in GWASs will be intermediate or strongly selected, with correspondingly large effect sizes (i.e., S≈2Nnw2a12>5). The functional form of the distribution of variances among these loci (Eq 15 and Fig 3B) implies that for GWASs to capture a substantial proportion of the genetic variance, their threshold variance for detection v* has to be on the order of the expected variance contributed by strongly selected sites, vs, or smaller. We therefore expect the threshold study size for the discovery of intermediate and strongly selected loci to be proportional to VP/vs. When the study size exceeds this threshold, the number of associations detected and proportion of variance explained depend on the study size measured in units of VP/vs (Fig 4) and follow from the functional forms that we derived (Eq 15 and Table A1 in S1 Text). The dependence on VP/vs makes intuitive sense, as the total phenotypic variance VP is background noise for the discovery of individual loci whose contributions to variance are on the order of vs. Some results are modified when variation in a trait is only weakly pleiotropic, which is probably less common: notably, the threshold study size for strongly selected loci would be higher, and loci under intermediate selection would begin to be discovered only after the strongly selected ones (Fig A22 in S1 Text, Eq 13, and Eq A35 in S1 Text). Regardless of the degree of pleiotropy, effectively neutral loci would only begin to be discovered at much larger study sizes, after the bulk of intermediate and strongly selected variance has been mapped (Fig 4 and Fig A22 in S1 Text). Thus, the dependence of the explained heritability on study size is largely determined by VP/vs and by the proportion of heritable variance arising from intermediate and strongly selected loci, whereas the number of associations also depends on the mutational target size, providing a tentative explanation for why the performance of GWASs varies among traits.
Importantly, these theoretical predictions can be tested. As an illustration, we consider height and body mass index (BMI) in Europeans, 2 traits for which GWASs have discovered a sufficiently large number of genome-wide significant (GWS) associations (697 for height [16] and 97 for BMI [15]) for our test to be well powered. We fit our theoretical predictions to the distributions of variances among GWS associations reported for each of these traits, assuming that these distributions faithfully reflect what they would look like for the true causal loci (see Section 6.3 in S1 Text). We further assume that these loci are under intermediate or strong selection (as our predictions suggest) and that they are highly pleiotropic (see "Introduction"; [37, 42, 45]). Under these assumptions, we expect the distribution of variances to be well approximated by a simple form (Eq A89 in S1 Text), which depends on a single parameter, vs. We find that the theoretical distribution with the estimated vs fits the data for both traits well (Fig 5A): we cannot reject our model based on the data for either trait (by a Kolmogorov-Smirnov test, p = 0.14 for height and p = 0.54 for BMI; Section 7.5 in S1 Text). By comparison, without pleiotropy (n = 1), our predictions provide a poor fit to these data (by a Kolmogorov-Smirnov test, p < 10−5 for height and p = 0.05 for BMI; Fig A14 in S1 Text).
Fitting the model to GWAS results further allows us to make inferences about evolutionary parameters (Sections 7.1 and 7.3 in S1 Text). By including the degree of pleiotropy (n) as an additional parameter, we find that for both height and BMI, n is sufficiently large for it to be indistinguishable from the high pleiotropy limit. Based on the shape of the distributions in this limit and on scaling the threshold values of v* in units of our estimates for vs, we estimate that the proportion of variance arising from mutations within the range of detectable selection effects is approximately 50% for height and approximately 15% for BMI. Further relying on the number of associations that fall above the thresholds, we infer that, within this range, height has a mutational target size of approximately 5 Mb, whereas BMI has a target size of approximately 1 Mb (Table A2 in S1 Text).
These parameter estimates can help to interpret GWAS results. They suggest that, despite their comparable sample sizes, the GWAS for height succeeded in mapping a substantially greater proportion of the heritable variance than the GWAS for BMI (approximately 20% compared to approximately 3%–5%) primarily because the proportion of variance arising from mutations within the range of detectable selection effects for height is much greater than for BMI. Moreover, the estimates of target sizes and the relationship between sample size and threshold contribution to variance can be used to predict how the explained heritability and number of associations should increase with sample size (Fig 5B and 5C). These predictions are likely underestimates as the range of detectable selection effects itself should also increase with sample size.
We can also examine to what extent our inferences are consistent with data and estimates from earlier studies. For example, the distribution of variances that we inferred for height fits those obtained in a recent GWAS of height based on exome genotyping (Kolmogorov-Smirnov test, p = 0.99; Fig A15b and Section 8.1 in S1 Text). In addition, the proportion of variance that we estimate to arise from the range of selection effects detectable in existing GWASs for height and BMI is consistent with estimates of the heritable variance tagged by all single-nucleotide polymorphisms (SNPs) with MAF > 1% [60, 61]; Section 8.2 in S1 Text.
While we have assumed that quantitative traits have been subject to long-term stabilizing selection, recent studies indicate that some traits, and height in particular, have also been subject to recent directional selection [78–82]. Under plausible evolutionary scenarios, recent directional selection can induce large changes to the mean phenotype through the collective response at many segregating loci while having a negligible effect on allele frequencies at individual loci [21,83]. This very subtle effect on allele frequencies is likely one reason why polygenic adaptation is so difficult to detect and why studies have to pool faint signals across many loci to do so [78–82]. In Section 5.1 of S1 Text, we show that the distribution of allele frequencies on which our results rely is insensitive to sizable recent changes to the optimal phenotype. Importantly then, even when recent directional selection has occurred and its effects are discernable, the genetic architecture of a trait is nonetheless likely to be dominated by the effects of longer-term stabilizing selection.
In contrast, recent changes in the effective population size are likely to have had a dramatic effect on allele frequencies and thus on the genetic architecture of quantitative traits [84,85]. In particular, European populations in which the GWASs for height and BMI were performed are known to have experienced dramatic changes in population size, including an Out-of-Africa (OoA) bottleneck about 100 KYA and explosive growth over the past 5 KY [86–89]. To study how these changes would have affected genetic architecture, we simulated allelic trajectories under our model and historical changes in population sizes in Europeans (relying on the model of [89]; Section 9 in S1 Text).
Our results suggest that the individual segregating sites with the greatest contribution to the extant genetic variance have selection coefficients around s = 10−3 and are due to mutations that originated shortly before or during the OoA bottleneck (Fig 6A and Section 9 in S1 Text). These mutations ascended to relatively high frequencies during the bottleneck and minimally decreased in frequency during subsequent, recent increases in population size, thereby resulting in large contributions to current genetic variance. Segregating sites under weaker selection contribute much less to variance because of their smaller effect sizes (i.e., for the same reason that applied in the case with a constant population size). Finally, and in contrast to the case with a constant population size, individual segregating sites under stronger selection (e.g., s ≥ 10−2.5) contribute much less to current variance than those with s ≈ 10−3. Mutations at these sites are younger and arose after the bottleneck, when the population size was considerably larger, resulting in much lower initial and current frequencies and therefore a lower per (segregating) site contribution to variance (as distinct from the proportion of strongly selected sites that are currently segregating, which will have greatly increased, resulting in the same total contribution to variance; [84, 85]). In Section 10 in S1 Text, we discuss one implication of these demographic effects: that the reliance on genotyping rather than resequencing in GWASs had a minimal effect on the discovery of associations.
Segregating loci with s ≈ 10−3 not only make the largest contributions to the current variance but also are likely to account for most of the GWS associations in the GWASs of height and BMI (Section 9 in S1 Text). When we account for the discovery thresholds of these studies, the expected distribution of variances for loci with s ≈ 10−3 closely matches the distribution observed among GWS associations (Fig 6B and Fig A20b in S1 Text). Moreover, these distributions closely match our theoretical predictions for s ≈ 10−3 and an Ne ≈ 5,000 (Fig 6B)—roughly the effective population size experienced by mutations that originated shortly before or during the bottleneck. This match likely explains why the results predicted on a constant population size fit the data well nonetheless. Our interpretation of GWAS findings is supported by other aspects of the data (Section 9 in S1 Text).
Our conclusions about the high degree of pleiotropy of genetic variation for height and BMI and the differences between these traits are likely robust to demographic effects, given how well our model fits the distributions of variances among loci, once we account for European demographic history. However, we might be underestimating the mutational target sizes and total heritable variances associated with the selection effects currently visible in GWASs, as simulations with European demographic history indicate that the proportion of variance arising from loci with s ≈ 10−3 explained by current GWASs is lower than our equilibrium estimates (approximately 42% compared to approximately 53% for height and approximately 29% compared to approximately 38% for BMI). By the same token, we likely underestimated the future increases in explained heritability with increases in study sizes (Fig 5B and 5C).
In summary, a ground-up model of stabilizing selection and pleiotropy can go a long way toward explaining the findings emerging from GWASs. Important next steps involve explicitly using more information from GWASs in the inferences. In particular, we can learn more about the selection acting on quantitative genetic variation by explicitly incorporating information about frequency and effect size (rather than their combination in terms of variance) and by including information from associations that do not attain genome-wide significance. Doing so will further require directly incorporating the effects of recent demographic history on genetic architecture [84,85]. An extended version of the inference, applied to the myriad traits now subject to GWASs, should allow us to learn about differences in the genetic architectures of traits and answer long-standing questions about the evolutionary forces that shape quantitative genetic variation.
|
10.1371/journal.ppat.1004026 | The Role of Host and Microbial Factors in the Pathogenesis of Pneumococcal Bacteraemia Arising from a Single Bacterial Cell Bottleneck | The pathogenesis of bacteraemia after challenge with one million pneumococci of three isogenic variants was investigated. Sequential analyses of blood samples indicated that most episodes of bacteraemia were monoclonal events providing compelling evidence for a single bacterial cell bottleneck at the origin of invasive disease. With respect to host determinants, results identified novel properties of splenic macrophages and a role for neutrophils in early clearance of pneumococci. Concerning microbial factors, whole genome sequencing provided genetic evidence for the clonal origin of the bacteraemia and identified SNPs in distinct sub-units of F0/F1 ATPase in the majority of the ex vivo isolates. When compared to parental organisms of the inoculum, ex-vivo pneumococci with mutant alleles of the F0/F1 ATPase had acquired the capacity to grow at low pH at the cost of the capacity to grow at high pH. Although founded by a single cell, the genotypes of pneumococci in septicaemic mice indicate strong selective pressure for fitness, emphasising the within-host complexity of the pathogenesis of invasive disease.
| Decades of research on bacterial sepsis have been devoted to analysing the steps that lead from a local event, either carriage or a localised infection, to systemic disease. Our work analyses in depth the events determining systemic infection by one of the main human pathogens, Streptococcus pneumoniae. Consistent with similar findings on the pathogenesis of bacteraemia due to other commensal pathogens, our results show that after an intravenous inoculum of a million pneumococci, the resulting septicaemia is often founded by a single bacterial cell. Investigation into the nature of this monoclonal infection identified strong within-host selective pressure for metabolic fitness during outgrowth of the bacterial population.
| Streptococcus pneumoniae, one of the major human bacterial pathogens, is also part of the normal upper respiratory tract flora, where nasopharyngeal colonisation with one or more strains often lasts weeks to months with seasonal peaks in late winter [1], [2]. Carriage of S. pneumoniae (pneumococci) may result in disease as the consequence of contiguous spread from the nasopharynx to other sites in the upper or lower respiratory tract causing, for example, otitis media or pneumonia. More rarely, there is hematogenous dissemination of pneumococci resulting in septicaemia and metastatic disease such as meningitis [1], [3]–[5]. In experimental models of pneumococcal infection, the challenge dose required to induce disease is dependent on the route of infection, the genetic background of the host and the virulence of the infecting strain [6] and may vary from a very few to millions of organisms [7]. Following intravenous inoculation of mice with laboratory grown pneumococci, a hallmark of experimental bacteraemic infections is the rapid and efficient clearance of most of the inoculated bacteria [8]–[10]. In non-immune rodents, major factors mediating this clearance are splenic macrophages and complement mediated opsonisation [11]–[14]. A challenge dose of about one million virulent, encapsulated pneumococci is generally needed to induce bacteraemia in about half of challenged animals (the effective dose or ED50) and which is the dose generally used to address investigations into the early events shaping an infectious process.
Most work on the pathogenesis of infectious disease focuses on specific virulence determinants which are generally presented as the cause, either alone or in combination with other factors, of the events leading to the infection of the host where the microbial population is considered to be a uniform entity. However, several investigations have addressed the within host population dynamics, especially on the early phases of host-pathogen interactions [15]. There are different models which address these early events that include: (i) the model of independent action, which postulates that at the LD50 (lethal dose for 50% for the hosts) the hosts develop infection “following the multiplication of only one of the inoculated bacteria, regardless of the total number of bacteria inoculated” [16], (ii) the hypothesis of synergy which “postulates that inoculated bacteria co-operate and that fatal infections will be initiated by more than one bacterium and that this will lead to the predominance of several variants” [16], and models which introduce time as a factor into the process and propose a two-stage model where a birth–death phase would be responsible for generation of the heterogeneity within the population later during the infection [17].
Both for viral and bacterial infections it has been shown that the effective number of infectious agents which actually start the disease is generally many orders of magnitude below the actual dose used for challenge [16], [16], [18]–[21]. In particular a series of reports, generally based on experimental challenge using an inoculum containing an approximately equal mixture of two isogenic variants at the LD50, has shown experimentally that systemic infections may be initiated by the multiplication of as few as a single organism [22]–[27]. Models differ widely, and due to the nature of pathogenesis of many infections, they rely on experimental challenge at a site different to that investigated for disease implying that multiple bottlenecks occur and that potentially a series of invasive events could be enucleated [16], [18]–[27]. We here have investigated the host and microbial determinants that underpin the occurrence of the single cell bottleneck in the pathogenesis of pneumococcal septicaemia following inoculation of mice with three isogenic variants by the intravenous route. This route has an advantage over other experimental infection models as there are less biological events between the initial challenge and full blown disease so that rigorous analysis of the events is facilitated.
CD1 mice were inoculated intravenously (i.v.) with a mixture of pneumococci comprising approximately equal numbers of each of three isogenic TIGR4 mutants (FP122, FP321 and FP318) with different resistance markers (Table 1) [28]. Following inoculation, blood samples were collected at different times and spread on selective plates. Colony counts allowed quantitation of the distribution of the different mutants making up the pneumococcal population in the blood (Figure 1). Two hours after bacterial challenge, blood samples from all mice, with one exception, grew all three of the variants that had been included in the challenge dose. Samples of the second group of mice, sampled one hour thereafter, showed a mixed population of the variants: two in 6 mice, three in 17 mice and one mouse had negative blood culture (Figure 1A). At 7 h after challenge, this pattern was distinctly different: there were 10 positive and 2 negative blood cultures and the numbers of bacteria were significantly reduced. At 8–9 h post-infection, most blood cultures (25/29 mice) were negative. The remaining 4 mice had monoclonal blood cultures in that each grew colonies of only one mutant (Figure 1A). In the subsequent hours of infection, bacteria were detected in the blood at high concentrations (up to 1×106 CFU) in 17/55 (31%) mice at 24 h, 16/43 (37%) mice at 48 h and 15/24 (62%) mice at 72 h (Figure 1B). At all these time points, most blood samples yielded colonies of only one of the variants: 12 out of 17 blood cultures at 24 h, 12/16 at 48 h and 8/16 at 72 h. Among the 32 single-variant blood samples each variant was more or less equally represented as the progenitor, although owing to the smaller challenge dose of strain FP318 in one experiment, this strain was recovered at lower density from the blood and caused fewer single-variant infections (Table S1). In three of twelve mice for which three serial blood cultures were taken we observed an increase in the number of variants. One of these mice showed evidence of infection with 3 variants at 48 h after earlier having a monoclonal infection and a further two mice had three variants infection at 72 h after having previously had infection with only one or two variants respectively. Seven of the single-variant bacteraemia isolates were checked for colony morphology and each was found to have the opaque phenotype, in contrast to the challenge strains (not-mouse-passaged) that yielded a mixed population of about similar proportions of opaque and translucent colonies [29].
The bacterial counts from cultures of spleen tissue were assayed in twelve mice at each time point (Figure S1). All mice which had positive blood cultures showed bacterial counts also in splenic samples. In addition, small numbers of organisms were cultured from spleen tissue both at 24 and 48 h in two mice each which presented with sterile blood cultures. Similarly at 72 h, bacteria were only detected in the spleen in one mouse (Figure S1). These data show that infectious foci can be detected in the spleens of mice that have negative blood cultures, indicating that the spleen is the probable site where the infection originates.
Our data indicate that the near totality of bacteria in the challenge with three isogenic pneumococcal variants is cleared by the immune system (predominantly by splenic macrophages; see below) and that few bacterial cells remain viable within a defined site of the host (i.e. the spleen). This small number of bacteria may start to grow and re-invade the host giving rise to bacteraemia. In our experimental infection most of the mice challenged were not bacteraemic and, of those becoming bacteraemic, most were infected by a single bacterial variant. In addition, we detected in some mice an increase in variants within blood cultures over time. These data indicate that over time more than one invasion event may occur. Theoretically bacteraemia may be generated in two ways: (i) by a single bacterium establishing a population in the blood in a single invasion event or several bacteria each independently establishing a population in distinct invasion events (independent action); (ii) by a defined number (more than one) of bacteria acting together to invade once or several times (co-operative synergism) [22]. We hypothesise that the former explanation pretains.
To statistically evaluate the number of bacteria involved in founding the blood population in each invasion event, we construct a model that assumes that bacteria invade and establish a population in the blood at random (Supplementary text) [23]. In this model, the number of invasion events in each mouse is assumed to follow a Poisson distribution so the expected number can be estimated from the proportion of the non-bacteraemic mice. Then we determined the expected numbers of mice infected with one, two or three variants, assuming that the number of bacteria (w) responsible for establishing the blood population in each invasion event were 1, 2, 3, etc (Table 2). Given that some mice were culled during the course of the experiment and some got multiple samplings, we limited the statistical analysis to the observations at the 24 h time point. In table 2, we report the comparison between the expected versus the observed numbers of the infected mice with one, two and three variants for different numbers of bacteria (w) potentially responsible for founding a population blood. The statistical analysis shows that the most probable number of bacteria responsible for establishing a blood population is 1 (Table 2). Note that the p-value was calculated by combining data for blood cultures with two or three variants because of the small expected frequencies in these two categories. Given that w is equal to one, we conclude that polyclonal blood infections are the result of more than one invasion event, each event founded by a single bacterium, consistent with the observed time-dependent increase in the frequency of polyclonal bacteremia over the 72 h of the experiment.
Prior to the assessment of the impact of host factors in the control of bacteraemia, we compared bacterial counts in the blood of two mouse strains known to be resistant to pneumococcal infection (outbred CD1 mice and inbred BALB/c mice) and a susceptible mouse strain CBA/Ca [30]. The mice were inoculated i.v. with a mixture of three encapsulated pneumococcal strains of different serotypes, D39 (type 2), TIGR4 (type 4) and G54 (type 19F). Bacterial clearance in CD1 and BALB/c mice showed similar kinetics (Figure 2 A–B), while CBA/Ca mice were less able to reduce the initial number of bacteria (Figure 2 C). All mouse strains cleared G54 bacteria immediately (no positive blood culture 10 min after infection) and showed a first phase of rapid clearance also for both D39 and TIGR4. Only the two resistant mouse strains showed bacterial numbers in the blood that were less than the limit of detection. In contrast, bacterial numbers increased in the susceptible strain after the first phase (Figure 2 A–C). This indicates that, depending on which host-pathogen pair was investigated, the bottlenecks may vary considerably. Since BALB/c mice showed a more uniform clearance of bacteria, subsequent experiments were conducted with this mouse strain in order to keep experimental groups to a minimal size.
To identify the host immune cells responsible for the initial clearance of bacteria from the blood, we performed a set of experiments in BALB/c mice depleted either of macrophages or neutrophils. Macrophage depletion was achieved by intraperitoneal (i.p.) injection of clodronate liposomes and neutrophil depletion by using anti-GR-1 monoclonal antibody [31]–[33]. Control groups were treated either with PBS-containing liposomes as control for clodronate experiments or with isotype-matched antibody in the case of experiments with anti-GR-1. The results obtained for the control groups were comparable to the untreated control mice and differed from the groups of mice treated with clodronate (Figure S3 A–C) or anti-GR-1 (Figure S3 E–G). To verify macrophage or neutrophils depletion, spleen samples were analyzed by flow cytometry. The reduction of macrophages in the spleen of clodronate-treated mice was 61%±14.2 measure by anti-F4/80 and 47%±10.4 by anti-CD11b compared to naïve mice. Similar results were obtained when liver samples were analyzed (Figure S3 D). In anti-GR-1-treated mice, the neutrophil number was reduced by 83%±2.7 as compared to control mice (Figure S3 H). To check for anti-pneumococcal antibodies in naïve mice, we evaluated the reactivity of mouse serum towards whole pneumococcal cells. Mice had no detectable serum antibodies to any of the pneumococcal serotypes (Figure S3 I–K). A result supported by the observation that addition of type specific rabbit serum to the blood from naïve mice conferred specific bactericidal activity (P<0.01).
To analyse the role of macrophages and neutrophils, mice were divided into three groups: untreated (Figure 3 A1–A4), clodronate-treated (Figure 3 B1–B4), and anti-GR1-treated (Figure 3 C1–C4). After i.v. challenge with a mixture of four different strains (TIGR4, D39, DP1004 and G54), time course of bacterial counts was monitored by sampling blood, spleen, lung, liver and kidney (Figure 3 and S2). Analysis of control mice allowed categorisation of the pneumococcal strains into two groups: TIGR4 and D39, which were slowly cleared (virulent strains), and G54 and DP1004, which were cleared from the blood within minutes. The counts of TIGR4 and D39 were higher in the blood than in the spleen at 5 min and at 4 h compared to the other two strains (P<0.05). Bacterial loads in the other organs were similar to those found in the spleen (Figure S2). In contrast, mice infected with strains DP1004 and G54 showed higher CFU counts in the spleen than in the blood and other organs (P<0.05 at 5 min for both strains, P<0.01 at 4 and 8 h for DP1004) (Figure 3 A3–4).
The groups of mice depleted of macrophages showed significantly reduced ability to clear bacteria from the bloodstream. An increase in bacterial numbers in blood from 5 min to the later time points was observed in mice infected with TIGR4 (Figure 3 B1) and D39 (Figure 3 B2) (P<0.01). Blood bacterial counts were significantly higher in the clodronate-treated mice than in the control group (P<0.05 for all time points for both D39 and TIGR4). Bacterial counts of TIGR4 and D39 in liver and spleen were lower but with a similar trend, over time, to those in the blood (Figure 3 B1–B2 and Figure S2 B1–B2). In clodronate-treated mice, the numbers of non-virulent bacteria (strains G54 and DP1004) were higher in the blood than in the spleen (P<0.05 at 5 min for DP1004 and P<0.05 at 5 and 4 h for G54) and paralleled the trend observed for the virulent strains TIGR4 and D39 in untreated animals (Figure 3 B3–B4 compared to A1–A2).
In neutrophil-depleted mice, bacterial counts of both TIGR4 (Figure 3 C1) and D39 (Figure 3 C2) in blood and spleen decreased in the first 4 h after challenge with a similar trend to that observed in untreated animals. Thereafter, blood and organ counts remained stable (Figure 3 C1–C2 and Figure S2 C1–C2). For both virulent strains, the number of bacteria were higher in the blood than in the spleen (P<0.01 at 5 min). Interestingly, strain G54 had a peculiar behaviour in neutrophil-depleted mice, as it persisted in the spleen at high levels throughout the whole experiment despite being cleared from the blood within a few min of infection (P<0.001 at 4 and 8 h), (Figure 3 C4). At 4 h post-infection G54 bacteria reappeared in the blood (Figure 3 C4). The experiment was repeated for the later time points, and the pattern of counts was identical. The rough DP1004 strain was cleared from each body site as well as in untreated mice (Figure 3 C3).
To determine more precisely the early events occurring in the clearance of pneumococci, we have plotted separately the data on blood bacterial counts obtained 5 min and 30 min after challenge (Figure 2 D–E). In untreated mice, bacterial blood counts of the invasive strains D39 and TIGR4 were respectively 6.2×104 and 5.4×104 CFU/ml, while those of the non-invasive strains, DP1004 and G54, were three times lower (reduction of 60 to 75%). Differences between the virulent and non-virulent strains were statistically significant (P<0.01).
Given the key role of splenic macrophages in pneumococcal clearance, we evaluated the capacity of splenic macrophages to internalise pneumococci. Splenic BALB/c macrophages were grown as primary cell cultures, washed and re-cultured for seven days in M-CSF supplemented medium. Cytoflourimetric data showed expression of the characteristic markers of splenic macrophages, CD11b, CD11c, F4/80 and SIGLEC-1 (Figure 2 H) [34]. Adhesion was evaluated by counting pneumococci after 45 min and phagocytosed bacteria were enumerated by plating after a further 30 min of antibiotic treatment (viable intracellular bacteria). Despite similar values in adherent cells (Figure 2 F), our data show higher numbers of intracellular bacteria for the rough DP1004 and for the G54 strain and less for the virulent D39 and TIGR4 (Figure 2 G). Essentially identical data where obtained when performing the experiment with splenic macrophages from C57BL/6 mice, while in contrast bone marrow macrophages from BALB/c mice and RAW264.7 macrophages showed different patterns of surface markers expression to the spleen macrophages and their phagocytosis of pneumococci showed no correlation to the extent of early clearance in the host (Figure S4). These data emphasise the importance in the choice of cell lines for performing phagocytosis assays in vitro to assess pneumococcal clearance in vivo.
Pneumococci grown from the blood of 6 mice were subjected to whole genome sequencing (Table 3). In each case, the isolates had the identical antibiotic resistance phenotype and were therefore presumptively monoclonal. Two of the blood cultures were obtained from the same mouse, but at 24 or 48 h respectively, (mouse 3.1.5; Table 3). To identify possible mutations characterising the founding cell of the monoclonal blood culture, we searched for single nucleotide polymorphisms (SNP) present in all cells isolated from a given blood culture. Such a SNP would demonstrate that the re-expanded population arose from a single cell. We identified one or two SNPs in 100% of the bacterial populations from four out of six mice (3.1.5, 4.1.4, 4.2.2 and 4.2.6), when compared to bacteria from the challenge inoculum (Table 3). The identification of a SNP common to all bacteria of a given sample is conclusive genetic evidence that the pneumococcal populations were monoclonal. Further, the SNPs were either inter-genic, silent or in regions not predicted to be functional in pathogenesis. Thus, we conclude that these mutations were unlikely to be associated with changes in within-host fitness. This argues strongly that bacteremia was founded as the result of a stochastic process rather than the selection of fitter variants. However, further analysis identified a second set of SNPs in pneumococci of 5 of 6 blood cultures. Crucially, this second set of SNPs were only found in a proportion of the bacteria obtained from mouse blood and therefore must have occurred after the bottleneck. Further, these SNPs differed between isolates of different mice, but all were located within distinct sub-units of the pneumococcal F1/F0 ATPase operon (Table 3). In three bloods, more than one SNP was detected. To determine if more than one SNP in the ATPase operon occurred in a single cell, we sequenced single colony isolates of these populations. In all cases, where a multi-SNP profile would have been possible according to the genomic data, only clones with a single SNP within the F1/F0 ATPase operon were recovered. These isolates included FP490, a 3.2.4 derivative with a SNP in atpA, FP487, a 3.1.5 derivative with a SNP in atpC, and FP489 and FP498, two 4.1.6 derivatives with different SNPs in atpD (Table 1 and 3). In few cases subpopulations with mutations in other genes were detected (pilus sortase, potassium uptake protein, metE, and SP0760) (Table 3), but no confirmation by direct sequencing was performed for these genomic data and we do not think that these mutations are of major biological relevance.
Phenotypic analysis of eight independent ex vivo blood isolates each having a mutation (SNP) in the ATPase (Table 1 and 3), showed normal colony morphology on agar plates and no significant change in their susceptibility to optochin. In liquid culture, the mutants showed normal or more efficient growth in Todd Hewitt Yeast Extract (Figure 4 A), but were unable to grow in other media (Tryptic Soy Broth) (Figure 4 B). Given that the F1/F0 ATPase is involved in multiple aspects of proton trafficking, we investigated the impact of pH, buffer composition and salt concentration on bacterial growth. Using a phenotype microarray for osmotic susceptibility using Biolog microtiter plates PM9, we compared the phenotype of parental strains derived from strain TIGR4 to the ex vivo mutants. The mutants had acquired a series of metabolic characteristics, also shared by strain D39 (Figure S5). Growth experiments performed in serial buffer and salt dilutions showed that TIGR4 and its isogenic derivatives used in the challenge experiments had a restricted pH optimum when compared to D39, which limited growth at potassium phosphate concentrations below 10 mM and pH below 6.8 (Figure 4 D). Interestingly many of the mutants had gained this capacity, making them equally able to grow at low pH as D39. In contrast, high buffer concentrations (80 mM K2HPO4 and pH 8), inhibited growth of all mutants (Figure 4 C).
To investigate effects on intracellular pH homeostasis of the ATPase mutations we transformed the frame-shift in the atpC gene into the non-encapsulated strain DP1004. Using in vivo NMR, the atpC mutant and its parental strain were both shown to have an identical intracellular pH of 6.52 to 6.56 during active metabolism of glucose. No differences in susceptibility to neutrophil killing were observed when mutants were assayed in an opsonophagocytosis assay in the presence of type specific antibodies (Figure 4 E). Also, data of macrophage phagocytosis were unaltered in primary cultures of splenic macrophages (Figure 4 F). To check for any fitness cost in vivo, the encapsulated atpC mutant was compared to the challenge strain FP321 in our i.v. mouse sepsis model. At early time points both strains showed comparable blood counts (data not shown). Also at 72 h post-challenge, bacterial counts in blood were similar, but bacterial spleen counts for the atpC mutant were significantly increased when compared to the wild-type (Figure 4 G).
We have investigated the pathogenesis of pneumococcal bacteraemia following intravenous inoculation of mice with three isogenic clones (variants). In our model, the infection followed the classic, three phase pattern in which a majority of pneumococci are cleared in the first minutes post-challenge. This leads to an “eclipse phase” of several hours in which bacterial numbers decline further or are undetectable. This is followed by the emergence of sustained and high density bacteraemia in a proportion of the challenged animals [8], [10]. By analysing the survival in the blood of three isogenic variants of S. pneumoniae, we observed that the majority of blood cultures arose from only one of the three variants. We used a statistical model to characterise the infection dynamics in which the number of bacteria starting the infection in each invasion event is w and the number of times this happens is k [23]. From the model, we could infer that the number of bacteria at the origin of infection is below 2 (w = 1). Thus, it follows that bacteraemia was generated by either (a) a single bacterium establishing a population in the blood in a single invasion event or (b) several bacteria each of which independently established a population in distinct invasion events. The probability of (b) is small (about 5% in our data, because the probability of two or more invasion events occurring is about 5%). Genome sequencing provided genetic evidence in 4/6 cases that monoclonal bacteraemia did actually start from a single bacterial cell (w = 1) confirming the first statement. For the remaining 2/6 cases we could not determine w = 1 by genome sequencing as we could not distinguish several invasions of a single bacterial variant from one invasion of several cells of the same variant without any SNPs. More complex is the experimental observation of invasive events. For this we could document bacteraemia in mice with previous negative blood samples (k≥1) and in other mice the increase of variants in serial blood samples (k>1). Since after the first 24 h the observed numbers of both these types of invasion events are similar, this strongly favours the occurrence of polyclonal infections resulting from independent, not cooperative action. In the case of H. influenzae it had been hypothesised that the single cells giving rise to the monoclonal infection might be selected by within-host evolution [23]. Our work now tests this hypothesis by whole genome sequencing. The data show in two cases absence of any SNPs and in four cases SNPs that apparently do not indicate selection for virulence. Despite the low numbers, it suggests that the single cells at the origin of infection apparently have no advantage (higher virulence) over the other cells in the population. Such results show that the bacteria in the challenge dose act independently to give rise to infection, that each has a similar probability of causing infection and that a dose near the LD50, a single cell may initiate disease. These criteria satisfy the theory of independent action [16], [17], [22], [23]. As such our investigation provides strong evidence that the single founding cell of an invasive infection is the result of a stochastic event. However, it must be emphasised that epigenetic variations would have eluded our genetic and genomic analysis.
Previously published studies have shown a major role for splenic macrophages in the initial clearance of pneumococci. In the seminal investigations of Brown et al. [35], they conclude: “… it appears that an anatomically normal spleen plays a unique role in the clearance of experimental pneumococcal bacteraemia, and that this role is of increasing importance as the pathogenicity of the invading organism increases”. Our data provide evidence that splenic macrophages have properties not found in those derived from other tissue sites, with respect to their efficiency to ingest and kill pneumococci. It is worth noting that the impressive efficiency of splenic clearance in vivo in the non-immune host is somewhat at odds with the relatively inefficient ingestion and killing of pneumococci in standard in vitro phagocytosis assays [34]. The innate host factors that result in the removal of the vast majority of bacteria within 45 minutes of challenge [11], [35], [36] deserve further attention.
Despite the efficiency of splenic macrophages in clearance, sustained bacteraemia eventually occurs after an eclipse phase of several hours during which bacteria are largely undetectable in blood. Similar data were obtained in work based on intranasal inoculation of H. influenzae, where also mixed blood cultures were detected in the first minutes after challenge and before the eclipse period [25]. We propose for our intra venous injection model that during this time, a fraction of the inoculated bacteria are sequestered in extravascular tissues, most probably in the spleen, in accordance with our data on positivity of bacterial spleen counts also in mice with negative blood cultures (Figure S1). This emergence of a clone from the potential splenic focus into the “sterile” bloodstream can be viewed as equivalent to the “invasive events” described for models which consider more than one organ system [16], [18], [21], [23]–[25]. Sustained bacteraemia is initiated from replication of one bacterial cell, perhaps a stochastic event in which the first replicon to reach a threshold biomass sufficient to seed the blood “wins the day”. The exponential increase in the number of bacteria in the blood is consistent with contributions from both intravascular and extravascular replication of pneumococci. We favour a scenario in which, at a challenge dose below the LD50, the rate of replication occurring in the extra-vascular site, followed by seeding of bacteria to the blood, exceeds host clearance rates thereby resulting in progressively more severe bacteraemia. Our data do not infer that only one pneumoccocus survives the initial host clearance, but rather that from those that do survive; only one cell initiates bacteraemia. The observed increase of polyclonal infections over time, as predicted also by the independent action hypothesis, is in accordance with the doubling of the ratio of infected mice at those time points (0.31 at 24 h to 0.67 at 72 h) [22]. The strong positive selection which drives the emergence of the ATPase-SNP subclones during the later phase of the infection is novel with respect to previous models (i.e. the live-death model), which postulates a neutral selection during this phase [16], [17], [37]. The in depth genomic analysis, in contrast to previous works [16], [17], [37], shows evidence for a more dynamic behaviour of the infecting bacterial population with an increase in heterogeneity of the monoclonal population over time due to a strong positive selection after the single cell bottleneck.
However, we observed added complexity; the residual, but inadequate, innate clearance mechanisms exert a selective pressure resulting in the emergence of adaptive mutants. Sequencing of bacteria from blood revealed that in most mice the bacterial clones had each acquired SNPs in different sub-units of the pneumococcal F1/F0 ATPase gene. This apparently high frequency of mutations, given the relatively small biomass of pneumococci in each animal, is consistent with the estimated mutation rates of up to 5×10−4 per genome described recently for pneumococci during one-cell bottleneck in vitro passages [38]. The selection for altered function of the ATPase, was found only in a proportion of the bacteria making up the population obtained from blood, compelling evidence that the ATPase mutations must have occurred after the single cell bottleneck. As stated above, the observation of subclones being selected during the bacteremic phase underlines a highly dynamic situation, which extends over the neutral two stage infection models [16], [17], [37].
In pneumococci, it has been recognised that ATPase mutations occur at high frequency during pneumococcal infection in humans, possibly in response to oxidative stress [39], and have been described both in vitro and in clinical isolates [40]–[44]. Polymorphisms in F0_atpA and F0_atpC (the trans-membrane part of the ATPase) were found to confer phenotypes of reduced susceptibility to optochin, quinine and mefloquine [40], [42], [43]. In particular, the detection of optochin resistant pneumococci in clinical samples is well described [44], as it has a practical impact on pneumococcal identification in the diagnostic laboratory [41]. None of the ex vivo ATPase mutants in our investigation were optochin resistant and the SNPs accordingly did not map to the optochin resistance conferring regions.
The F1/F0 ATPase is encoded by a highly conserved eight-gene operon and, as in aero-tolerant anaerobes, it is involved in the maintenance of intracellular pH through the generation of a membrane proton gradient [45]. In some of the mutants we were able to identify a clear metabolic benefit of the mutations which enabled growth at pH lower than 6.8, albeit all mutants showed that loss of capacity to grow at pH above 7.8. Interestingly the phenotype of TIGR4 mutants recovered from blood was not different from other virulent pneumococcal stains, such as D39. The high frequency of mutation observed here, given by the many different sites mutated, strongly suggests within-host adaptation through selective pressure during sepsis. While in vitro susceptibility of the ATPase mutants to antibody mediated neutrophil killing and macrophage phagocytosis was essentially unaltered, the phenotypic consequence of the ATPase mutations may be linked to a gain in fitness related to the increased survival of bacteria within the splenic, extravascular focus that provides the source of pneumococci re-seeding the blood and sustaining the progressively escalating and ultimately lethal bacteraemia. In agreement with this hypothesis is the recent description of inhibition of the own F1F0 ATPase by both Salmonella enterica and Mycobacterium tuberculosis as strategy to withstand phagolysosomal activity [46].
In summary, we propose that after the majority of the bacteria of the challenge inoculum have been removed, a few bacteria survive the predominantly lethal activity of splenic macrophages and neutrophils. From these rare survivors, single pneumococcal cells may start to replicate and initiate seeding of the blood resulting in a steady state bacteraemia in which efficient host clearance is off-set by re-seeding from the original, persisting extravascular reservoir of bacteria. These extravascular bacteria are subjected to strong selection for adaptive mutations. Later during infection, selected subpopulations of the initial clone may become part of the bacterial population causing disease. These observations are in accordance with a two stage model of infection where independent action generating the initial stochastic event is followed by a dynamic birth-death phase which increases heterogeneity due to strong selection [16], [17], [24]. In the case of the model organism S. pneumoniae, our data show different selective pressures shaping the invasive bacterial population during different phases of infection [16]. Given the demonstration that pneumococci are independent in generating disease in our rodent model and that less than twenty per cent of human pneumonia cases are bacteraemic [3], we hypothesize that human pneumococcal bacteraemia is generally monoclonal originating from a single cell in analogy to the monoclonal meningitis case recently described [47]. Presentation of a model which foresees development of invasive disease from a single bacterium and strong selection during outgrowth represents an important example on which to model fitness selection during invasive infection.
Three isogenic zmpC knock-out mutants of TIGR4 (FP122, FP318 and FP321) that differed only for the resistance marker, ermB (erythromycin resistance), aad9 (spectinomycin resistance) and aphIII (kanamycin resistance), respectively were constructed for co-infection studies with isogenic clones [28], [48]. The experiments with mice depleted of macrophages and neutrophils were done with four different strains: TIGR4 (serotype 4; strain FP321 zmpC::aphIII), G54 (serotype 19F; erythromycin and tetracycline resistant) [49], [50], D39 (serotype 2; strain FP335 bglA::aad9; gift of Hasan Yesilkaya, Leicester), and the streptomycin resistant non-encapsulated D39 derivative DP1004 [51], [52]. The transfer of the atpC frame-shift into DP1004 was performed by transformation of a marker flanked by two PCR fragments, one of which containing the frame-shift. This was possible since the atpC SNP is only 76 bp from the end of the operon. Two representative transformants FP499 and FP500 were confirmed by sequencing. The series of ATPase mutants isolated are described in Table 3, while all other strains are listed in Table 1. Strains were cultured in Tryptic Soy Broth (TSB, Liophilchem, Teramo) or Todd Hewitt (THY, Oxoid, Milano) supplemented with 0.5% Yeast Extract (Liophilchem). Solid media were blood agar plates (Tryptic soy agar, Difco) supplemented with 3% horse blood (Biotech, Grosseto). The colony morphology was checked on Todd-Hewitt agar plates containing 200 units/ml of catalase (Sigma-Aldrich, Milano, Italy) [53], [54]. Antibiotics were used at the following concentrations: 1 mg/L erythromycin, 500 mg/L kanamycin, 100 mg/L spectinomycin and 500 mg/L streptomycin (all from Sigma-Aldrich).
The intracellular pH was determined by Nuclear magnetic resonance (NMR). In brief, 400 ml of mid log pneumococcal cells grown in Todd Hewitt broth were pelleted and mixed with 1 ml of sodium alginate 6% (w/v 0.9‰ NaCl). Mixture were extruded manually through 25G needle on a surface of 0.25 M CaCl2 solution. The small drops were washed and transferred in the 10 mm NMR tube. NMR 31P spectra were recorded on a Bruker DRX 600 instrument operating at 242.9 MHz. 31P spectra were recorded with a 1.5 s repetition time and 45°flip angle. Line broadening of 10 Hz were applied before Fourier Transform. 31P chemical shift were determined by comparison with external standard Trisodium trimetaphospate at −20.80 ppm. Intracellular pH was determinate by Pi (intracellular phosphate) chemical shift in phosphate-free perfusion model [55]. Active metabolism of pneumococci was confirmed by acidification of the extracellular medium during the experiment carried out at 37°C.
Growth profiles of wild type strains and ATPase mutants were assayed both in standard laboratory media and in defined media. Standard laboratory media included TSB (Liophilchem) and Todd-Hewitt broth supplemented with Yeast Extract (0.5%) (Oxoid). Defined media were prepared in CAT medium by adding serial concentration of potassium phosphate buffer with different range of pH (6 to 8) and by adding several concentration of K2HPO4 as source of salt. CAT medium was composed by: Casitone (10gr/l) (Becton Dickinson), Tryptone (10 gr/l) (Oxoid), Yeast Extract (1 gr/l) (Liophilchem), NaCl (5 gr/l) (Panreac, Milano, Italy), Catalase (200 U/ml) (Sigma-Aldrich) and Glucose (0.2%) (J.T. Baker, Milano, Italy). Metabolism of pneumococcal strains including wild type and ATPase mutants were assayed by Phenotype MicroArray (PM) microplate PM9 containing a total of 96 different osmolyte sources. PM technology measures active metabolism by recording the irreversible reduction of tetrazolium violet to formazan as an indirect evidence for NADH production. PM procedures were carried out as previously described (Viti C 2009). Quantitative colour change were recorded automatically every 15 min for a period of 72 h. Analyses were performed by the Omnilog-PM Software (Biolog, inc.) and data were filtered using average height as a parameter.
Animal experimentation in Italy is regulated by Decreto Legislativo 116/92 and Directive 210/63/EU. The animal protocol was approved by the “Comitato Etico Locale” of the Azienda Universitaria Ospedaliera Senese and received thereafter the relative project licence issued by the Italian Ministry of Health (193/2008-B).
Six to seven-weeks old female CD1, BALB/c, and CBA/Ca mice were purchased from Charles River Italia (Lecco, Italy). For the bottleneck experiments, outbred CD1 mice were used, while BALB/c mice were chosen for both in vivo macrophage and neutrophil depletion and ex vivo experiments. CBA/Ca data are shown only for comparison of the dynamics of the early phases of infection. Animals were sacrificed by intraperitoneal (i.p.) injection of xylazine hydrochloride and zolazepam tiletamine cocktail (Xilor 2%, Bio 98 S.r.l., Bologna, Italy and Zoletil 20, Virbac S.r.l., Milano, Italy). Mice were kept at the animal facility of the LAMMB, University of Siena, according to its guidelines for the maintenance of laboratory animals [48], [56]–[58]. Blood samples from mice were collected by sub-mandibular vein or cardiac puncture under terminal anaesthesia. To prevent blood coagulation, 100 U/ml of heparin (MS Pharma, Milano, Italy) was added. All the collected organs (spleen, lung, liver and kidney) were homogenized in 1 ml of TSB, and then frozen at −80°C after making to 10% v/v of glycerol.
Two series of experiments were performed in order to define the bottleneck for invasive pneumococcal infection with a total of 68 CD1 mice. Mice were challenged intravenously (i.v.) as described [48], [56]–[58] with a mixture of the three isogenic TIGR4 derivatives (FP122, FP318 and FP321) at 3.3×105 CFU each/mouse. At pre-set time points blood samples were collected and selected groups were sacrificed for obtainment of spleen samples. Two blood samples from each animal, taken at different time points, are reported in Figure 1. Bacteria were enumerated by plating on selective media. The dose of the experiment was decided after having observed in a preliminary experiment 5/8 mixed and 3/8 monoclonal infections using two pneumococcal clones at a dose of 2×10∧6 (data not shown).
A pilot experiment for comparison of virulence in CD1, BALB/c and CBA/Ca mice was carried out by infecting i.v. four mice each with a mixture of G54, D39 and TIGR4 (3×105 CFU each/mouse). Three blood samples per mouse were obtained.
For depletion of macrophages, BALB/c mice were treated 24 h prior to challenge by i.p. injection with 750 µl of a suspension of clodronate (CL2MBP) liposomes. One control group received PBS-containing liposomes [31] and the other was untreated. Clodronate was encapsulated in liposomes, as described earlier [31] and was a gift of Roche Diagnostics (Mannheim, Germany). Neutrophil depletion was performed by a single i.p. injection of 150 µg/mouse of anti-GR-1 antibody (Ly6G and Ly6C, clone RB8-8C5; Becton Dickinson) 24 h prior to infection [32], [33]. Two control groups were either left untreated or administered with a rat isotype control antibody IgG2b K (kappa) (Becton Dickinson). Groups of mice depleted of macrophages or neutrophils were infected i.v. with 1×106 CFU/mouse containing 2.5×105 CFU of each TIGR4, D39, DP1004 and G54. Bacterial viable counts were determined at preset time points.
The virulence of the ATPase mutant FP487 (atpC mutant) was assayed in parallel with TIGR4. BALB/c mice (n = 6) were infected with 1×106 CFU/mouse i.v. and blood and spleen samples collected at 72 h.
Spleen and bone marrow macrophages were isolated from mice using a modified protocol previously described [34]. Cells were cultured in medium supplemented with 25 ng/ml of recombinant M-CSF (Invitrogen) and re-seeded at day 7 at the concentration of 2×105 cell/ml. After 24 h, 0.1 ml of pneumococci cultured to OD590 0.25 were added. After 45 min plates were washed and reincubated with 10 mg/L of penicillin and 200 mg/L of gentamicin for 30 min. Intracellular bacteria were enumerated after lysis with saponin 1%. Phagocytosis of RAW264.7 macrophages followed the same protocol, but in addition samples were reincubated after removal of the antibiotics for an additional hour in fresh medium.
Flow cytometric analysis was conducted on bacteria suspended in 1% v/v paraformaldehyde in PBS on a FACScalibur machine (Becton Dickinson, California, USA). To verify macrophage and neutrophil depletion, homogenised organ samples were washed in DMEM (Sigma-Aldrich) and non-specific binding was blocked with FcR blocking agent [59]. Cells were incubated 30 min with 1 µg of specific fluorochrome-conjugated antibodies per 106 cells. Neutrophils were stained using a rat anti-GR-1 antibody (MACS, Bologna, Italy). Macrophages were detected with rat anti-F4/80 mAb (BM8 clone; Abcam, Milano, Italy), and a rat anti-mouse CD11b mAb (Becton-Dickinson). Surface markers of macrophages were analysed using the following antibodies: anti-F4/80 mAb, anti-mouse CD11b mAb, anti-CD11c mAb (eBioscience), anti-mouse SIGNR1/CD209b Ab, goat IgG control Ab, anti-mouse Siglec-1 mAb, rat IgG2A Isotype control Ab, anti-mouse MARCO mAb and rat IgG1 isotype control Ab (R&D Systems).
To assay for the presence of anti-pneumococcal antibodies in mouse sera, the four pneumococcal strains TIGR4, G54, D39 and DP1004 were blocked in PBS-BSA 1% v/v for 30 min at 37°C and incubated for 1 h at 37°C with sera (1∶100) obtained from BALB/c mice and the positive anti-serotype 2 control serum (Staten Serum Institute, SSI, Copenhagen, DK). Samples were marked with anti-mouse IgG (1∶64) or anti-rabbit IgG (1∶160; Sigma-Aldrich).
In order to evaluate the capacity of whole blood to kill or inhibit the multiplication of pneumococci and to investigate the effect of specific antibodies, ex vivo experiments were set up. Blood from BALB/c mice was collected into tubes containing heparin and infected with pneumococci. For the assay of opsono-phagocytosis of ATPase mutants 1×104 CFU/ml of parental and ATPase mutant were inoculated in blood and incubated in rotation. The anti-type 4 serum (SSI) was used at 1∶50 dilution. For the evaluation of growth of pneumococci in blood 3×105 CFU/ml of G54, D39 and TIGR4 were inoculated in rotating blood. The efficacy of type 2 anti-serum (SSI) on D39 and its non-encapsulated derivative DP1004 was assayed as above using a inoculum of 3×105 CFU/ml and a 1∶100 dilution of the serum.
Chromosomal DNA was extracted using the High Pure PCR Template preparation kit (Roche). Whole genome sequencing was performed by the Institute of Applied Genomics and IGA Technology Services srl (University of Udine, Italy) using an Illumina (Solexa) Genome Analyzer II platform [60]. Reads of both, parent and mutant strains, were aligned to the reference genome of TIGR4 (accession NC_003028) using the Mosaik Assembler suite (The MarthLab, Boston College, Massachusetts, USA). Single nucleotide polymorphisms (SNPs), insertions and deletions (INDELs) were retrieved with VarScan software [61]. SNPs and INDELs of the challenge strains were subtracted from those found by aligning the blood isolates. All F1/F0 ATPase mutations were re-sequenced by the Sanger method and deposited in GenBank (accession KF705516 to KF705525).
In order to evaluate the number of bacteria at the origin of blood infection, a model derived from that previously described [23], was developed. A full description of the statistical model is given in the supplementary materials. Statistical analysis of bacterial counts in blood and organs was performed by the Student's t-test for data reported in Figures 3, 4, S2 and S3. The analysis of different bacterial blood clearance at 5 and 30 min and the differences in bacterial phagocytosis and data of phenotype microarray were performed using Kruskal-Wallis and Dunn's multiple comparison post test (Figure 2 C–F and S5). Values of P<0.05 were considered statistically significant. The Fluorescence Index (Figure S3 K) was calculated by multiplying the percentage of positive events with the geometric mean fluorescence intensity (GeoMean).
|
10.1371/journal.ppat.1006781 | Association of papillomavirus E6 proteins with either MAML1 or E6AP clusters E6 proteins by structure, function, and evolutionary relatedness | Papillomavirus E6 proteins bind to LXXLL peptide motifs displayed on targeted cellular proteins. Alpha genus HPV E6 proteins associate with the cellular ubiquitin ligase E6AP (UBE3A), by binding to an LXXLL peptide (ELTLQELLGEE) displayed by E6AP, thereby stimulating E6AP ubiquitin ligase activity. Beta, Gamma, and Delta genera E6 proteins bind a similar LXXLL peptide (WMSDLDDLLGS) on the cellular transcriptional co-activator MAML1 and thereby repress Notch signaling. We expressed 45 different animal and human E6 proteins from diverse papillomavirus genera to ascertain the overall preference of E6 proteins for E6AP or MAML1. E6 proteins from all HPV genera except Alpha preferentially interacted with MAML1 over E6AP. Among animal papillomaviruses, E6 proteins from certain ungulate (SsPV1 from pigs) and cetacean (porpoises and dolphins) hosts functionally resembled Alpha genus HPV by binding and targeting the degradation of E6AP. Beta genus HPV E6 proteins functionally clustered with Delta, Pi, Tau, Gamma, Chi, Mu, Lambda, Iota, Dyokappa, Rho, and Dyolambda E6 proteins to bind and repress MAML1. None of the tested E6 proteins physically and functionally interacted with both MAML1 and E6AP, indicating an evolutionary split. Further, interaction of an E6 protein was insufficient to activate degradation of E6AP, indicating that E6 proteins that target E6AP co-evolved to separately acquire both binding and triggering of ubiquitin ligase activation. E6 proteins with similar biological function clustered together in phylogenetic trees and shared structural features. This suggests that the divergence of E6 proteins from either MAML1 or E6AP binding preference is a major event in papillomavirus evolution.
| Papillomaviruses are a large family of viruses with great medical and veterinary importance. This study explores the viral E6 oncoproteins from diverse papillomavirus genera to determine how E6 distinguishes in interaction between cellular proteins. E6 proteins have been previously found to interact with a ubiquitin ligase called E6AP and thereby target particular cellular proteins for degradation, or to interact with MAML family proteins to repress Notch signaling and thereby alter cellular differentiation. It has been unclear if diverse families of papillomavirus E6 proteins interact with only E6AP or MAML (or possibly both), how E6 distinguishes between these interactions, and if interaction of E6 with E6AP is coupled to ubiquitin ligase activation. We find here that none of the tested E6 proteins physically and functionally interacted with both E6AP and MAML1, indicating an evolutionary split that clustered E6 proteins by sequence similarity analysis. Currently, the categorization of papillomaviruses is complex, with thirty-eight genera so far described. This study establishes an early evolutionary split among most papillomavirus genera between those viruses that encode E6 proteins that physically and functionally associate with MAML compared to E6AP. This provides a structural and functional basis for categorizing most currently described papillomaviruses into two major functional groups.
| Papillomaviruses are a large group of viruses with hundreds of different fully sequenced types and additional types partially characterized by metagenomic sequencing [1–4]. All papillomaviruses express early genes E1 and E2 that are necessary for viral transcriptional control and DNA replication, as well as late gene capsid proteins L1 and L2 that package progeny viral DNA. Almost all papillomaviruses also express accessory early proteins with oncogenic properties, which are subject to transcriptional control by E1 and E2 (termed E5, E6, E7, and other designations) [5–9]. While many papillomaviruses encode all three oncoproteins, one or more of the oncoproteins may be absent within a particular group of related papillomaviruses. Papillomaviruses are classified based upon nucleotide sequence similarity of the major L1 capsid protein into taxonomic levels of genus (different genera share less than 60% nucleotide sequence identity in L1), species (share between 60% and 70% nucleotide identity), and types (types within a species are between 71% and 89% identical in sequence within L1) [10]. Thirty-eight genera of papillomaviruses have been described, with additional genera expected as additional animal papillomaviruses will be detected in the future [4, 11].
Papillomavirus clustering on the basis of early gene relatedness instead of L1 has also been performed [12, 13], and shows more coherent clustering of E6 by clinical phenotype compared to L1 [14]. However, L1 clustering is practical, primarily because L1 is more highly conserved than the early region genes, and all papillomaviruses encode the L1 gene. But given the large and growing number of genera, a parallel approach that clusters papillomaviruses into smaller numbers of groups with biological relatedness might be more comprehensible.
Although all virus-induced papillomas are initially benign, some virus types produce papillomas that may progress into malignancy. Some Alpha genus human papillomaviruses (HPVs) are associated with anogenital and upper airway cancers in primates (reviewed in [15]), and certain animal papillomaviruses (such as cotton tailed rabbit papillomavirus [16], and Ovies Aires Papillomavirus type 3 [17], produce cutaneous papillomas that can progress to malignancy. The subset of HPVs associated with cancer is referred to as “high-risk” HPV types (HPV types 16, 18, and 31 are model systems), and the related mucosal Alpha genus viruses that do not cause malignancies are called “low-risk” HPVs (prototypes being HPV types 6 and 11) [18].
The propensity of a particular papillomavirus to produce a cancer is a property of the virally encoded oncoproteins, and in some viruses, environmental exposure to carcinogens (such as Bovine Papillomavirus type 2 when cows consume bracken fern [19]). High-risk HPVs encode an E7 oncoprotein that associates with retinoblastoma family proteins and targets them for degradation, thereby ablating cell cycle checkpoint control and contributing to genomic instability [20]. The HPV high-risk E6 oncoproteins associate with a cellular E3 ubiquitin ligase called E6AP (a product of the UBE3A gene) [21]. E6 associates with E6AP by docking upon an alpha-helical LXXLL-containing peptide motif in E6AP [21–24]. Upon binding an LXXLL peptide, E6 undergoes a conformational change that promotes the association of E6 with p53 [25], p53 ubiquitination by E6AP, and p53 degradation by the proteasome. Therefore, like E7, high-risk E6 promotes genomic instability [26]. Low-risk mucosal HPV types also express E7 proteins but these do not target the degradation of Retinoblastoma (RB) [20]. While low-risk E6 proteins bind to E6AP and trigger E6AP degradation, they have not been shown to interact with p53 [27].
It is clear that E6 proteins from non-Alpha genera do not primarily associate with E6AP. The Delta genus E6 protein from Bovine Papillomavirus type 1 (BPV1 E6) associates with focal adhesion proteins paxillin (PXN) and HIC5 by binding to LXXLL motifs on those proteins that are similar to but distinct from the LXXLL motif of E6AP [28, 29]. PXN expression, BPV1 E6 interaction with LXXLL motifs, and docking of BPV1 E6 on particular LXXLL motifs of PXN are all required for BPV1 E6 to induce anchorage-independent colony formation [30–33]. Recently, E6 proteins from HPV Beta genus cutaneous papillomaviruses as well as BPV1 E6 were found to interact with MAML1 transcriptional coactivators, and thereby repress Notch signaling [34–37]. The interaction of these E6 proteins with MAML1 resembled that of BPV1 E6 with PXN and HIC5, and HPV16 E6 protein with E6AP, in that the cutaneous type E6 proteins bound to an LXXLL peptide motif located at the carboxy-terminus of MAML1. BPV1 E6 and Beta genus HPV E6 proteins preferentially interact with MAML1 compared to E6AP despite the similarity of the LXXLL binding motifs of E6AP and MAML1 [34].
Relatively few E6 proteins have been characterized for their associations with E6AP compared to MAML1. In this study, we have expressed 45 different E6 proteins from 21 animal and human genera. We have determined the preferential association of each E6 protein with either E6AP or MAML1, and performed functional assays for either transcriptional repression of MAML1 or the stimulation of E6AP degradation by E6 in vivo. The results clearly divide genera of papillomaviruses that express E6 proteins into those that physically and functionally target MAML1 and not E6AP, and those that physically and functionally target E6AP and not MAML1. We further make mutations in E6AP and thereby demonstrate that Alpha genus E6 proteins have a second function separate from binding to the LXXLL motif of E6AP that is required to efficiently stimulate E6AP degradation, demonstrating that the ability of Alpha genus E6 proteins to confer E6AP dependent degradation of cellular proteins is not triggered solely by association with an LXXLL motif on E6AP. This analysis delineates a profound functional split among papillomavirus genera and suggests that a functional grouping of papillomaviruses into those that target Notch signaling through interaction with MAML proteins, and those that target cellular protein degradation through association with E6AP provides insight into papillomavirus taxonomy.
Fig 1 shows LXXLL motifs from cellular proteins that associate with E6 proteins. The LXXLL motifs from E6AP and IRF3 are bound by some Alpha genus E6 proteins [27, 38, 39]. PXN and HIC5 motifs are bound by Delta genus BPV1 E6 [31], and MAML1 by Delta, Mu, Pi and Beta genus HPV E6 proteins [34–37, 40]. Although there are clear similarities among these few known LXXLL motifs, the range of possible binding sites from a very large set of random possible ligands has not been explored.
In order to explore LXXLL binding motifs in an unbiased way, BPV1 E6 was expressed and used to select peptide interactors from a very large (> 1012 independent clones) phage display library where random sequence 12-mer peptides were displayed as fusions to M13 gene-III. After five rounds of selection, 25 plaques were sequenced revealing 14 unique sequences; all the unique sequences are shown in Fig 1 and aligned where a ØXXØØ-like sequence (Ø representing a hydrophobic side chain) could be discerned; a consensus sequence and position numbering are at the bottom of Fig 1. The core ØXXØØ was observed in all 14 peptides, as LXXLL six times and as LXXLF eight times. Notably, a clear selection for a hydrophobic residue at position -3 upstream of the LXXL/F was observed in 12 of 14 selected peptides, making this the second most prevalent selected feature, followed by the presence of acidic side chains at positions two (10 of 14 peptides), positions -1 and 3 (found in six of 14 peptides) and position 7 (four of 14 peptides). A consensus for preferred BPV1 E6-bound peptides from this experiment is thus ØX(D)L(D/E)(D/E)L(L/F)X(D/E). This consensus is quite similar to the LXXLL motifs in cellular proteins that bind BPV1 E6, shown in Fig 1A.
The strong prevalence of a hydrophobic side chain at position -3 in BPV1-bound phage-selected peptides contrasts with the LXXLL peptides found in E6AP and IRF3 that interact with high-risk Alpha genus HPV16 E6, where position -3 in both LXXLL peptides is a glutamic acid. The crystal structures of BPV1 E6 and HPV16 E6 bound to LXXLL motifs of PXN and E6AP respectively have been solved, and show a conserved overall fold and mode of interaction between the two E6 proteins and their respective LXXLL peptides [41] [42]. In the BPV1 E6 structure, the amino acid side chain M1 of the bound PXN-derived LXXLL peptide resides within a hydrophobic pocket comprised of E61, L64 and W65 and the side chain of R116 (that interacts with E61 of BPV1 E6 and D5 of the PXN LXXLL peptide); in the HPV16 E6–E6AP structure, the analogous E1 residue of the E6AP LXXLL peptide interacts with the analogous E6 residues of HPV16 E6 (S74, H78, R77 and R129) [41]. The overall structure and detail illustrating this is shown in Fig 2. The strong selection for a hydrophobic amino acid side chain at position -3 among the BPV1 E6 phage-selected peptides illuminates the important role of this position when E6 proteins discriminate between interactions with the LXXLL peptide derived from E6AP, compared to peptides from PXN, HIC5, MAML1 or MAML3. Interestingly, BPV1 E6 selected peptides with an LXXLF sequence as well as LXXLL (Fig 1B). LXXLF is found in MAML3, and MAML3 as well as MAML1 were found in association with BPV1 E6 [37].
Given that analogous amino acids in HPV16 E6 and BPV1 E6 served similar structural functions in binding to position -3 of their respective LXXLL peptides, we wondered if examination of these contact amino acids in diverse E6 proteins might predict the preference of a given E6 protein for either E6AP or MAML1. Within the Alpha genus, precise conservation of these contact residues was overall modest. For the four amino acids with closest contacts with position -3 of E6AP, S74 was poorly conserved, R77 was highly conserved and H78 was moderately conserved (present in 52/78 Alpha genus E6 proteins); interestingly, H78 was replaced by a hydrophobic residue (similar to that seen in BPV1 E6) in 26/78 Alpha E6 sequences, and the remaining contact residue, R129, was poorly conserved (a multiple sequence alignment of the Alpha genus E6 proteins is shown in S1 Fig). Thus, examination of the contact residues of Alpha genus E6 proteins found they were not sufficiently conserved and could not indicate that E6 proteins would interact with E6AP and/or possibly an additional LXXLL binding site such as on MAML1.
As noted above, the Beta genus HPV E6 proteins interact with the LXXLL motif of MAML1. A similar examination of the predicted contact residues between E6 and position M1 of the MAML1 LXXLL was more consistently predictive, but not completely. The position analogous to HPV16 E6 S74 was highly conserved as a glutamic or aspartic acid, the position of HPV16 E6 R77 was not conserved, but the position analogous to HPV16 H78 was hydrophobic as expected (being either phenylalanine or tyrosine in 48/50 examined Beta genus sequences), and the R129 position of HPV16 E6 (analogous to R116 of BPV1 E6) was completely conserved in the Beta and Gamma genus E6 proteins (S2 and S3 Figs).
The heterogeneity in the E6 contact residues for interaction with position -3 of the LXXLL binding sites suggested the possibility that some E6 proteins within a genus might have multiple different LXXLL interaction targets, perhaps interacting with both E6AP and MAML proteins as has been recently suggested for Gamma genus HPV197 E6 [43]. We wished to determine if within or between papillomavirus genera, E6 proteins had distinct binding and functional preferences for either E6AP or MAML family proteins. We synthesized 45 different E6 genes where the E6 protein shared the canonical E6 structure of two zinc structured domains connected by an alpha helix, and one E6 protein (OcPV1) where an additional zinc structured domain is found at the carboxy-terminus (as is present in Shope Papillomavirus E6). Table 1 shows the tested E6 proteins and their host species. A multiple sequence analysis of the 45 tested E6 proteins is shown in Fig 3, and shows the amino acid positions in HPV16 and BPV1 E6 amino acids that interact with position -3 of the E6AP or MAML1 LXXLL peptides. The position of the tested E6 proteins within a phylogeny map of animal E6 proteins and a subset of HPV E6 proteins is shown in S4 Fig.
To test the preference of each E6 protein for either MAML1 or E6AP association, we co-expressed FLAG-E6 together with an excess of both HA-E6AP_Ub- (Ub- means mutated in ubiquitin ligase activity) and HA-MAML1, such that HA-E6AP_Ub- and HA-MAML1 were similar in abundance, and less than 10% of the input HA-tagged proteins were immunoprecipitated by FLAG-E6. We reasoned that overexpression of the HA-tagged E6-binding targets compared to E6 would promote competition between HA-E6AP and HA-MAML1 for the available E6, and thereby serve as an indication of the preference of each E6 protein for either MAML1 or E6AP_Ub-. Fig 4A shows a typical experimental result with nine different E6 proteins, some of which the preference for E6AP or MAML1 association was previously described [34, 81]. Representative western blots for the remaining E6 proteins are shown in S5 Fig. These experiments were repeated to obtain quantitative results with error which is shown in Fig 5A. Typically, E6 proteins showed a clear preference for either E6AP or MAML1 association. In some cases, an E6 protein clearly discriminates in binding between E6AP compared to MAML1, but that discrimination did not reach statistical significance in the three experimental replicates (such as HPV11, 17, and 8). Typically, those E6 types had low levels of association to both E6AP and MAML1; low expression levels reduce net levels over background and therefore increases the associated error. Even in those cases, a preference of binding to either E6AP or MAML1 is clear on the multiple western blots, allowing a qualitative assessment of binding preference to be made.
We desired to correlate the physical association of E6 with E6AP to an in vivo biological function. We and others had previously observed that overexpression of E6 together with E6AP WT results in reduced expression of E6AP because of E6AP ubiquitin ligase activity and proteasome mediated degradation; in contrast, BPV1 E6 does not stimulate E6AP degradation [27, 82]. Each E6 protein was co-transfected with E6AP WT, and the reduction in E6AP expression upon co-expression with E6 was ascertained by western blot (Fig 4B and S6 Fig) with quantitation after three experimental replicates shown in Fig 5B. As expected, E6 proteins from the Alpha genus (HPV types 7, 10, 11, 16, 18 and MmPV1) that associated preferentially with E6AP also stimulated E6AP degradation. While MmPV1 (also known as RhPV1 for Rhesus Papillomavirus Type 1) was less active than other Alpha proteins in stimulating E6AP degradation, it did so significantly and also promoted degradation of p53 in an E6AP-dependent manner (S7 Fig). Other genera whose E6 proteins preferentially associated with E6AP also stimulated E6AP degradation, including Omega (UmPV1), DyoDelta (SsPV1), Omikron (PhPV1, TtPV5, PsPV1), and Dyopipa (PhPV4).
To correlate the association of E6 proteins with MAML1 to an in vivo biological activity, we co-transfected the test set of E6 proteins together with a GAL4_MAML1 fusion, and measured the repression of a GAL4 responsive luciferase reporter as previously described [34]. Fig 5C shows that E6 proteins that preferentially associate with MAML1 also repress GAL4_MAML1 transcriptional activation. Some E6 proteins that weakly associated with MAML1 by IP/WB still significantly repressed GAL4-MAML1 activity (HPV8, HPV17, HPV1, HPV123, MnPV1, MfPV2, and McPV2), indicating that some of the weak interactions with MAML1 in our binding assay were biologically significant. Only HPV41 E6 strongly associated with MAML1 but failed to significantly repress GAL4-MAML1 transcriptional activation (Fig 5A and 5C).
Some E6 proteins that did not clearly associate with either E6AP or MAML1 also failed to repress either GAL4_MAML1 activity or reduce E6AP expression levels (TmPV1, TmPV2, RaPV1, BpPV1, and OcPV1). Examination of the LXXLL motif at the carboxyl-terminus of MAML1, 2, or 3 reveals significant differences between the three LXXLL motifs (S1 Table). We therefore tested the E6 proteins that interacted poorly with E6AP and MAML1 for their possible interaction with either MAML2 or MAML3 by co-expression and co-immunoprecipitation. Only BPV1 E6 co-immunoprecipitated MAML3, which is consistent with the prior finding of BPV1 E6 in association with MAML3 from the Howley lab [37]. Only CPV7 E6 showed preferential interaction with MAML2 compared to MAML1 (Fig 6). Thus overall, there was a strong preference for MAML1 association with E6 from many divergent host species.
The LXXLL motifs of E6AP, MAML2, and MAML3 were quite conserved among the host species in our test set of E6 proteins, however several host species had one or more amino acid changes in the LXXLL motif of MAML1 (S1 Table). To test the relevance of host-specific changes in the MAML1 LXXLL motifs, we mutated the LXXLL motifs in human MAML1 into the LXXLL motifs of rodent or canine species. These chimeric MAML1 molecules were co-transfected in comparison with human MAML1 to determine if species preference for cognate LXXLL motifs accounted for the limited associations of these E6 proteins with human MAML1 observed in Fig 5. Surprisingly, all tested rodent and canine E6 proteins showed similar preference for human or their cognate species MAML1 chimeric protein except MmaPV1, which surprisingly, showed a modest preference for human MAML1 compared to the human-rodent MAML1 chimera (S8 Fig).
Using the multiple sequence alignment (Fig 3), a cladogram was generated to illustrate the clustering of the tested E6 proteins by sequence relatedness. Qualitative binding data and statistical significance for the physical association of each tested E6 protein with MAML1 and E6AP is shown, compared together with E6AP degradation and GAL4-MAML1 repression to illustrate clustering of preferential interactions with biological functions (Fig 7). Several observations were apparent.
First, E6 proteins from most tested genera preferentially associated with MAML1 compared to E6AP. Where the physical association with MAML1 was statistically significant, in all but one case (HPV41) GAL4_MAML1 activity was significantly repressed. All the tested Gamma and Nu (HPV41) genera E6 proteins preferentially associated with MAML1 and all except HPV41 E6 significantly repressed GAL4_MAML1.
Second, E6 proteins that bound neither E6AP nor MAML1 did not cluster with E6 proteins that did associate with MAML1 or E6AP: OcPV1, TmPV1, TmPV2, RaPV1, and BpPV1. RaPV1 bound both MAML1 and E6AP at low levels, but neither decreased E6AP expression nor repressed GAL4_MAML1 transcriptional activation. The characterization of our test set of E6 proteins within the larger set of all animal E6 proteins and a subset of HPV E6 proteins is shown in S9 Fig.
Third, E6 proteins that preferentially associated with E6AP compared to MAML1 both reduced E6AP expression levels and failed to repress GAL4_MAML1 transcriptional activation. Phylogenetically clustered together with the Alpha genera E6 proteins are E6 proteins from Omega, Omikron, DyoDelta, and Dyopipa genera. Since only E6 proteins that preferentially associated with E6AP were able to reduce its expression level, it is unclear from these experiments if it was the docking of an E6 protein to the LXXLL motif of E6AP that triggers degradation, or if a second property of E6, subsequent to the initial and required binding to the LXXLL motif, was also required to initiate degradation.
To address this question, an E6AP molecule was constructed where the LXXLL binding motif of E6AP (ELTLQELLGEE) was mutated into the LXXLL motif of MAML1 (MSDLDDLLGS, the new E6AP mutant is termed E6AP_LDDLL). While BPV1 E6, and E6 proteins from HPV types 112, 4, and 131 (Gamma genus) preferentially associate with MAML1 and not E6AP (Fig 5), all interacted robustly with E6AP_Ub-_LDDLL (Fig 8A). Interestingly, HPV16 E6 bound both E6AP_Ub- and to a lesser extent, E6AP_Ub-_LDDLL, despite poor interaction with MAML1 (Figs 5 and 8A). HPV16 E6 still initiated degradation of both E6AP WT and E6AP_LDDLL. In contrast, BPV1 and HPV types 4, 112, and 131 E6 proteins did not interact with E6AP_Ub-, but interacted robustly with E6AP_LDDLL, even more than did HPV16 E6. However, despite robust interaction with E6AP_LDDLL, BPV1, and HPV types 4, 112, and 131 E6 did not reduce the expression of E6AP_LDDLL (Fig 8B). This demonstrates that while association of E6 with E6AP is required to initiate E6AP degradation, some difference in the binding or an additional property of Alpha genera E6 beyond simply associating with the LXXLL (that is not present in Delta or Gamma genera E6), is required to initiate degradation of E6AP. In either case, docking of an E6 protein at the E6AP LXXLL site was insufficient to initiate E6AP degradation.
HPV16 and BPV1 E6 both interact with extended LXXLL alpha-helical peptides, are stabilized by the association in vivo, and solubilized by the association in vitro, which led to their crystallization and structural determination [25, 41]. But significant questions remain. How do E6 proteins discriminate in interaction among many potential binding partners? Can we predict the interaction targets of different genera of E6 proteins? What are the additional structural features of E6 beyond interacting with a cellular target LXXLL docking site that determine the biological properties of E6? How are E6 proteins that associate with particular cellular proteins evolutionarily related to each other?
There are five genera of HPV. Our work demonstrates the Alpha genera is the sole genera where E6 targets E6AP while the remaining (Beta, Gamma, Nu, and Mu) preferentially target MAML1 or another protein(s) with a similar LXXLL motif as MAML1. While HPV41 E6 bound MAML1 preferentially, it was the only E6 protein that while preferentially binding MAML1 failed to repress MAML1 transactivation. It is possible that ancillary E6 structure(s) that may be important for transcriptional repression are not present in HPV41 E6, or that in the genera, E6 has different biological functions.
The crystal structures of BPV1 and HPV16 E6, together with sequence alignments of other E6 proteins predicted that other E6 proteins would also interact with acidic amphipathic alpha helical peptides [41]. The conservation of contact residues between BPV1 and HPV16 E6 and the LXXLL ligands, plus the recent discovery that MAML1 was a common target of E6 proteins from several genera, prompted us to speculate that the preferred LXXLL targets of diverse genera might be predictable from multiple sequence alignment data, and that MAML1 or MAML1-like targets would be the most common E6 binding targets of diverse genera of papillomaviruses. We have taken three separate approaches to test these hypotheses.
Examination of the crystal structures of BPV1 E6 plus the PXN LXXLL peptide, compared to HPV16 in complex with the LXXLL peptide of E6AP showed a distinctive difference to be hydrophobic BPV1 E6 contacts with M1 of the LXXLL PXN peptide compared to charged interactions between HPV16 E6 and glutamic acid at position -3 of the E6AP LXXLL peptide (Fig 2). To test the importance of this contact, we prospectively selected peptides from a large random peptide phage library using BPV1 E6 as a binding substrate, and found a strong selection for peptides that had a hydrophobic residue in the same position as M1 of the PXN or MAML1 peptides (Fig 1). This illustrates the importance of this hydrophobic interaction when peptides compete for interaction with BPV1 E6. In the comparison of HPV16 E6 and BPV1 E6, both experimental selection of peptides and examination of the structure highlighted the important interactions between E6 and position -3 of the LXXLL peptides. Because only two E6 structures have been determined so far; it is possible that differences between the E6AP and MAML1 peptides other than at position -3 are more important to other E6 proteins to discriminate between E6AP and MAML1 association.
In testing 45 E6 proteins for interaction with E6AP and MAML1, we were surprised that most E6 proteins had a clear binding preference, that few interacted with both, and then only at low levels, and that no E6 protein both stimulated E6AP degradation and repressed GAL4_MAML1 transcriptional activation (Fig 5). Why would an Alpha E6 protein not evolve to both interact with E6AP to target cellular proteins for degradation, and simultaneously also interact with MAML1 to repress the Notch tumor suppressor pathway in squamous epithelia (like a Beta or Gamma genus E6 protein)? Our experiments do not yet shed light on this question.
All E6 proteins that had a preference for association with E6AP also stimulated the degradation of E6AP, indicating the linkage of physical association with biological function. However, binding of an E6 protein to E6AP at the LXXLL site was not sufficient to initiate in vivo degradation of E6AP. This was demonstrated by the use of the E6AP_LDDLL chimera, where BPV1 E6 or Gamma genera HPV E6 (HPV types 4, 112 and 131) bound robustly to E6AP_Ub-_LDDLL yet failed to stimulate degradation of E6AP_LDDLL. In contrast, HPV16 E6 (which bound E6AP_Ub-_LDDLL relatively poorly) could still initiate degradation of E6AP_LDDLL, indicating the functionality of the E6AP_LDDLL mutant. Thus, while binding of an E6 protein to the LXXLL site of E6AP is necessary, it is not sufficient for degradation of E6AP (Fig 8). This implies that two properties of alpha genera E6, binding and initiation of degradation of E6AP, co-evolved as the Alpha and Alpha-clustered E6 proteins diverged from interaction with other LXXLL target proteins (possibly MAML1).
The Alpha-clustered genera we tested in this study, being from Alpha, DyoDelta, Dyopi, Omega, and Omikron genera, behaved similarly in binding to E6AP and stimulating its degradation. Phylogenetic clustering would suggest that Upsilon and Dyoomikron genera E6 proteins (that were not tested in this study) would also interact with E6AP and stimulate its degradation because they cluster together next to the Dyodelta, Omega, Omikron, and Alpha genera. These genera, where E6 proteins interact with E6AP and stimulate its degradation, are related in binding, in vivo biological effect, and phylogenetic clustering. We would propose that these genera should be clustered as a super-genera or clade in the description of papillomavirus oncoproteins. It is remarkable that it is not the presumed ultimate target of E6 action (such as p53, or type I cellular PDZ proteins [83] for the high-risk Alpha genera HPV E6) that phylogenetically clusters these diverse E6 proteins together, but is the common mechanism of MAML1 or E6AP association. While the activities of Alpha high-risk E6 proteins have been speculated to compensate for the activities of E7 (such as E7 activation of p53 leading to E6 mediated p53 degradation), within this larger super-genera, that is not the case, since the cetacean papillomaviruses (such as Omikron and Dyopipa genera) do not encode an E7 gene [5, 69], although additional E7-like functionality could be encoded elsewhere in the genome, such as within E6 or elsewhere. Earlier, it was customary in the field to think of papillomaviruses as either mucosal or cutaneous types. However within the Alpha genera, there are both cutaneous (like HPV types 7 and 10) and mucosal infections (like HPV types 11 and 16); similarly, both mucosal and cutaneous sites also harbor Beta and Gamma papillomaviruses [84, 85]. What is common to all Alpha compared to Beta or Gamma genera, is the preference of the E6 proteins for either E6AP or MAML1 association; this provides a simplified factor for categorizing many (but not all) papillomavirus types, compared to sequence divergence within L1 capsid sequence.
Our experiments suggest that E6 association with MAML1 is responsible for the clustering of most of the remaining E6 proteins that do not associate with E6AP. Only BPV1 E6 interacted with MAML3 in addition to MAML1, and only CPV7 E6 showed a preference for association of MAML2 compared to MAML1. To our surprise, rodent and canine MAML1 chimeras interacted with their cognate E6 proteins similarly to human MAML1, despite some amino acid differences in the LXXLL motif. This raises an unresolved question in how E6 association with MAML1 represses Notch signaling overall, since MAML1 can be deleted without ablating the gross development of a mouse. MAML1 -/- mice are born without gross developmental defects but are growth retarded and die at about 10 days (before weaning) and have lymphoid and muscle development defects [86, 87]. MAML3 null mice appear normal, but MAML3 combined with MAML1 deletion causes a severe defect in organogenesis similar to a Notch1 deficiency [88]. MAML2 null mice have not yet been described. Our results show that most E6 proteins interact with MAML1, with only BPV1 E6 interacting with both MAML1 and MAML3, and only CPV7 interacting preferentially with MAML2. MAML1, 2, and 3 are expressed in the skin at the RNA level by RNAseq analysis [89]. So how is it that E6 interaction with MAML1 represses Notch transcription overall when transcription complexes might be expected to alternatively contain MAML2 and/or MAML3 instead of MAML1? It is possible that individual MAML protein expression differs within the differentiating layers of squamous epithelium or that E6 may have a gain of function when bound to MAML1-Notch complexes to repress Notch signaling overall by an as yet uncharacterized mechanism, rather than acting upon MAML1 alone.
The E6 proteins that interacted poorly with either E6AP or MAML1, 2, or 3, and failed to either promote E6AP degradation or repress GAL4-MAML1 did not cluster with those that targeted either MAML1 or E6AP. OcPV1, similar to SfPV1 (also known as CRPV1 or Shope papillomavirus) has a different structure than most E6 proteins with three zinc structured domains, so it is not surprising that it does not share interaction targets with other canonical E6 proteins. The other non-E6AP/non-MAML1 interacting E6 proteins like TmPV1, TmPV2, RaPV1, and BpPV1 may have different LXXLL binding sites on cellular targets that remain to be determined. Alternatively, a technical reason such as incompatibility with the amino-terminal FLAG epitope tag may have prevented some of the tested E6 proteins from interacting well with either E6AP or MAML1.
Our results implicate E6 targeting of MAML1 in animal squamous cell carcinomas. Several of the MAML1-directed E6 proteins in this study were isolated from papillomaviruses in association with squamous cell cancers: OaPV3, CPV2, CPV3, CPV7, FcaPV2, and McPV2. OaPV3, CPV2, CPV3, and FcaPV2 all interact with MAML proteins, repress MAML1 transcriptional activation, and did not stimulate E6AP degradation. CPV7 and McPV2 did not bind MAML1 well by immunoprecipitation, but did repress GAL4_MAML1 transcriptional activation (Fig 7). Complete disruption of Notch signaling during development of the squamous epithelium of transgenic mice by either tissue specific Notch deletion [90, 91], epithelial deletion of the RPB/j -binding subunit of the Notch transcription complex [92, 93], or expression of a dominant negative MAML1 [94] all result in loss of squamous differentiation and squamous cell carcinomas. These results demonstrate that Notch signaling is a tumor suppressor in normal squamous epithelium, and the critical role of MAML1 in this process, making the targeting of MAML1 by these animal E6 proteins plausibly related to carcinogenesis. While these few papillomaviruses were associated with squamous cell cancers, most MAML1-targeting E6 proteins were not cancer-associated. This could reflect a low frequency of cancer development (similar to high-risk HPV types), the relative potency of different E6 proteins in repression of Notch, the possible differential sensitivity of developing compared to adult epithelia to Notch signaling disruption, and the role of E7 and or other oncoproteins in these animal papillomaviruses. Among the HPVs, Beta genera E6 proteins repress MAML1, but the consistent and continued expression of Beta or Gamma genus viral oncogenes in cutaneous squamous cell carcinomas has not been observed. A recent publication on Gamma genus HPV found the presence of HPV197 DNA associated with cutaneous squamous cell cancers, but the expression and role of viral early gene products in the cancers is as yet uncharacterized [3]. The role of Notch signaling in high-risk HPV infections and progression to cervical cancer has been contentious. Initially, Notch signaling was thought to be synergistically oncogenic with E6 and E7 in the development of cervical cancer, and experiments were presented demonstrating overexpression of Notch and synergistic transformation with papillomavirus oncogenes [95] which was then contradicted by different experimental approaches [96]. A study from the Kast lab [97] showed expression of high-risk E6 and E7 augmenting Notch1 expression while a recent study from the Doorbar lab [98] shows E6 repression of Notch1 expression through degradation of p53. While varying expression and cell culture systems have given rise to different results, head and neck cancers accumulate mutations in Notch signaling genes [99], and p53 null mice have normal appearing skin after development.
The phylogenetic alignment of E6 proteins shows a clear split between genera whose E6 proteins target E6AP and the genera that target MAML1. Consideration of the host species and the E6 sequence divergence give insight into when this occurred.
While some fish (SaPV1) and bird papillomaviruses express E6 proteins, these proteins are shorter and have only a single zinc structured domain, thus differing in structure from canonical E6 proteins with two zinc-structured domains that were tested here. It seems unlikely that fish and bird E6 proteins form a similar pocket for interaction with LXXLL motifs and as yet we have not tested these E6 proteins for either LXXLL interactions or modulation of Notch signaling. There is a single papillomavirus isolate from a marsupial (BpPV1) where the E6 protein has a canonical structure, but unfortunately our results with this E6 protein were ambiguous, binding both E6AP and MAML1 poorly, and neither significantly stimulating E6AP degradation nor repressing MAML1 transactivation (Fig 7). The Eutheria (placental mammals), diverged into two categories approximately 100M years ago: the Laurasiatheria (shrews, bats, ungulates, and cetaceans) and the Euarchontoglires (tree shrews, rodents, primates). Since E6 proteins that target E6AP have hosts within both of these groups of mammals, it is reasonable to speculate that the divergence of E6AP-directed E6 proteins must have predated the split between these two groups of Eutheria. Additional samples of papillomaviruses from marsupials, shrews, and rodents may provide further insight into the split between E6AP and MAML directed E6 proteins.
Among animal models of high-risk HPV infections, we show here that the Rhesus papillomavirus (Macaca mulatta Papillomavirus type 1, MmPV1, also known as RhPV1) E6 protein associates with E6AP, targets p53 for degradation (Fig 7 and S7 Fig), and causes genital squamous cell cancers, making this an excellent animal model of high-risk HPV disease [76]. Other animal papillomavirus infections do not model the activities of the Alpha genera high-risk HPV E6 and E7 proteins. Our findings suggest the possibility that additional sampling of rodents might reveal a papillomavirus whose E6 protein preferentially targets E6AP, which would apply the power of mouse genetic analysis to the activities of viral oncoproteins that are relevant to lethal human papillomavirus infections.
E6AP-null 8B9 cells (A gift of Dr. Lawrence Banks, ICGEB, Italy) [100], and 293T cells (American Type Culture Collection) were maintained in DMEM media supplemented with 10% fetal bovine serum, glutamine and antibiotics. Transient transfections were performed using polyethylenimine (PEI). Transient Vaccinia virus expression of proteins was performed in CV-1 cells (American Type Culture Collection) as described [101].
E6 genes were synthesized (Genewiz Corp.) using native unmodified codons, terminated at the native stop codon, and were cloned as amino-terminal 2X-FLAG-tagged fusions in pcDNA3. HA-tagged E6AP WT and E6AP containing a C843A mutation that eliminates ubiquitin ligase activity (E6AP_Ub-), HA-tagged MAML1, and GAL4-MAML1 fusions have been previously described [34]. cDNA’s for human MAML2 and MAML3 were generously provided by Brandon White (San Jose State University). Epitope-tagged E6AP, GFP, MAML, and E6 were all expressed from the pcDNA3 plasmid.
12-well plates of subconfluent HEK293T cells were transfected with 0.5 ug HA-tagged E6AP or E6AP-Ub-, 2 ug HA-MAML1, 1 ug FLAG-E6, and 0.02 ug HA-GFP (as an internal transfection control) with a total of 3.4 ug of DNA complexed with 4 ug of polyethylenemine (PEI). Eighteen hrs. later, cells were lysed on ice in 0.5 mls of 0.5X IGEPAL lysis buffer (1X IGEPAL lysis buffer contains contains 150 mM NaCl; 50 mM Tris pH 7.5; 0.5 mM Dithiothreitol (DTT); 50 mM NaF; 5 mM NaPPi; 1% IGEPAL; 0.01% phenylmethylsulfonyl flouride; 1mM sodium vanadate; 1ug/ml leupeptin/aprotinin), pelleted at 15,000 X G at 4°C for 15 min. The supernatant was immune precipitated with 10 ul anti-FLAG M2 antibody coupled beads (Sigma) for 1 hr. and then beads were washed three times with ice-cold 1.0 ml 0.5X lysis buffer. Samples were eluted with SDS and resolved on SDS-PAGE gels, blotted to PVDF membranes and probed sequentially with rabbit anti-HA antibody, and rabbit anti-FLAG antibodies (Sigma). Bound antibodies on the blot were detected with secondary antibodies coupled to horseradish peroxidase and chemiluminescent substrates, and images captured and analyzed using an Alpha-Innotech Fluorchem detector and associated image analysis software. For quantitation, input samples and immune precipitations were run in adjacent lanes for each transfection, and immune precipitations calculated as percent of input signal and then normalized to the input HA-GFP signal. Statistical analysis of normalized, independently replicated experiments was performed with Excel software. For E6AP degradation assays, 0.5 ug HA-E6AP, 1 ug FLAG-E6 proteins, 0.02 ug HA-GFP and 0.5 ug of empty pcDNA3 (total of 2 ug) were complexed with 3 ug PEI and transfected into HEK293T cells and harvested by denaturing lysis 18 hrs. post-transfection.
0.1 ug of GAL4_MAML1, 0.2 ug of FLAG_E6, 0.05 ng GAL4-responsive luciferase reporter, 0.01 ug pCMV renilla internal transfection control plus 0.64 ug pUC19 to a total of 1.0 ug DNA complexed with 3 ug PEI, was transfected into HEK293T cells and harvested 18 hrs. later. Duplicate wells for each sample were used in each assay and the assay was independently repeated 4 times for statistical analysis.
Glutathione S-Transferase (GST) or Chitin binding domain (CBD) were fused at the carboxy-terminus to a TEV protease cleavage site, a FLAG epitope and then BPV1 E6; the fusion protein was expressed overnight in CV-1 cells using the pTM1 vaccinia virus expression system [101]. Either Glutathione-agarose or Chitin-agarose beads were blocked overnight in 2% BSA in PBS with 0.3% IGEPAL. An aliquot of an M13 phage display library expressing random 12-mer peptides as fusions to the phage pIII gene (Ph.D.-12 library, New England Biolabs) containing 1.5 X 1011 phage was blocked overnight on ice in Luria Broth containing 1% BSA in PBS with 0.1% IGEPAL. A 6 cm plate of CV1 cells expressing the GST-BPV1 E6 or CBD-BPV1 E6 fusion proteins were lysed on ice in 250 ul 0.1 X IGEPAL lysis buffer containing 1% BSA. Cell lysates were clarified by centrifugation (15,000 XG for 30 min), and the supernatant applied to 10 ul of blocked glutathione-agarose or chitin-agarose beads rocking for 30 min. Beads were then washed three times with 1X Luria Broth containing 1% BSA, 0.1% IGEPAL, and 3 mM DTT, and phage applied in 100 ul Luria Broth containing 1% BSA, and 0.1% IGEPAL at 4°C for 1 hr. Beads were washed 6X with 1.0 ml of binding buffer, 1X with TEV protease cleavage buffer, transferred to a new tube and then cleaved with 5 U TEV protease in 25 ul of buffer (Invitrogen). Released phage were amplified and purified by PEG precipitation in XL1-blue bacteria according to the manufacturer’s instructions. Five rounds of phage display selection were performed starting with CBD as the fusion partner alternating with GST at each selection round. Individual phage were picked at the fifth round of selection and eluted, amplified, phage DNA isolated, and the pIII-peptide fusion segment DNA-sequenced.
Multiple protein sequence files were downloaded from the papillomavirus episteme [102], alignments were performed with the MUSCLE 3.8.31 program [103] using the phylogeny.fr toolset [104]. The output from MUSCLE was used as input for phylogenetic analysis and graphing using PhyML 3.1/3.0 aLRT [105], and tree rendering with TreeDyn 198.3 [106].
|
10.1371/journal.pcbi.1000583 | Oxidized Calmodulin Kinase II Regulates Conduction Following Myocardial Infarction: A Computational Analysis | Calmodulin kinase II (CaMKII) mediates critical signaling pathways responsible for divergent functions in the heart including calcium cycling, hypertrophy and apoptosis. Dysfunction in the CaMKII signaling pathway occurs in heart disease and is associated with increased susceptibility to life-threatening arrhythmia. Furthermore, CaMKII inhibition prevents cardiac arrhythmia and improves heart function following myocardial infarction. Recently, a novel mechanism for oxidative CaMKII activation was discovered in the heart. Here, we provide the first report of CaMKII oxidation state in a well-validated, large-animal model of heart disease. Specifically, we observe increased levels of oxidized CaMKII in the infarct border zone (BZ). These unexpected new data identify an alternative activation pathway for CaMKII in common cardiovascular disease. To study the role of oxidation-dependent CaMKII activation in creating a pro-arrhythmia substrate following myocardial infarction, we developed a new mathematical model of CaMKII activity including both oxidative and autophosphorylation activation pathways. Computer simulations using a multicellular mathematical model of the cardiac fiber demonstrate that enhanced CaMKII activity in the infarct BZ, due primarily to increased oxidation, is associated with reduced conduction velocity, increased effective refractory period, and increased susceptibility to formation of conduction block at the BZ margin, a prerequisite for reentry. Furthermore, our model predicts that CaMKII inhibition improves conduction and reduces refractoriness in the BZ, thereby reducing vulnerability to conduction block and reentry. These results identify a novel oxidation-dependent pathway for CaMKII activation in the infarct BZ that may be an effective therapeutic target for improving conduction and reducing heterogeneity in the infarcted heart.
| Calmodulin kinase II (CaMKII) is a multifunctional serine/threonine kinase that regulates diverse functions in heart. Recently, a novel pathway for CaMKII activation was discovered where oxidation of the kinase at specific methionine residues produces persistent activity. This alternative oxidation-dependent pathway has important implications for heart disease where oxidative stress is increased (e.g., heart failure and following myocardial infarction). We hypothesized that myocardial infarction caused by occlusion of a coronary artery would increase levels of oxidized CaMKII. Moreover, we hypothesized that oxidative CaMKII activation represents an important mechanistic link between increased oxidative stress and life-threatening heart rhythm disturbances (arrhythmias) in heart disease. We report a dramatic increase in levels of oxidized CaMKII following myocardial infarction in the canine. Based on these experimental data, we developed a novel mathematical model of CaMKII activity to study the role of oxidation-dependent CaMKII activation in regulating cardiac cell excitability. Our findings identify a novel role for oxidation-dependent CaMKII activation following myocardial infarction and provide a mechanistic link between oxidative stress and lethal cardiac arrhythmias in heart disease.
| Calmodulin kinase II (CaMKII) mediates diverse roles in the heart, including excitation-contraction coupling, sinus node automaticity, apoptosis, hypertrophy, and gene transcription [1],[2]. Mounting experimental evidence demonstrates an important role for CaMKII in heart disease and arrhythmias. Specifically, CaMKII overexpression occurs in human heart failure [3] and transgenic mice overexpressing CaMKII develop dilated cardiomyopathy [4],[5]. Conversely, transgenic inhibition of CaMKII prevents structural remodeling and improves heart function following myocardial infarction (MI) [6] while knockout mice lacking the predominant cardiac CaMKII isoform (CaMKIIδ) are resistant to development of pressure overload-induced hypertrophy and/or heart failure [7],[8]. Finally, CaMKII inhibition prevents arrhythmias in several different mouse models of heart disease [9],[10].
CaMKII is activated by binding of Ca2+/calmodulin and may undergo inter-subunit autophosphorylation that allows the kinase to retain activity even upon dissociation of Ca2+/calmodulin (autonomy) [11]. Recently, a novel CaMKII activation pathway was identified where oxidation at specific methionine residues in the CaMKII regulatory subunit results in persistent activity independent of autophosphorylation [12]. While oxidative-dependent CaMKII activation has been shown to mediate apoptosis in response to chronic AngII treatment in the mouse [12] as well as arrhythmogenic afterdepolarizations in isolated cardiomyocytes treated with hydrogen peroxide [13], nothing is known about its role in large animal models of heart disease. Considering that levels of reactive oxygen species (ROS) such as H2O2 and superoxide are elevated following myocardial infarction [14], we hypothesized that oxidation of CaMKII represents an important pathway for CaMKII activation in the infarct border zone (BZ) that may provide a mechanistic link between increased ROS production, Na+ channel remodeling and conduction slowing following MI.
In this study, we describe a dramatic increase in levels of oxidized CaMKII in a well-validated large animal model of arrhythmias following MI [15]–[22]. To investigate a role for oxidized CaMKII in regulating refractoriness and conduction in the infarct BZ, we develop a novel mathematical model of CaMKII activity that includes oxidation and autophosphorylation activation pathways. Our computer simulations show that enhanced CaMKII activity in the BZ, due primarily to increased oxidation, leads to slowed conduction, prolonged refractory periods and increased vulnerability to conduction block at the BZ margin (a prerequisite for reentry initiation). Our results identify oxidation-dependent CaMKII activation as a potential link between oxidative stress and electrical remodeling after myocardial infarction. Furthermore, our findings support CaMKII inhibition as a potential therapy for reducing susceptibility to ventricular tachycardia by improving conduction and reducing refractory gradients in the infarcted heart. Finally, it is important to note the oxidative activation of CaMKII allows for independent regulation of the kinase by a host of unique upstream activators and signaling partners (e.g. oxidases/reductases) with great potential relevance to human disease. As details emerge regarding regulation of the kinase by this newly identified pathway, they may be incorporated into our model to study electrophysiological consequences of CaMKII activation via this independent signaling pathway.
Based on the recent discovery of a novel oxidation-dependent pathway for CaMKII activation [12], immunoblot analysis was first performed in a well-validated large animal model of arrhythmias [15]–[22] to determine whether oxidization of CaMKII occurs in the infarct BZ five days post-occlusion (Figure 1). Interestingly, levels of oxidized CaMKII were over eight-fold greater in the five-day BZ compared to normal (non-infarcted) (p<0.01), but were unchanged in remote regions of the same hearts (Figure 1, p = NS vs. normal). These data together with our previous findings that CaMKII autophosphorylation is significantly increased in the five-day infarct BZ [20] indicate that CaMKII activity is enhanced in the infarct BZ.
To determine whether enhanced CaMKII activity, due in part to oxidation (Figure 1), regulates conduction in the infarct BZ, we revised our model of the canine ventricular action potential [20],[24] to include a new model of CaMKII activity based on the simplified scheme proposed by Dupont et al [26]–[28] (Figure 2). Importantly, our model includes an oxidized active state in addition to a Ca2+/CaM bound active state and an autophosphorylated active state. Inclusion of an additional autonomous active state (Ca2+/CaM dissociates from phosphorylated subunit) was found to have no impact on model behavior (state occupancy <0.001%, not shown) and was therefore not included in the final model. Consistent with experimental observations [12], Ca2+/CaM must bind to a subunit before oxidation may occur (no direct transition from inactive to oxidized active state). Rate constants for state transitions were taken from the literature or chosen to fit experimental data (Table S3, Figure 2B–D). Model equations are provided in supplementary information (Text S1). Our experimental data demonstrate a significant increase in both oxidized (Figure 1) and autophosphorylated CaMKII [20] in the infarct BZ. Even though autophosphorylation and oxidation occur through distinct pathways, the model assumes that the same subunit may be both oxidized and autophosphorylated. Furthermore, consistent with previous work [26]–[28], the model assumes that any active subunit (including oxidized) may autophosphorylate another Ca2+/CaM bound subunit. Thus, the model predicts a secondary increase in the fraction of autophosphorylated CaMKII subunits with an increase in oxidized subunits due to oxidative stress (Figure 2E). Currently, the upstream pathways responsible for increased CaMKII autophosphorylation are unknown. However, our model predicts that oxidative stress may account for the increase in both oxidized and autophosphorylated subunits measured in the infarct BZ (Figure 2E). Thus, for the purpose of this study, we assume that the primary defect responsible for activated CaMKII in the BZ is oxidative stress.
While absolute measures of H2O2 levels are limited (likely less than 0.25 µM at baseline [29],[30]), an increase in ROS levels from 10- to 100-fold have been reported following ischemia-reperfusion [31]–[33]. Furthermore, ROS levels of 10 µM in vitro have been shown to recapitulate the level of oxidative stress observed in vivo in the BZ [34]. Unless otherwise stated, we assume [ROS] = 1.0 µM in the BZ, likely a conservative estimate. However, based on the fact that the exact level of ROS is unknown in the BZ and is likely to be highly heterogeneous, we also explore a range of ROS levels from 0 to 10 µM. Note that for [ROS] = 1 µM, the fraction of autophosphorylated subunits in the BZ is much lower than the fraction of oxidized subunits (0.11 for autophosphorylation compared to 0.75 for oxidation, Figure 2E), indicating that oxidation rather than autophosphorylation is the primary determinant of increased CaMKII activity in the BZ model.
Importantly, our model of the BZ myocyte also accounts for observed remodeling changes to the density and/or kinetics of several ion channels, including the L-type Ca2+ current, transient outward K+ current, and Na+ current [20]. Specifically for the Na+ current, changes to kinetics and peak current have been observed [22]. Since CaMKII has been shown to alter Na+ channel kinetics but not peak current, our model assumes that the reduction in total Na+ channel density occurs through a CaMKII-independent pathway [20]. NZ (control) and BZ cell models are incorporated into one-dimensional fibers to study conduction (Figure 3).
INa inactivation and recovery from inactivation were first determined in the NZ and BZ fiber after pacing to steady state (Figure 4). INa recovery from inactivation was determined by applying a premature stimulus (S2) at a varying S1S2 interval and plotting channel availability (calculated as product of inactivation gates, h*j) vs. recovery interval (S1S2 interval - APD90). INa steady-state inactivation is shifted to more hyperpolarized potentials (Figure 4A) and recovery from inactivation is slower (Figure 4C) in the BZ fiber compared to NZ, consistent with our single cell simulations [20] and experimental measurements [22]. CaMKII inhibition (CaMKII activity held constant at zero) shifts INa steady state inactivation to more depolarized potentials (Figure 4A) and accelerates recovery from inactivation in the BZ fiber (Figure 4C) but has little effect in NZ (steady-state inactivation and recovery curves superimpose curves from BZ+CaMKII inhibition model, not shown). Thus differences in INa inactivation between NZ and BZ observed under control conditions are largely eliminated upon CaMKII inhibition (Figure 4).
Based on the effects of CaMKII on INa availability, we hypothesized that enhanced CaMKII activity would promote slow conduction in the BZ. While resting transmembrane potential is comparable between isolated BZ and NZ myocytes [19], membrane depolarization is observed in multicellular BZ preparations [35],[36]. Therefore, we measured conduction velocity in NZ and BZ fibers over a range of end diastolic potentials (Vm,dia, −86 to −63 mV), by increasing [K+]o incrementally from 5.4 to 13 mM. Conduction velocity was measured across the central 100 cells (Figure 5). Conduction was dramatically slower at every Vm,dia in the BZ compared to NZ (65–100% slower) (Figure 5A). Furthermore, while successful conduction was observed in the NZ for potentials up to −64 mV, conduction block occurred in the BZ for Vm,dia >−72 mV. In fact, conduction velocity is steeply dependent on the concentration of ROS in the BZ over a range of concentrations from about 0.01 µM to 1 µM (Figure S1). To determine the role of oxidation-dependent CaMKII activity in conduction slowing in the BZ, we measured conduction velocity in the BZ model resistant to CaMKII oxidation (CaMKIIox = 0). Making CaMKII resistant to oxidation increased conduction velocity at all Vm,dia in the BZ with a greater effect at more depolarized Vm,dia (Figure 5A). Moreover, successful conduction was restored in the BZ for Vm,dia up to −68 mV (compared to −64 mV in the control). In contrast, the BZ model resistant to CaMKII autophosphorylation showed very little improvement in conduction (Figure S2). Furthermore, inhibiting total CaMKII activity showed a similar improvement in conduction as the oxidation-resistant model (Figure 5B), indicating that oxidation is the primary determinant of enhanced kinase activity in our BZ model. To verify that CaMKII-dependent effects on conduction were mediated by regulation of INa kinetics, we also calculated conduction velocity in the BZ model with INa resistant to CaMKII phosphorylation. As expected, this model showed a similar improvement in conduction as the oxidation-resistant model (Figure 5B), indicating that enhanced CaMKII activity regulates conduction by altering INa kinetics. In summary, enhanced CaMKII activity contributes to reduced conduction velocity in the BZ fiber, even promoting conduction block in the setting of depolarized transmembrane potential.
Remodeling in BZ tissue involves not only changes to ion channel properties and rest potential, but also intracellular communication [37],[38]. In order to address whether cellular uncoupling affects the role of oxidized CaMKII in regulating conduction, we determined conduction velocity in the fiber over a range of gap junction resistances (Figure 5C). Increasing gap junctional resistance (Rg) from 1.5 to 60 Ωcm2 produced a similar decrease in conduction velocity in NZ and BZ fibers (86% and 90% decrease, respectively). Successful conduction occurred in the NZ fiber for Rg up to 300 Ωcm2, while conduction block was observed in the BZ fiber for Rg>76 Ωcm2. Eliminating oxidation-dependent CaMKII activity increased conduction velocity in the BZ fiber at all Rg and restored conduction for Rg up to 150 Ωcm2 (Figure 5C), indicating that CaMKII regulates conduction even in the setting of gap junction uncoupling.
Effective refractory period (ERP) of the action potential is dramatically prolonged in BZ compared to NZ, despite comparable action potential durations [18],[19],[35]. Moreover, large gradients in refractoriness at the BZ margin have been associated with conduction block and the initiation of reentrant arrhythmias [15],[17],[18]. Based on these data and the ability of CaMKII to regulate INa recovery from inactivation (Figure 4C), we hypothesized that enhanced CaMKII activity would contribute to prolonged refractoriness in the BZ. To test our hypothesis, ERP was determined in NZ and BZ fibers by applying a premature (S2) stimulus during the repolarization phase of the action potential at cell 1 (site of S1 stimulus). The S1S2 interval was increased until a second propagating wave was generated in the wake of the final S1 stimulated AP (Figure 6). ERP is defined as the largest S1S2 interval that fails to generate a propagating excitation wave and is a function of both action potential duration (APD) and postrepolarization refractoriness. Consistent with experiment [18],[19],[35], ERP is much greater in the BZ model (213 ms compared to 181 in the NZ) (Figure 6). Small differences in APD (173 ms and 187 ms in NZ and BZ, respectively) account for only a portion of this difference in ERP. Rather the primary determinant of prolonged ERP in the BZ is increased postrepolarization refractoriness due to the much slower time course of recovery from inactivation of INa (Figure 4C). Making CaMKII resistant to oxidative activation reduces ERP to 207 ms in the BZ model despite a slight prolongation of APD (Figure 6C) by eliminating differences in postrepolarization refractoriness (measured as ERP – APD, Figure 6D). Likewise, total CaMKII inhibition and making INa resistant to CaMKII phosphorylation reduce ERP by normalizing postrepolarization refractoriness (Figure 6D). These results suggest that oxidation-dependent CaMKII activation contributes to large gradients of refractoriness, particularly at the margins of the infarct BZ, by regulating INa kinetics.
Electrophysiological mapping during programmed stimulation to induce ventricular tachycardia has revealed that premature excitation block occurs in areas of large refractory gradients at the BZ margin [18]. Our findings that enhanced CaMKII activity substantially increases ERP in the BZ led us to hypothesize that CaMKII promotes formation of conduction block at the transition between normal and border zone tissue by introducing large refractory gradients. To test our hypothesis, we used a heterogeneous fiber comprised of coupled NZ (cells 1–75) and BZ (cells 126–200) cells with a central transitional region (cells 76–125) across which BZ parameters were linearly scaled. The size of the transitional region corresponds to the approximate width of the outer common pathway (about 5.0 mm [15]) (Figure 7A). The fiber was paced to steady state at cell 1 and a premature S2 stimulus was applied at the same cell. The S1S2 pacing interval was varied to determine the critical range of S1S2 intervals (vulnerable window, VW) that resulted in conduction block at the transition from the NZ into the BZ region. S1S2 intervals from 181 to 197 ms (VW = 18 ms) resulted in an action potential that propagated successfully through the NZ region but failed at the transition into the BZ region (Figure 7A). S1S2 pacing intervals shorter than this critical range failed to propagate even through the NZ region while S1S2 intervals greater than this range successfully propagated through the entire fiber (Figure 7B). Vm and INa availability (h*j) spatial profiles as the wavefront reaches the BZ margin indicate that INa availability is a critical determinant of whether or not a premature wavefront blocks at the BZ margin (Figure 7C).
Figure 8 shows VW in the heterogeneous fiber as a function of Vm,dia in the BZ region ([K+]o was scaled linearly across the transition region as with other BZ parameters). VW shows a monophasic increase as Vm,dia is depolarized from −85 mV to −72 mV. Vm,dia greater than −72 mV results in transient block at the BZ margin even at the basic cycle length of 500 ms. Making CaMKII resistant to oxidative activation greatly reduces VW at all Vm,dia (Figure 8A). Furthermore, transient block is not observed at the basic cycle length until Vm,dia is depolarized above −68 mV. Total CaMKII inhibition and making INa resistant to CaMKII phosphorylation also effectively prevented formation of block at the BZ margin (VW less than 1 ms for Vm,dia up to −72 mV and −74 mV, respectively, not shown).
While these data show that oxidation-dependent CaMKII activation increases the vulnerability to conduction block at the BZ margin, a prerequisite for initiation of reentrant arrhythmias, it is important to note that CaMKII-independent remodeling of ion channels (notably INa) also likely play an important role. To address the role of CaMKII-independent remodeling in conduction block, we measured VW in the heterogeneous fiber with normal INa conductance throughout. Eliminating differences in INa conductance successfully reduced vulnerability to transient conduction block across a wide range of Vm,dia (Figure 8A), indicating that both CaMKII-dependent (altered kinetics) and CaMKII-independent (reduced channel conductance) effects on INa increase the vulnerability to conduction block and reentrant arrhythmias at the BZ margin. Interestingly, eliminating either CaMKII-dependent (oxidation-resistant CaMKII) or CaMKII-independent (normal ) heterogeneities in the fiber resulted in a similar increase in the effective INa availability () in the transition region (Figure 8B).
We next determined whether cellular uncoupling in the BZ region would alter the role of CaMKII-dependent or CaMKII-independent INa changes in formation of conduction block at the BZ margin (Figure 8C). Increasing Rg in the BZ region (Rg scaled linearly across the transition region as with other BZ parameters) had very little effect on VW for a moderate degree of uncoupling (Rg up to 38 Ωcm2). However, VW increased sharply for Rg>38 Ωcm2 until block occurred at the BZ margin even at the basic cycle length (Rg>60 Ωcm2). Eliminating oxidation-dependent CaMKII activation or normalizing INa conductance reduced VW at all Rg (VW = 0 ms for Rg up to 120 Ωcm2 and 150 Ωcm2, respectively) indicating that CaMKII-dependent and CaMKII-independent effects on INa regulate VW even in the presence of cellular uncoupling.
Our data provide the first evidence for oxidation of CaMKII as an important component of the remodeling process following MI. Furthermore, our simulation results provide the following insight into regulation of CaMKII signaling by this novel oxidative pathway: 1) Significant oxidative activation of the kianse occurs under pathophysiological conditions; 2) Oxidative stress may activate the kinase not only through direct oxidation but also through a secondary increase in autophosphorylation; and 3) Changes in Na+ channel kinetics due to oxidative CaMKII activation are sufficient to impact conduction in the BZ.
Conduction in the canine infarct border zone is highly irregular with areas of very slow and discontinuous conduction during sinus rhythm [17]. During programmed stimulation, lines of conduction block often form transverse to the fiber axis leading to initiation of reentry and sustained ventricular tachycardia [15],[16]. The mechanisms responsible for conduction block and reentry in the BZ are unknown although remodeling changes in tissue refractoriness and electrical coupling most likely play important roles [15]–[18]. At the cellular level, it is clear that Na+ channel dysfunction contributes to reduced action potential upstroke and action potential amplitude in myocytes isolated from the infarct border zone [19],[22]. Previous modeling studies have shown that decreased Na+ channel availability contributes to prolonged refractoriness in BZ cells despite AP duration comparable to NZ [39] and that slowing of Na+ channel recovery from inactivation or reducing Na+ channel conductance increases the vulnerable period for unidirectional block in cardiac tissue [40]. Our simulations demonstrate that CaMKII may regulate both conduction velocity and refractoriness in the BZ through its effects on voltage-gated Na+ channel kinetics. Furthermore, by introducing gradients in INa availability and refractoriness, CaMKII activation, due in part to oxidation, increases vulnerability to conduction block at the BZ margin, a prerequisite for reentrant excitation. Moreover, our simulations suggest that CaMKII inhibition improves conduction (particularly in depolarized tissue) and reduces ERP in the BZ, thereby reducing the risk for conduction block and reentrant excitation. These results are significant in light of experimental mapping studies showing premature excitation block in areas of large gradients in refractoriness [18]. While the current study focuses on conduction defects in the BZ, it is important to note that CaMKII activation is also expected to regulate intracellular Ca2+ cycling that itself may promote arrhythmias [2].
Our findings regarding the effects of CaMKII on conduction are consistent with experimental studies in mice that over-express CaMKIIδ. Specifically, consistent with our simulations, CaMKIIδ over-expression results in prolonged QRS intervals (marker of slowed intraventricular conduction) and increased arrhythmia susceptibility [41]. In contrast, our finding that CaMKII acts to increase post-repolarization refractoriness and therefore ERP in the BZ does not agree with shorter refractory periods in CaMKIIδ mice [41]. While the nature of this discrepancy is unclear, it is difficult to reconcile the reported effects of CaMKII on Na+ channel recovery (slowing) with the measured effects on refractoriness in transgenic mice. It is possible that the decrease in ERP measured in transgenic mice is due to secondary effects of chronic CaMKII over-expression rather than acute signaling effects. Regardless, further studies are needed to define the role of CaMKII in regulating refractoriness in the heart.
Clearly, many factors besides remodeling of voltage-gated Na+ channels influence conduction in the infarct BZ. Specifically, alterations in cell-to-cell coupling due to gap junction remodeling and/or fibrosis undoubtedly play an important role in abnormal conduction. In fact, studies have shown a close correlation between location of the central common pathway of the reentrant circuit and connexin43 redistribution suggesting that gap junction remodeling is required for maintenance of reentrant excitation [37],[38]. Furthermore, preferential uncoupling along the transverse fiber axis is thought to increase the degree of anisotropy in the BZ and facilitate initiation and maintenance of reentry. In addition, remodeling of the extracellular space in the BZ likely interacts with changes in gap junction coupling to affect conduction [42]. Thus, cell communication is regulated by a complex set of ion channel and structural changes in the BZ. Importantly, we find that CaMKII-dependent changes in INa kinetics regulate conduction in the BZ independent of the degree of cell coupling. Moreover, we report that CaMKII inhibition may restore conduction in the BZ even in the setting of very poor coupling (Rg up to 150 Ωcm2).
Previous studies have shown increased levels of ROS in five-day infarct BZ regions [34]. Moreover, exposure of cardiac Na+ channels expressed in HEK cells to ROS recapitulates the remodeling phenotypes observed in BZ myocytes. Our results suggest that the newly discovered oxidation-dependent pathway for CaMKII activation serves as a critical link between oxidative stress, enhanced CaMKII activity, Na+ channel dysfunction, and abnormal conduction in the infarct BZ. Of course, CaMKII is unlikely to be the only pathway through which oxidative stress alters cell excitability as oxidation affects many proteins in the heart, including kinases, transcription factors, ion channels, pumps, transporters, Ca2+ handling proteins, and contractile machinery [43].
While studies from our group and others demonstrate an important role for activated CaMKII in remodeling following MI, the upstream signaling pathways responsible for enhanced CaMKII activity remain to be fully elucidated (Figure 8). In this study, we assume that oxidative stress is the primary cause of enhanced CaMKII activity through direct oxidation of the kinase that also produces a secondary increase in the fraction of autophosphorylated subunits (Figure 2E). Clearly, CaMKII oxidation downstream of increased ROS production is one possible pathway for CaMKII activation in the BZ (Figure 1). However, a number of other upstream factors likely play an important role in regulating CaMKII activity following MI. For example, β-adrenergic stimulation, observed in the setting of myocardial infarction, activates CaMKII [6] and may also contribute to electrical remodeling after MI. Another possible mechanism for dysfunction in the CaMKII signaling pathway involves loss of coordinate regulation by phosphatases. Recently, it was discovered that miR-1 over-expression causes CaMKII-dependent hyperphosphorylation of RyR2 and afterdepolarizations due to reduced expression of the B56α regulatory subunit of the serine/threonine protein phosphatase 2A, PP2A [44]. Previous studies from our group have shown that B56α binds to, and is targeted by the adapter protein ankyrin-B in heart [45],[46]. Furthermore, B56α expression is reduced in cardiomyocytes lacking ankyrin-B [45]. More recently, we have shown that expression levels of ankyrin-B are significantly reduced in the BZ leading to altered expression and distribution of ankyrin-B associated membrane proteins including PP2A [21]. Interestingly, previous modeling studies have shown that loss of local phosphatase signaling may greatly potentiate levels of autophosphorylated CaMKII [27]. Therefore, loss of ankyrin-B may provide another mechanism for abnormal CaMKII signaling in the BZ through abnormal localization and/or activity of PP2A. Interestingly, patients with ankyrin mutations show catecholaminergic-induced afterdepolarizations [47]–[49] as observed in cells with reduced B56α. Future studies are needed to define the upstream signaling pathway(s) responsible for CaMKII activation, Na+ channel remodeling and increased susceptibility to reentrant arrhythmias after MI.
Finally, it is important to note that activation of CaMKII through direct oxidation has only recently been discovered. Consequently, much remains unknown regarding the signaling mechanisms that regulate this pathway. Analogous to regulation of autophosphorylation by local kinase/phosphatase activity and local concentrations of Ca2+/calmodulin, oxidative activation is likely controlled by a delicate balance of oxidase/reductase activity, mitochondrial function and local calcium signaling. Furthermore, degree of crosstalk between oxidative and autophosphorylation pathways and their relative importance in response to the complex set of upstream stressors in heart disease remain to be determined. These details, as they emerge, may be incorporated into our model to analyze the functional consequences of upstream signals that converge through distinct pathways to alter CaMKII activity.
While our mathematical model accounts for central aspects of the newly identified oxidation-dependent pathway for CaMKII activation, it has important limitations based on the available experimental data. Many questions remain to be answered regarding the function of CaMKII oxidation in the normal and diseased heart. For example, how do oxidation and autophosphorylation of CaMKII interact to control regulation of the holoenzyme? Do these pathways interact synergistically to activate the kinase and what are the unique/shared targets for each activation pathway? Finally, what is the relative importance of oxidized versus autophosphorylated CaMKII in normal and diseased hearts? Answers to these questions will not only facilitate the development of more comprehensive models but also will provide critical information necessary to design novel cell-specific therapies for regulating cardiac excitability.
It is important to note that CaMKII in the model detects a subspace pool of Ca2+ that reaches concentrations somewhere between cytosolic and dyadic concentrations (peak concentration 10–20 µM). Previous modeling studies have shown that the dynamic response of CaMKII may vary greatly between dyadic and cytosolic pools based on variability in concentrations of Ca2+ and CaM [27]. Consistent with previous studies [27], we found that cytosolic Ca2+ transients do not significantly activate CaMKII activity at baseline or even in the presence of 1 µM H2O2 (<1% maximal activity, not shown). However 10 µM H2O2 was able to activate CaMKII (25% maximal), suggesting that a sufficiently high level of oxidative stress may be able to activate even cytosolic CaMKII. Clearly, local regulation of CaMKII in well-defined subcellular domains is an exciting area for future research with important implications for human disease. As we learn more about CaM and CaMKII signaling in the vicinity of Na+ channels, it will be important to incorporate these data into the model.
|
10.1371/journal.pntd.0001656 | Prediction Score for Antimony Treatment Failure in Patients with Ulcerative Leishmaniasis Lesions | Increased rates for failure in leishmaniasis antimony treatment have been recently recognized worldwide. Although several risk factors have been identified there is no clinical score to predict antimony therapy failure of cutaneous leishmaniasis.
A case control study was conducted in Peru from 2001 to 2004. 171 patients were treated with pentavalent antimony and followed up to at least 6 months to determine cure or failure. Only patients with ulcerative cutaneous leishmaniasis (N = 87) were considered for data analysis. Epidemiological, demographical, clinical and laboratory data were analyzed to identify risk factors for treatment failure. Two prognostic scores for antimonial treatment failure were tested for sensitivity and specificity to predict antimony therapy failure by comparison with treatment outcome.
Among 87 antimony-treated patients, 18 (21%) failed the treatment and 69 (79%) were cured. A novel risk factor for treatment failure was identified: presence of concomitant distant lesions. Patients presenting concomitant-distant lesions showed a 30.5-fold increase in the risk of treatment failure compared to other patients. The best prognostic score for antimonial treatment failure showed a sensitivity of 77.78% and specificity of 95.52% to predict antimony therapy failure.
A prognostic score including a novel risk factor was able to predict antimonial treatment failure in cutaneous leishmaniasis with high specificity and sensitivity. This prognostic score presents practical advantages as it relies on clinical and epidemiological characteristics, easily obtained by physicians or health workers, and makes it a promising clinical tool that needs to be validated before their use for developing countries.
| The manuscript is relevant because of the finding of a new risk factor for chemotherapy failure and the development of a prognosis score for cutaneous leishmaniasis. The proportion of patients that have multiple lesions in American Tegumentary Leishmaniasis (ATL) is considerable. Publications and our experience permit to estimate that they represent around 20% of the affected population from the Amazon basin with cutaneous lesions. In addition, about 1/3 of them would correspond to the concomitant distant lesions category, the novel risk factor identified with a very high odds ratio (20–30) associated. Such numbers merit study of concomitant distant ulcers category on its own, not only because of clinical management implications, but also to search for factors that are contributing to chemotherapy failure. Finally, the simple equation proposed in the manuscript can be easily adapted to smart phone technologies. Similar prognosis equations are scarce for other pathologies and do not exist for Cutaneous Leishmaniasis at all. The simplicity of this tool should be followed by subsequent epidemiologic studies in other ATL endemic regions.
| Leishmaniasis is caused by protozoan parasites of the genus Leishmania sp. There is an estimated 2 million new cases and almost 70 000 attributable deaths worldwide every year [1]. Leishmaniasis includes a cluster of diseases with widely diverse clinical manifestations. These include three major groups of clinical disorders: visceral leishmaniasis (VL), cutaneous leishmaniasis (CL), and mucocutaneous leishmaniasis (MCL) [2]–[4]. New world cutaneous Leishmaniasis (also known as American Tegumentary Leishmaniasis) is endemic in the Andean region affecting the most deprived socioeconomic groups.
Pentavalent antimonials (SbV) have been the first line drugs for leishmaniasis treatment for more than 50 years. Studies conducted in Latin America have reported rates of antimony failure ranging from 7% up to 39% [5]–[7], raising a serious concern on health policy makers. Several factors like the parasite species [8], [9], the duration of the period living in the endemic area [10], the duration of the presence of skin lesions before the start of treatment, and the presence of multiple cutaneous lesions, were found to be significantly associated with SbV therapy failure [9], [10].
The use of clinical scores as predictive tools is being more frequently used in several conditions. It provides a valuable tool for clinical management, orienting physicians to establish the most appropriate treatments [11]. Given the considerably high level of antimonial resistance [12] and the need to develop strategies to improve the treatment of patients [12], [13], a clinical score to predict SbV treatment failure would have an important impact on the control efforts for CL leishmaniasis.
This study evaluates potential risk factors and disease severity parameters for SbV treatment failure of ulcerative cutaneous leishmaniasis. A predictive score for antimonial treatment failure (PSATF) is generated in a nested case-control study conducted in Peru.
We performed a nested case-control study from a prospective cohort that took place between November 2001 and December 2004 at the Leishmaniasis Clinic of the Instituto de Medicina Tropical “Alexander von Humboldt,” Universidad Peruana Cayetano Heredia, in Lima, Peru. The clinic serves patients from nearly all areas of endemicity in the country.
Subjects from both sexes and all ages with a first episode of parasitologically confirmed CL by direct Giemsa stained smear or positive culture were recruited. Patients with only ulcerative lesions and with a first diagnosis of CL without mucosal, disseminated or diffuse lesions, who received at least 20 doses of antimonials, and who were followed for at least 6 months after starting SbV treatment were included.
The patient's population included in this study was a sub-cohort of a larger sample previously reported [8]. Written, informed consent was obtained from all patients. In case of children their parents or guardians gave consent. Research protocols complied with national and international ethics policies. The human experimentation guidelines of the Institute of Tropical Medicine Antwerp were followed. Ethics clearance was obtained from the ethical committees of the Universidad Peruana Cayetano Heredia and the Institute of Tropical Medicine Antwerp, Belgium. Clinical and epidemiological data for each patient were available from a previous study [8].
Patients received treatment on site, with standard supervised daily administration of generic sodium stibogluconate (SSG) from Colombia (Viteco SA, lots: 10700, 10800, 20600, 20700 and 30600) or India (Albert David Ltd, lot: 3P12001) following Ministry of Health (MHO) of Peru guidelines (20 mg Sb5+/kg/day), for 20 days by intravenous or intramuscular injection. Quality control for SbV concentration in all batches was performed by the International Dispensary Association (Amsterdam, The Netherlands). Follow-up visits were scheduled for 1, 2, 3, 6, and 12 months after treatment ended. Patients with treatment failure received either a repeat course of antimonials with or without topical imiquimod (Aldara; 3 M Pharmaceuticals) or intravenous amphotericin B (amphotericin B deoxycholate; Bristol-Myers Squibb).
After completion of the therapy, patients were schedule at 1, 2, 3, 6 and 12 month to evaluate the progression of the disease. On each visit the patient was clinically classified as (1) cure, if complete wound healing, with epithelization and absence of any sign of activity or inflammation or (2) failure, if increased inflammation around the initial lesion, with or without epithelization, clinical reactivation of a healed lesion, or presence of new lesion(s) or a satellite lesion around the initial one was evidenced. Pending was record in patients that evidence lesion in progress to closure. Treatment failure was considered if lesion received clinical classification of failure after 3 months of follow up. Patients clinically classified as cure continued the follow up scheme in order to monitor relapse.
Information available for patients included age, sex, body mass index (BMI), main occupation, geographical region where disease was acquired, and disease severity parameters. Main occupation was classified as low or high risk for exposure to insect bites. High-risk occupations included agriculture, mining, and logging. Lesion description included number, type, location, and lymph node compromise. Typing of the leishmania species isolated from patients was performed as previously described [8].
Parameters that reflect the severity of the ulcerative lesions were measured. These included time of the disease and lesion size, which was calculated as (transversal diameter/2)×(sagital diameter/2)×3.14. On patients with multiple ulcerated lesions, the total area was calculated considering areas of the three largest lesions. Time of disease was defined as the time period since the recognition of the first lesion until the start of treatment.
Another parameter considered was the presentation of “concomitant-distant” lesions at the time of enrollment. This parameter was defined as the appearance of more than one lesion in different segments of the body (head, arms, trunks and legs) within 15 days. Cutaneous lesions that appeared in the same body segment were considered satellital lesions but not “concomitant-distant”.
In all cases age was expressed in years, time of disease in days, and the total area of lesion in cm2. Gender was considered 0 if female and 1 for male. Genotype L. braziliensis was considered “1” if leishmania type is L. braziliensis, otherwise it was considered “0”. The covariate low/high risk activity is considered “1” in case of a high-risk occupation otherwise it was considered “0”. The concomitant-distant ulcers category was “1” if it was present, otherwise it was considered “0”.
Categorical variables were described by frequencies and proportions. Continuous variables were described by means, medians and standard deviation. Categorical variables were compared by using chi-squared test. t-test and ANOVA were used to compare normally distributed variables while U Mann-Whitney and Kruskal-Wallis tests were used for non-normal distributed variables.
The probability of treatment failure was modeled in a multiple logistic regression. Potential clinical risk factors and severity indicators were tested as predictors after adjusting for potential confounders. Interactions and second order effects were tested. Multi-dimensional outliers were assessed with the test of Hadi [14]. Covariates that were significant in the univariate analysis and were not over-correlated with other covariates (correlation coefficient less than 0.75) were included in the multiple regression analysis. Nested models were compared with the likelihood ratio test. The best sensitivity and specificity of the PSATF was estimated by maximizing the Youden's index, J = sensitivity+specificity−1 [15].
A clinical tool to predict antimony treatment failure was created using the best multiple logistic regression model. The probability of treatment failure was calculated as:Where L is the linear predictor of the best multiple logistic model (a0+a1 X1+a2 X2+….+an Xn, where Xi are the significant covariates included in the best multiple model, and ai are the regression coefficients. All the statistical analyses were conducted with a 5% significance level using the software Stata 10.
A total of 171 patients received a diagnosis of leishmaniasis. In all these patients the parasite was isolated and typed during the study period. Of these patients, 87 met the eligibility criteria for ulcerative lesions and were included in the analysis. Eighteen patients (20.7%) failed the treatment and 69 (79.3%) cured. The age distribution ranged from 0.3–85 yr with a mean of 29 yr. Thirty patients (34.5%) were female and 57 (65.5%) were male. Fifty-two patients (59.8%) presented single lesions. The number of lesions in patients classified as multiple lesions ranges from 2 to 7. Among patients with multiples lesions, 10 (23%) showed “concomitant-distant” lesions.
A total of 84 patients did not meet eligibility criteria and were excluded from the analysis: 14 had previously received treatment for leishmaniasis, 18 presented mucosal involvement, 10 did not complete the first round of SbV treatment, 14 were followed-up for only 6 months, and 28 presented non-ulcerative lesions (nodule, plaque, mixed).
Descriptive statistics were compared between patients that cured and failed to treatment (Table 1). In the univariate analysis, patients who failed treatment were younger (mean = 16.2 years) than patients that cured (mean = 32.4 years) (P<0.001). Treatment failure was significantly associated with the type of work activity. People living in areas with a high rate of insect bites presented lower risk of chemotherapy failure (P<0.001). Time of disease and the size of induration, although border line significant, were included in the logistic regression analysis.
Among the clinical parameters that define the severity of the disease, the total area of lesions and the presence of “concomitant-distant” lesions, were significantly associated with treatment failure. Total area of lesions in patients that cured was greater than in patients that failed treatment (2.7 cm2 vs 1.26 cm2, U-Man-Whitney test, P = 0.007). Fifty percent of people with concomitant-distant lesions failed to antimony chemotherapy whereas only 17% failed among the patients that presented either single or multiple non-concomitant-distant lesions.
The unadjusted odds ratios estimated from the univariate analysis for treatment failures are shown in Table 2. Covariates significantly associated with treatment outcome were: Age, occupation, parasite species induration size, the natural logarithm of the total area of lesion, and the presentation of concomitant-distant lesions. The presence of concomitant-distant lesions showed a remarkably highly significant association in the univariate model (OR = 4.92, P = 0.023). The number of lesions was not significantly associated with treatment failure. The logarithmic transformation of the total area of lesion evidenced a significant association (OR = 0.50, P = 0.006).
The best multiple logistic model to explain treatment failure included six covariates significantly associated with treatment outcome (Table 2). The regression coefficients of the linear predictor of the best multiple logistic model were used to calculate the PSATF-1 (PS1 = 1/(1+e −L1), where L1 = 6.617−0.12 (age)+3.24 (L. braziliensis)−0.027 (Time of disease)−0.64 (log total area lesion +1)+3.41 (Concomitant-distant)−2.65 (low/high risk activity)). Noteworthy, the presence of concomitant-distant lesions and the type of Leishmania species were strongly associated with treatment failure (odds ratios of 30.5 and 25.7 respectively). Treatment failure predictors associated with odds ratios ranging from 6 to 30, showed a statistical power from 0.64 to 0.99 respectively.
The best model explained 54% of the variability of treatment failure. The total area under the Receiver Operating Curve was 0.93 (Figure 1). The best sensitivity and specificity that maximized the Youden's index were 77.78 and 92.52 respectively, for a PS1 cutoff of 0.4 (if PS1>0.4 failure is prognosticated, otherwise it is expected a cure). Different sets of sensitivity and specificity of PS1 were tabulated for different cutoff values (Table 3).
In addition we evaluated a score to model the chemotherapy failure without considering the leishmania specie (PSATF-2 (PS2)). This multivariate model included four significant variables and was able to explain 38.7% of the chemotherapy failure (Figure 1). Concomitant-distant lesions remained the most important predictor of chemotherapy failure (OR = 6.27, although significance was 0.054). The optimal sensitivity and specificity of PS2 were determined by maximizing the Youden index and reached 66.67% and 92.54% respectively, which corresponded to a PS2 cutoff of 0.4.
The regression coefficients of the linear predictor of the reduced multiple logistic model (Table 4), were used to calculate the PSATF-2 (PS2 = 1/(1+e −L2), where L2 = 2.84−0.073 (age)−0.028 (Time of disease)+1.83 (Concomitant-distant)−2.06 (low/high risk activity)). A free Excel-based electronic calculator of PSATF-1 and PSATF-2 (Calculator S1) is available at the online supplementary material or upon request to the authors.
We report here for the first time a score for prognosis of antimonial therapeutic failure in ulcerative CL patients treated with SSG. This prognostic score PSATF (PS1) includes a newly identified risk factor that was highly associated with the risk of treatment failure: the appearance of concomitant-distant lesions. The proposed prognostic score could be used as a clinical prediction tool able to be adjusted (PS2 instead of PS1) according to the level of the technical capacity available on site.
The two most important factors included in the PSATF, were the appearance of concomitant-distant lesions and the species of Leishmania associated to infection. The appearance of concomitant-distant lesions compared to patients with unique or non-concomitant-distant multiple lesions appear as a promising important clinical finding to be considered during treatment of leishmaniasis. Although a relatively small number of cases give support to this finding, the high OR (30.5) is statistically significant (p = 0.023). However, a larger cohort study will be required to confirm this finding. Confirming the findings in other studies [8], [10], infection with L. braziliensis also accounts as an important risk factor for treatment failure (OR = 25.7).
A limitation is that concomitant-distant lesions may not be identified at a very early stage of the disease. However, considering that in developing countries, patients seek medical attention lately, this limitation would not prevent the appropriate use of the prognostic score in the majority of situations.
As showed in Table 1 Previous works by our group and others [10], [16] showed the relationship between the area of lesion and the chemotherapy failure. This suggests that contrary to what is a common concept of treatment of leishmaniasis, early treatment and in consequence a smaller lesion size seems to be a risk for chemotherapy failure.
The frequency of SbV treatment failure estimated in this study was as high as reported in other sites [8]–[10]. About 20% of patients fail to SbV treatment in a first treatment scheme. Therefore it is important to predict failure with a reasonable sensitivity and specificity. The PSATF proposed could provide different combinations of sensitivities and specificities, according to specific necessities. The PSATF cut off of 0.4 has privileged the specificity over sensitivity to optimize the safety and rational use of chemotherapy with SbVs. The optimal values of sensitivity and specificity indicate that 78% of patients who failed treatment were correctly predicted while 92.5% of patients who cured were classified as such. In this way a larger proportion of patients will be correctly treated with the drug that is provided free of costs by the ministry of health while avoiding the use of second line drugs that have adverse side effects and are more expensive [17], [18]. Similar approaches to determine prognostic scores of treatment failure have been proposed for other diseases such as tuberculosis [19]. Clinical scores were also developed to predict fatal outcome in patients with visceral leishmaniasis [20], [21].
Our proposed PSATF includes two different models depending on the availability of genotyping of Leishmania species. Given that in some settings genotyping is not possible to perform, the use of model PS2 that does not require genotyping appears to be an alternative to improve the management of this disease in those places. It is important to highlight that the sensitivity and specificity are not largely compromised when the species of leishmania is excluded.
The appearance of concomitant-distant lesions was highly associated with the risk of treatment failure in contrast to the presence of multiple lesions regardless the timing of onset. The total number of lesions has been previously suggested to be a risk factor for SbV treatment failure [9], [10], but in these studies, concomitant-distant lesions were not distinguished.
Cutaneous leishmaniasis is characterized by lesion(s) that progress from an erythema to the typical ulcerative form in a range of 2–24 weeks [22]. All the patients included in the study were clinically classified as ulcerative cutaneous leishmaniasis and the number of lesions range from 1 to 7. Concomitant distant lesions do not seem to be the typical clinical form of disseminated leishmaniasis since it is characterized by the presence of numerous ulcerative and papular lesions.
Given its remarkable importance, it is likely that the appearance of concomitant-distant lesions correspond to a different biological phenomenon that needs to be further studied. A possibility is that these lesions could be a consequence of an intrinsic immune failure that favors metastasis [23], or a consequence of multiple infected sand-fly bites on different parts of the body [24]. However, in both cases it might indicate a decreased capability of the immune system to undergo a cell-mediated immunity against leishmania parasites.
The PSATF here presented has practical advantages because it depends on observable clinical and epidemiological features, easily obtained by physicians or health workers. With the increased use of portable computational systems, the prognostic score PSATF could be easily used by physicians in tablet PCs and smartphones. Prospective clinical studies should probe its value as prognostic tool.
|
10.1371/journal.ppat.1000987 | Intergenomic Arms Races: Detection of a Nuclear Rescue Gene of Male-Killing in a Ladybird | Many species of arthropod are infected by deleterious inherited micro-organisms. Typically these micro-organisms are inherited maternally. Consequently, some, particularly bacteria of the genus Wolbachia, employ a variety of strategies that favour female over male hosts. These strategies include feminisation, induction of parthenogenesis and male-killing. These strategies result in female biased sex ratios in host populations, which lead to selection for host factors that promote male production. In addition, the intra-genomic conflict produced by the difference in transmission of these cytoplasmic endosymbionts and nuclear factors will impose a pressure favouring nuclear factors that suppress the effects of the symbiont. During investigations of the diversity of male-killing bacteria in ladybirds (Coccinellidae), unexpected patterns of vertical transmission of a newly discovered male-killing taxon were observed in the ladybird Cheilomenes sexmaculata. Initial analysis suggested that the expression of the bacterial male-killing trait varies according to the male(s) a female has mated with. By swapping males between females, a male influence on the expression of the male-killing trait was confirmed. Experiments were then performed to determine the nature of the interaction. These studies showed that a single dominant allele, which rescues male progeny of infected females from the pathological effect of the male-killer, exists in this species. The gene shows typical Mendelian autosomal inheritance and is expressed irrespective of the parent from which it is inherited. Presence of the rescue gene in either parent does not significantly affect the inheritance of the symbiont. We conclude that C. sexmaculata is host to a male-killing γ-proteobacterium. Further, this beetle is polymorphic for a nuclear gene, the dominant allele of which rescues infected males from the pathogenic effects of the male-killing agent. These findings represent the first reported case of a nuclear suppressor of male-killing in a ladybird. They are considered in regard to sex ratio and intra-genomic conflict theories, and models of the evolutionary dynamics and distribution of inherited symbionts.
| Normally, in sexually reproducing organisms, the sex ratio (ratio of males to females) is 1∶1. However, examples are known where this is not the case and there are more females than males in a population. Extreme bias in sex ratio can lead to females failing to find a mate. We studied Cheilomenes sexmaculata, a ladybird species that has females that produce more female than male offspring. In aphid-eating ladybirds, this phenomenon has been widely reported and is known to be due to the presence of bacteria that live inside the mother and are passed via her eggs to her offspring. In eggs destined to become male, the bacteria kill the embryo by some unknown mechanism. This is known as male-killing. Female offspring develop normally. Evolutionary theory predicts that in such systems, the genome of the host can fight back if a variant arises that stops the bacteria killing male offspring. In C. sexmaculata we found females that carried the male-killer but the sex ratio of their offspring depended on the male that they mated with. We carried out breeding tests to show that some ladybirds had a version of a gene that rescued the male offspring from the pathological effects of the male-killer.
| Cytoplasmic sex ratio distorters have been reported from many invertebrates [1]. One group of distorters comprises a diverse array of bacteria which distort the secondary sex ratio of their hosts towards females by killing male hosts early in embryogenesis [2], [3]. Infected females gain an advantage over uninfected females via inbreeding avoidance, resource reallocation or reduction in sibling competition. Theory predicts that, within populations biased towards females as a result of the action of maternally inherited cytoplasmic sex ratio distorters with incomplete vertical transmission, selection will favour autosomal sex ratio compensation [4], [5]. This could occur by distortion of the primary sex ratio or by distortion of the secondary sex ratio towards males, if loss of female offspring is compensated for by increased fitness of male progeny. No such case has been demonstrated in a diploid harbouring a male-killer (but see data of male-biased families in [6], [7]. In addition, selection may favour the evolution of autosomal genes that reduce the vertical transmission or the phenotypic effects of sex ratio distorting bacteria [8]. Autosomal genes that suppress the sex ratio phenotype are known for both cytoplasmic male sterility in plants [9] and sex chromosome meiotic drive in dipterans [10]. They are also suspected in the woodlouse Porcellinoides pruinosus which hosts a feminising Wolbachia [11]. A nuclear suppressor of male-killing could kill the male-killer, or reduce its vertical transmission, or prevent the symbiont from killing males, either by blocking male host recognition, or by blocking the killing act. Such a suppressor has been reported in Drosophila prosaltans where it is suggested there is a recessive allele that prevents transmission of the male-killer [12]. A suppressor conferring resistance has been demonstrated the butterfly Hypolimnas bolina infected with the male-killing Wolbachia strain wBol1, where infected Southeast Asian H. bolina produce a 1∶1 sex ratio [13]. It has been suggested that the widespread occurrence of males testing positive for known male-killers found via PCR screening of samples of 21 ladybird species, could be indicative of nuclear suppression [14]. Suppressors at fixation might also explain the findings in Drosophila recens and Ephestia cautella, where the Wolbachia strains they harbour cause male-killing when transferred to con-generic host species, although not in the original species [15], [16].
Cheilomenes sexmaculata harbours a male-killer [17]. This male-killer is transovarially transmitted, is horizontally transferable in haemolymph by microinjection and curable by both high temperature and tetracycline treatment [17], [18]. The causative agent associated with male-killing has previously been reported to be similar to that causing male-killing in Harmonia axyridis [18], a Spiroplasma [19].
Phenotypic assessments, based on egg hatch rates and progeny sex ratio, of a small sample from Tokyo showed that two of 15 matrilines had traits consistent with male-killing, with less than 50% of eggs hatching and female-biased progeny sex ratios (family Mk1 - 7 males, 73 females; family Mk2 - 2 males, 14 females) prior to antibiotic (tetracycline) treatment. Antibiotic treatment effected a cure of the trait, egg hatch rates and the proportion of males both increasing after treatment (Mk1 - 71 males, 76 females; Mk2 - 87 males, 93 females).
Identity of the male-killer was established by PCR amplification of the 16S rDNA gene using general eubacterial primers 27f and 1495r [20], cloning [21] and sequencing the gene. Briefly, genomic DNA was extracted [19] from females producing female biased sex ratios from both male-killer (m-k) lines (parental females, F1 and F2 progeny), from F1 eggs of both m-k lines, from F1 and F2 females from a normal (N) sex ratio line (N12) that never produced a biased sex ratio (five generations, 37 families), and from F1 and F2 progeny from antibiotic treated females from both m-k lines. The m-k line females from all three generations, the m-k eggs, but not the N line or progeny from antibiotic treated females, generated a 16S rDNA PCR product, which was then cloned and sequenced (1517 bp). A majority rule consensus sequence was generated for the bacterium to account for PCR errors (total 22 sequences, of a total of 26 sequences generated at this stage, 3 contaminant Streptococcus lactis and one pGEM non-transformant also sequenced). The sequence produced was submitted to EMBL (accession number AJ272038). This was compared with sequences in the EMBL, Genbank, DDBJ and PDB databases using BLASTN [22]. The sequence was identified as a γ-proteobacterium with closest similarity to a variety of secondary symbionts, of aphids (97-98% identity), in particular Hamiltonella defensa (98% identity), of whiteflies, particularly of Bemisia tabaci (97% identity) and to a number of bacteria of the genus Yersinia. To investigate these relationships further, a phylogeny was produced using 16S rDNA sequences from 25 of the closest matches from the BLASTN search (Figure 1). It is interesting to note that the male-killer clearly falls within the clade of aphid secondary symbionts and is distinct from both the whitefly symbiont clade and the Yersinia clade. The tree is not particularly informative regarding relationships within the aphid symbiont clade, with almost all of the bootstrap values being below 500. However, it is certainly suggestive of horizontal transfer of symbionts both between aphid species, as has been reported [23], [24], [25] and between prey and predator species as has been suggested for other coccinellid male-killers [26], [27].
The 16S rDNA phylogeny highlights one further interesting feature of this male-killer, as it supports the suggestion of horizontal transfer of symbionts with a change in phenotype between the alternative hosts. Secondary symbionts are known to confer protection to their hosts against natural enemies [25 for review], but natural enemy protection co-occurring with reproductive parasitism in a clade has to date only been established for Wolbachia [28].
This finding represents the second instance of a γ-proteobacterium causing male-killing [29], and the first recorded instance in a coccinellid. To confirm the absence of other known male-killers, in particular Spiroplasma, but also Rickettsia, Wolbachia and Flavobacteria, specific PCR assays were carried out on individuals from both male-killer lines, from the parental and F1 generations. In all cases these tests proved negative [27]. The male-killer in this sample of C. sexmaculata is thus not the same as that identified from H. axyridis [19], as has previously been assumed [18].
Crosses designed to test the inheritance of the trait, involving F1 females from the two putative male-killer lines (Mk1 and Mk2), with males from normal sex ratio matrilines produced unexpected progeny sex ratios (Table 1). All three crosses using males from one line (N5) produced significantly female-biased sex ratios, one (Mk1.1) being almost all female, the others (Mk1.4 and Mk2.3) giving approximately 2∶1 ratios of females to males. The remaining seven crosses, involving males from two other lines (N1 and N8) produced normal (approximately 1∶1) sex ratios.
These data suggest either an exceptionally low and variable vertical transmission of the sex biasing trait, or paternal influence on progeny sex ratios.
To test the latter possibility, male parents were transferred between some of the crosses, the male (N5) from Mk1.1 being mated multiply to the female from Mk1.2 and once to the female from Mk1.5, and the male (N1) from Mk1.3 being mated to the Mk1.1 female. In all cases the progeny sex ratios changed following these additional matings (Figure 2), that of Mk1.1 rising from 0.013 to 0.355, with those of Mk1.2 and Mk1.5 becoming significantly female-biased. The Mk1.2 female, mated multiply to the second male produced a strong female bias from three days after introduction of the new male and for the remainder of her life. However, the Mk1.5 female, mated just once to a second male, again after a lag (four days) produced a strong and significant female bias (14 male∶42 female) over four days, before her progeny sex ratio reverted towards normality. This reversion suggests that sperm from the singly mated second male was used for a block of time before utilisation reverted to sperm from the multiply mated first male.
Extended replication of this mate swapping procedure using both male-killing matrilines showed that changes in the sex ratio following change of male were reversible if males were changed back (data available on request). These data suggest the presence of a factor, acting through sperm or some other element in the ejaculate, that inhibits the male-killing action of the bacterium. The initial data could be explained by a unifactorial dominant nuclear gene.
Four questions were addressed to test this hypothesis and detail the nature of the suppressor system.
To address i) and ii) on the basis of the initial hypothesis, expected progeny sex ratios from monogamous crosses involving individuals of alternative genotypes with respect to both the male-killer and the suppressor locus (alleles: suppressor = res+, non suppressor = res−) were calculated (Table 2).
Crosses involving individuals of inferred suppressor genotype, the females being F1, F2 or F3 from male-killer matrilines, produced progeny sex ratios consistent with expectation on the basis of Mendelian inheritance (Table 2). The results indicate that the expression of the suppressor is not affected by the sex of the parent from which the suppressor is inherited. For example, cross Mk1.1.4.9, in which both parents were heterozygous for the suppressor gave a progeny sex ratio of 0.419 (n = 234), close to expectation and significantly different from both the maximum expected sex ratio if only paternally derived suppressors are expressed (χ21 = 4.869, p<.05), and from a 1∶1 sex ratio (χ21 = 6.171, p<0.05).
Mk1 and Mk2 were maintained for five generations, 118 families (minimum number of progeny = 10; mean number of progeny = 62.7) being reared. In 51 of these the genotype with respect to the rescue gene locus was inferred for both parents prior to progeny being obtained. In 46 families, results were consistent with the theorised autosomal (or pseudoautosomal) nature of the locus. The locus is not sex-linked. In the remaining five, 16S rDNA sequence analysis, performed post hoc, showed that the female parent lacked the male-killer. This proportion of revertants is roughly consistent with the estimated vertical transmission efficiency, a, of the male-killer seen in Mk1 (a = 0.89) and Mk2 (a = 0.83).
To address question iii), two female F2 progeny from predominantly female families of both the Mk1 and Mk2 matrilines were crossed to homozygous res+ males. The progeny sex ratio in each of these families was close to 1∶1. res+res− female offspring from these crosses were mated to res−res− males. Of eight crosses, one produced a normal sex ratio, the remainder producing ratios approximating to 1 male: 2 female (data not shown). Given the vertical transmission efficiency of this male-killer, these results show that the male-killer is inherited through females, even when its pathological effect on males has been suppressed. Verification was obtained by sequencing the 16S rDNA PCR product from parents and six progeny (3 male, 3 female) of two of the families that produced a 1 male: 2 female sex ratio. Female parents, all male progeny and five of the six female progeny were shown to bear the male-killer.
Verification that the suppressor acts as a rescue gene was obtained by demonstrating that adult males from suppressed male-killer females produced 16S rDNA PCR product with the same sequence as the γ-proteobacterium identified as the male-killer (question iv)). Two male progeny from two m-k families, one of which produced a 1∶2 sex ratio and one which produced a normal sex ratio (Mk1.4 and Mk2.5, respectively) were submitted to the same PCR cloning and sequencing protocol as used for the initial identification of the male-killer. In each case the γ-proteobacterium was shown to be present. The male-killer is thus present but not expressed in ‘rescued’ males.
Crosses of such males, inferred to be res+res− and to carry the male-killer, to res−res− females lacking the male-killer produced normal progeny sex ratios. The male-killer is thus not inherited from males. Taken together, these results suggest that the endosymbiont is not killed by the suppressor. Rather the suppressor acts as a rescue gene for males.
The discovery of a nuclear gene that rescues males from the pathological effects of a maternally inherited bacterium that otherwise kills males, to the benefit of the males' female siblings that carry and vertically transmit the bacterium, is in accord with theories of sex ratio [30], [31] and intra-genomic conflict [4], [5]. Selection favouring a suppressor gene will be a direct consequence of sex ratio distortion. Autosomal genes that act against sex ratio distorters have been recorded in isopods infected with feminising Wolbachia and in a butterfly infected with male-killing Wolbachia. In Armadillium vulgare, the main effect of such genes is to reduce bacterial transmission to progeny [32]. In contrast, in P. pruinosus, autosomal genes are conjectured to prevent the feminising effect of Wolbachia [11]. This is analogous to the situation in C. sexmaculata and H. bolina [13] where observations are compatible with a single, dominant autosomal locus suppressing the male-killing effect of the bacteria.
Most models considering the evolutionary interactions between sex ratio distorting symbionts and suppressors are based on the assumption that the suppressor will kill the symbiont or reduce its vertical transmission [33], [34]. The dynamics of a male rescue gene may be quite different, and will depend on its cost, if any, in the absence of the male-killer. A male-killer in the presence of a male rescue gene should be selected against, due to the cost on hosts of carrying the male-killer and the lack of fitness advantages to infected females resulting from male death [3]. Such a male-killer then has several alternative fates. It may be selected to extinction. It may become polymorphic, male killer prevalence being determined by its transmission dynamics and fitness compensation and the costs to hosts of both it and the rescue gene. It may circumvent the rescue gene by evolving a different mechanism to kill males (cf. the double feminising effect of Wolbachia in A. vulgare [35]). Finally, it may reduce its cost on hosts, becoming costless (persistence would require vertical transmission close to 1) or even beneficial - cytoplasmic male-killers are an exquisite testing ground for theories of virulence [3].
It is interesting to compare the situation in C. sexmaculata with that in H. bolina. In the latter, the suppressor has recently and rapidly spread to fixation in Southeast Asian populations, but is absent from Polynesian populations [13]. A recent model [36] examines the dynamics of such systems with reference to host suppressors of male-killing and Wolbachia that are also able to induce cytoplasmic incompatibility (CI). Here the model predicts that (in the absence of CI, and so pertinent to this study) the maximum cost of a dominant suppressor of male-killing that allows invasion of a host population will increase as the male-killer prevalence increases. The model also predicts (in the absence of CI) that a costly suppressor that does invade will become polymorphic and the frequency of the male-killer will be reduced. Further, the authors suggest that polymorphic suppressors of the male-killing action of non-Wolbachia male-killers (not known to induce CI) should be more common than of Wolbachia male-killers. Our findings fit well with this model.
In contrast to the male-killer in C. sexmaculata, the male-killing Wolbachia in H. bolina may also cause CI and where this occurs the model demonstrates that the dynamics of the system are altered, with corresponding changes in the rate of spread, fixation and frequency of infection that depend on the level of CI, the cost of the suppressor, the transmission efficiency and the initial male-killer frequency. Suppressor spread may be inhibited, where there is CI, but where a suppressor does spread it is predicted to lead to fixation of both itself and the infection, giving an appearance of a population exhibiting only CI [36].
The Fuchu population of C. sexmaculata studied here is polymorphic for the rescue gene. Further investigation of whether this is a balanced or transient polymorphism, and determination of whether the rescue gene imposes a cost on bearers, would provide valuable insight into the dynamics of the spread of suppressors, as well as the generality of the findings in H. bolina. Establishment of the frequencies of the rescue gene and male-killer in different populations of C. sexmaculata would be valuable. Further, molecular investigation of coccinellids with ecological traits making them liable to male-killer invasion, but in which searches for male-killers using phenotypic assays have proved negative, may reveal presence of beneficial symbionts that are a peaceful resolution of an evolutionary arms race between a male-killer and a suppressor system.
Lines were designated either Male-killer – Mk, or Normal – N, to reflect the phenotypic status of the P1 females in the original sample. The lines were then numbered sequentially within the two categories, hence Mk1 and Mk2 were the two male-killing lines and N1-N13 the 13 normal lines. Subsequent generations show the parental name followed by a ‘.’ to indicate a new generation, and then a number e.g. Mk1.3 is the third cross generated from F1 female progeny of Mk1 and Mk1.1.3 is the third cross generated from an F2 female, progeny of Mk1.1.
These numbers simply reflect P1 phenotype and indicate the matriline. They do not reflect rescue gene status and hence do not necessarily indicate F1 (or subsequent generation) phenotype. Rescue gene allelic status is indicated by res+ and res- for suppressor and non-suppressor, respectively.
For some tests, males known to lack the male-killer suppressor were necessary. These were obtained by tetracycline treatment of singly mated, male-killer bearing females (for both Mk1 and Mk2). Females were mated once and male-killer status confirmed. Females showing characteristic half hatch rates were treated with antibiotic. Where these initial clutches produced only female progeny, and assuming, as was subsequently shown, that res+ was expressed when inherited from the female parent, males produced in the later clutches of these crosses would be homozygous res−.
Different categories of ladybird were assessed for the presence of a bacterial male-killer by performing PCR using general eubacterial primers that amplify the 16S rRNA gene (primer pair 27f, 1495r) [20]. The PCR was carried out using Expand High Fidelity PCR System (Boehringer Mannheim). The product was purified using Microcon Microconcentrators (Amicon Ltd.) and ligated into pGEM T-vector (Promega). The resulting plasmids were transformed into E. coli DH5α as described by Hurst et al. [21]. Plasmids containing insert DNA were purified using Wizard Minipreps DNA purification system (Promega). Inserts were sequenced using the ABI PRISM BigDye Terminator cycle-sequencing ready-reaction kit (Perkin Elmer) and visualised on an ABI 377 automated sequencer. Primers pUC/M13 forward and reverse, 27f and 1495r and internal primers [37] were used to sequence both strands of the whole unit.
The 16SrDNA sequence generated above was aligned with 16S rDNA sequences from 25 different BLASTN matches with high alignment scores, which were downloaded from the nr database. The accession numbers of the 25 sequences used were: AY296733, CP001277, AF293622, EU348313, AY264676, AF293626, AF293616, AY692361, AY264675, AY136161, AY136136, AY136164, AY136162, AY136163, AY136145, AY136156, AY136148, EU178101, AB273745, AM403659, FM955884, AL590842, CP001048, NR028786, U90757. Sequences were aligned with ClustalW2 [38], minor manual adjustments were made using Seaview [39], and a neighbour-joining tree was generated excluding sites with a gap in any sequence, using Kimura's 2 parameter correction, with 1000 bootstrapped replicates, implemented in ClustalW2 [38]. The tree was displayed using NJPlot [40].
Under a model where the suppressor is a single gene, with alleles res+ and res−, where res+ is a dominant rescue allele and where the vertical transmission efficiency of the male-killer is between 0.8–1.0, the expected progeny ratios from different crosses can be calculated as follows:
In all cases female progeny will survive, whether infected or not; what varies is the proportion of males that inherit the infection and further the proportion of those inheriting the infection which carry a res+ allele and so (under this model) survive. If the proportion of progeny inheriting the male-killer varies between 0.8 and 1.0 a maximum of 20% of the progeny (10% male and 10% female) will lack the male-killer.
In a cross where both parents are free from the suppressor (res− res− x res− res−) and the parental female is infected with the male-killer, none of the progeny will inherit a suppressor and hence all the male progeny that receive the male-killer will die. This will be between 80 and 100% of the males, i.e. if vertical transmission is 1, 0 males will survive, if the vertical transmission is 0.8, 20% of the progeny will not inherit the male-killer, 10% of these are male, and will now survive, increasing the sex ratio to 0.1/0.6 = 0.167.
Similarly if the parents are res− res+ x res− res− half the progeny will inherit a suppressor allele. If vertical transmission is 100% one quarter of the progeny will die (males with no suppressor), and one third, 0.333, of the remaining progeny will be male. If vertical transmission is 80% this would mean of the 20% of the progeny that fail to inherit the male-killer, 10% will be male (as above), of which half will be res− and so will now survive, increasing the proportion male to 0.3/0.8 = 0.375.
If the parents are res+ res− x res+ res− three-quarters of the progeny will inherit a suppressor so 3/7 or 0.429 of the surviving progeny will be male assuming all inherit the male-killer. If vertical transmission is 80%, again 10% of the progeny that fail to inherit the male killer will be male. Now only a quarter of males are res−, hence another 2.5% will survive, making the proportion male 0.4/0.9 = 0.444.
Finally, if one parent is homozygous res+ then all progeny inherit a copy of the suppressor, all will survive and the proportion male will be 0.5, regardless of the vertical transmission efficiency.
These calculations assume that in the absence of the male-killer the sex ratio would be 1∶1, that the male-killer is inherited equally by males and females and that it always kills males unless there is a suppressor present.
Simulations were carried out to estimate the 95% confidence limits of the expected sex ratios. This can be illustrated using Mk 1.7. Here the range of possible sex ratios is from 0–0.167, and the sample size is 12. First, a random number is chosen, and on the basis of this, a sex ratio is randomly chosen from an even distribution from 0 to 0.167. Then, using this sex ratio, a binomial sample of 12 individuals is created, of which between 0 and (theoretically) 12 will be male. This process was repeated 100,000 times, to produce a distribution of the numbers of males seen in samples of 12. In this case the distribution is:
0 males: 41.8%
1 male: 30.7%
2 males: 17.2%
3 males: 7.3%
4 males: 2.3%
5 males: 0.5%
6 males: 0.1%
From these values it is concluded that any number of males in the data above 3 would give significant evidence against the hypothesis, since this has a chance of happening that is below 5%, and thus the 95% confidence limits run from zero to three males from 12, or from 0–0.25 as the sex ratio. Corresponding calculations were carried out for each cross listed in Table 2.
EMBL: AJ272038
|
10.1371/journal.pcbi.1000572 | Stochastic Drift in Mitochondrial DNA Point Mutations: A Novel Perspective Ex Silico | The mitochondrial free radical theory of aging (mFRTA) implicates Reactive Oxygen Species (ROS)-induced mutations of mitochondrial DNA (mtDNA) as a major cause of aging. However, fifty years after its inception, several of its premises are intensely debated. Much of this uncertainty is due to the large range of values in the reported experimental data, for example on oxidative damage and mutational burden in mtDNA. This is in part due to limitations with available measurement technologies. Here we show that sample preparations in some assays necessitating high dilution of DNA (single molecule level) may introduce significant statistical variability. Adding to this complexity is the intrinsically stochastic nature of cellular processes, which manifests in cells from the same tissue harboring varying mutation load. In conjunction, these random elements make the determination of the underlying mutation dynamics extremely challenging. Our in silico stochastic study reveals the effect of coupling the experimental variability and the intrinsic stochasticity of aging process in some of the reported experimental data. We also show that the stochastic nature of a de novo point mutation generated during embryonic development is a major contributor of different mutation burdens in the individuals of mouse population. Analysis of simulation results leads to several new insights on the relevance of mutation stochasticity in the context of dividing tissues and the plausibility of ROS ”vicious cycle” hypothesis.
| Aging is characterized by a systemic decline of an organism's capacity in responding to internal and external stresses, leading to increased mortality. The mitochondrial Free Radical Theory of Aging (mFRTA) attributes this decline to the accumulation of damages, in the form of mitochondrial DNA (mtDNA) mutations, caused by free radical byproducts of metabolism. However, there is still a great deal of uncertainty with this theory due to the difficulties in quantifying mtDNA mutation burden. In this modeling study, we have shown that a random drift in mtDNA point mutation during life, in combination with the experimental sampling can explain the variability seen in some of the reported experimental data. Particularly, we found that while the average mutation increases in a linear fashion, the variability in the mutation load data increases over time, and thus a low number of data replicates can often lead to a deceptive inference of the mutation burden dynamics. The model also predicted a significant contribution from the embryonic developmental phase to the accumulation of mtDNA mutation burden. Furthermore, the model revealed that the replication rate of mtDNA is a major determinant of new mutations during development and in fast-dividing tissues.
| Mitochondria are the main energy producing organelles present in eukaryotic cells. Mitochondria are the only organelles aside from the nucleus which harbor their own genetic material. Mitochondrial DNA (mtDNA) encodes a small number of polypeptides needed for the electron transfer chain (ETC). The ETC is responsible for cellular energy synthesis via oxidative phosphorylation (OXPHOS), during which some of the electrons leak from the ETC and are captured by oxygen to form reactive oxygen species (ROS) [1]. Most ROS are detoxified by cellular antioxidant defenses, but some escape and cause damage to cellular biomolecules like lipids, protein and nucleic acids [2]. Mitochondrial DNA may be particularly susceptible to such oxidative insult due to its proximity to the ROS production sites of the ETC [3]. Oxidative damage of mtDNA and its implications on cellular aging form the basis of the mitochondrial Free Radical Theory of Aging (mFRTA) [3]. One of the predictions of the mFRTA is the possibility of ROS ‘vicious cycle’ (Figure 1), referring to the hypothesized positive feedback mechanism in which mtDNA mutations cause an increase in the ROS production resulting in a higher de novo mutation rate [3]. Major challenges and questions with respect to the mFRTA have been summarized in some of the recent reviews [4],[5].
Despite uncertainties related to the assumptions of mFRTA, the importance of mitochondria as both the source and target of ROS in aging is supported by some transgenic mouse studies. For example, a 15% increase in the maximum and median lifespan is observed in knock-in mice expressing human catalase, an enzyme that decomposes H2O2 into water and oxygen, in mitochondria (MCAT), but not in the nucleus or the peroxisome [6]. Furthermore, MCAT mice heart tissue accumulates less than 50% of the mtDNA point mutations of age-matched wild-type mice [7]. Also, studies of homozygous knock-in mice with an error-prone polymerase-γ (POLG mutator mice) show that a dramatic increase in mtDNA mutation burden, most importantly deletions [7], is associated with shortened lifespan and some phenotypes that may resemble accelerated human-like aging [8],[9].
Although there is reasonable evidence for an age-dependent increase in mtDNA mutations, the dynamics by which these mutations accumulate is still largely unclear. Inferring dynamics and more importantly, the mechanism by which mtDNA mutations accumulate critically depends on accurate quantification of oxidative and mutational burden, which poses significant experimental challenges [4]. Many of these challenges stem from the limitations associated with experimental protocols in measuring oxidative damages and mutational frequency [10],[11], which typically exist at extremely low magnitude. Consequently, published reports show conflicting results regarding the levels of oxidative damages and mutation dynamics of mtDNA during aging [12]–[14]. A highly sensitive method based on the random mutation capture (RMC) assay has recently been developed for the quantification of mtDNA mutation frequency [15]. This method is based on restriction enzyme digestion and amplification of mtDNA molecules carrying mutations at the corresponding recognition site [12]. Application to wild-type mice has revealed mtDNA mutation burdens that were two orders of magnitude lower than previously determined using PCR-cloning and sequencing protocols [8],[9]. This indicates that PCR artifacts may have been a major contributor of errors in the past reports. Furthermore, quantification of age-dependent accumulation of point mutation burdens using the RMC assay in wild-type mice suggested an exponential increase, apparently supporting the existence of a ‘vicious cycle’ in the mutation accumulation [3],[13]. However, the low levels of burden suggest that point mutations may not be a major determinant of lifespan [12] and it is difficult to see how a positive feedback mechanism could set in at such a miniscule level of point mutation burden.
One requirement for addressing these uncertainties is a better understanding of the inherent stochasticity of cellular processes [16]. The accumulation of mtDNA mutations likely involves complex stochastic factors, such as the inherent random nature of mutations and related cellular processes in the context of aging. For instance, enzyme staining for ETC deficient tissue of substantia nigra neurons in aged subjects and Parkinson patients revealed a high degree of mosaicity of COX respiratory deficient cells [17]. This mosaicity has also been seen in skeletal muscle cells associated with sarcopenia in aged subjects [18]. Also, studies on Caenorhabditis elegans indicate that individual worms and their cells harbor a wide spectrum of mtDNA deletion loads [19].
Here we aim to address these challenges using a systems approach by way of constructing mathematical models that encompass the most relevant biological processes and also features related to experimental protocols to comprehend the origin and consequence of mutation variability that arises in individuals of a mouse population. Additionally, we seek to better understand the influence of intrinsic stochasticity of the mutation process on the variability observed in the experimental data. Such understanding may reveal possible causes of disagreements amongst published reports and further facilitate optimization of experimental design. In this study, we have constructed an in silico stochastic mouse model using the Chemical Master Equation (CME) [20]. Here, the accumulation of point mutations in mtDNA is simulated to arise as a consequence of what we believe to be a minimal process required for the maintenance of mtDNA integrity.
The in silico mouse model accounts for the accumulation of mtDNA point mutations across two stages of mouse life: development and postnatal (Figure 2). In this study, the number of wild-type mtDNA (W) and mutant mtDNA (M) molecules are tracked for each cell in whole mouse heart (∼2.5×107 cells) and liver tissues (∼4×108 cells) [21]. Each mutant mtDNA molecule is assumed to contain only a single mutation in the TaqI recognition site (TCGA), following the RMC experimental design [12]. The probability of finding two or more mutations at the same site is negligible [15]. The model simulates two mtDNA-related maintenance processes: mitochondrial turnover, comprising of relaxed replication and degradation of mitochondria, and de novo point mutation, based on a minimal conservative assumptions. First, the mtDNA population of each cell is assumed to exist as a well-mixed pool due to fast fusion and fission dynamics of mitochondria [22]. Second, due to the low overall mutation burden, point mutation burden is assumed to remain below the level of functional significance (i.e. no nuclear retrograde signaling [23],[24]). While the latter assumption is conservative, our simulations indicate that the incorporation of functional effects of mutations into the model, by assuming that mutant mtDNA are non-functional and cells respond to a decrease in the number of wild-type (WT) mtDNA by increasing replication, does not result in any significant changes to the mutation burden (see Text S1 and Figure S3). A Langevin formulation using relaxed replication assumption demonstrated that stochastic drift can lead to a clonal expansion of mtDNA mutations in human [25].
Following experimental evidence, each mitochondrion is assumed to carry 10 mtDNA molecules and these mtDNA are assumed to undergo replication and degradation due to mitochondrial turnover [26]. In a turnover event (Figure 2B), ten molecules of mtDNA are chosen randomly from a well mixed population of mtDNA in a cell and are either degraded or replicated according to the CME described below. The selection of ten wild-type and mutant mtDNA molecules from the population can be described as a hypergeometric random sampling following the probability distribution: [27](1)where x represents the number of wild type mtDNA chosen for replication or degradation.
De novo point mutation can occur during replication of mtDNA due to mis-pairing associated with ROS-induced mutagenic lesions such as 8-hydroxy-2-deoxyguanosine (8OHdG) [2] or as random errors arising due to finite polymerase-γ (POLG) fidelity [28]. Consequently, each replication of a wild-type mtDNA has a finite probability, given by the mutation rate constant (km), to produce a mutant. Here, the number of de novo mutant mtDNA is randomly chosen from a binomial distribution: [27](2)where y denotes the number of de novo mutations resulting from replication of x wild-type mtDNA.
Based on these probabilities, the in silico mouse model is formulated as a CME in which each mtDNA-related process: replication without mutation, replication with de novo mutations and degradation, is described as a jump Markov process with the following state transitions: (3)The first two transitions reflect replication without mutation, the third represents de novo mutation, and the last pair represents degradation. A general formulation of CME is given by: [20](4)where is the state vector denoting the total number of each molecular species present in the system and the function denotes the probability of a system to assume the state configuration at time t, given the initial condition at time . The function aj denotes the propensity function, while is the state change associated with a single j-th event. The propensity function gives the probability of the j-th event to occur in the time interval [t, t+dt). As analytical solution to CME is usually not available even for moderately sized systems [29], Monte Carlo algorithms have been employed to solve the CME numerically [30], e.g. using Gillespie's SSA (Stochastic Simulation Algorithm) [31]. In SSA, two random variables (, j ) determine the temporal evolution of the states in a system, where is the time for the next event to occur and j is the type of event that will take place. The probability density functions of and j are evaluated based on the propensity function of the events involved [29].
A modified version of the SSA is used in this work for simulating in silico mice tissues based on the following CME:(5)
The density function denotes the probability of a cell in a given tissue to contain W and M number of wild-type and mutant mtDNA, respectively, given the initial conditions of the states (not explicitly stated here for brevity, refer Equation 4). The parameters kR, kd and km represent the specific probability rate constants for mtDNA replication, degradation and de-novo point mutations, respectively. The terms in the curly braces describe the hypergeometric sampling of mtDNA from the population. Particularly, the first two terms of the CME above represent mtDNA replication without mutation, the second pair of terms corresponds to replication with de-novo mutation, and the last two terms represent the degradation of mtDNA. The CME can be solved numerically using a Monte Carlo approach following the SSA. The implementation of the modified SSA is described below:
To predict mtDNA mutation burden in a single organ or tissue, millions of such simulations are performed to capture the mtDNA dynamics of all cells in a tissue.
Simulations were performed using an IBM high performance computing cluster with 112 Intel 1.6 GHz processors. The simulation code (Text S2) was compiled using GNU FORTRAN compiler G77 (v4.1.1) and run on a CentOS Linux platform. On average, a single simulation of a heart tissue (∼25 million cells) from development to 3 years of age required approximately 3 hours.
The embryonic cell divisions begin after fertilization of an oocyte. Mouse oocytes harbor a large number of mitochondria (∼1.5×105 mtDNA) [32], which allow the zygote to multiply initially without the need to replicate mtDNA [33],[34]. Mouse embryos with dysfunctional mitochondrial replication are able to proceed through the implantation and gastrulation stages, but eventually die, presumably when the mtDNA synthesis becomes necessary to maintain ATP level [35],[36]. Furthermore, the total mtDNA number in mouse embryo does not increase until the late stage of blastocyst, which is roughly the 7th to 8th cell divisions in development (i.e., 4.7 to 5.5 days post coitum (d.p.c)) [33],[34],[37]. During these stages, mtDNA are segregated among the dividing progenitor cells (Figure 2A). Consequently, each progenitor cell of the developing embryo has only few copies of mtDNA at the early egg-cylinder stage [33],[34].
In order to account for the mtDNA segregation without replication during the initial cell divisions, the developmental simulations start from the end of the 8th stage (5 d.p.c) with an initial wild-type mtDNA count of roughly 580 molecules per cell (W = 580, M = 0) [33]. Mitochondrial DNA replication is tied to the cellular division to maintain a steady state number of total mtDNA after each division [38]. Mouse development lasts until 20 d.p.c [39] with a doubling time of roughly 15.5 hours [40]. The mtDNA replication rate is estimated assuming that mtDNA doubles its population every 15 hours while still undergoing degradation. Here, a cell division occurs when the total number of mtDNA count reaches twice the steady state homeostatic count (Table 1). The segregation of wild-type and mutant mtDNA between the daughter cells is assumed to occur at random, without any selective advantage according to a hypergeometric distribution: [27](6)where denotes the number of wild-type mtDNA in one of the daughter cells after segregation and n is the total number of mtDNA in a single daughter cell (i.e., n = (W+M)/2). During development, polymerase-γ, the care taker of the mtDNA replication fidelity, is the main contributor for point mutations in mtDNA, with negligible oxidative activity and damage [28],[41].
After birth, many tissues like heart do not undergo further cellular division. However, mtDNA in these tissues are still continuously turned over independent of cellular division, a process called “relaxed replication” [26]. The functional significance of relaxed replication in postmitotic tissues like heart and brain is to maintain a healthy population of mtDNA to satisfy the cellular energy requirements [26],[42]. The postmitotic simulations continue from cells produced at the last stage of development (Figure 2A), in which each cell maintains mitochondrial biogenesis to balance degradation. The mutation rate in this stage is a summation of contributions from oxidative damage and POLG-related error.
The in silico mouse model is also used to simulate POLG mutator heterozygous (POLG+/mut) and homozygous (POLGmut/mut) mice by changing the rate of de novo point mutations. Mutator mice carry a proofreading-deficient allele of POLG which has 200 times the error rate of the wild-type enzyme [28],[43]. Thus, in the simulations of POLG mutator mice, the model formulation remains the same in all aspects with the exception that the POLG error rate corresponding to the mutant allele is assumed to be 200 times higher (Table S2 and S3). In heterozygous POLG mutator mouse, the replication of mtDNA molecules is carried out by either wild-type or mutant allele with equal probability.
Model parameters are compiled from published data for mice and we have ensured that they are consistent with the current literature and the state of the art techniques. The basic model parameters are listed in Table 1, while more detailed information of the rest of parameters used in all mouse models is given in Tables S1, S2 and S3.
In silico wild-type (WT) mouse population of 1100 individuals was generated starting from embryo up to three years of age, the approximate life span of mice (Figure 2). The overall point mutation frequency in 2.5×107 cells of whole heart tissues was recorded at the end of each cell division during development and every fortnight during the postnatal stage. Figure 3 illustrates the percentile and distribution function of the mutation frequency arising from two important sources of variability related to the quantification of mtDNA point mutations. The probability density functions indicate the distribution of mutation frequencies in the population as a function of time. Each contour on the percentile plot represents the maximum mutation frequency that a given percentage of the population harbors (e.g. 99% of mice harbor mutation frequencies up to and including the level indicated by the 99th percentile curve (Figure 3A, 3C)).
The main source of randomness is the intrinsic stochastic nature of the aging process, which arises from the mtDNA maintenance processes (Figure 2B). Note that the intrinsic stochasticity prevailing in the in silico population has a long tailed non-Gaussian density function (Figure 3B, 3D), indicating that a small fraction of the population harbors a significantly higher mutation burden. Cell-to-cell variability of mtDNA mutation load is also observed as a result of the random processes (Figure S1). Figure 4 illustrates the evolution of mtDNA states (W and M) in two cardiomyocytes during the postnatal stage of a mouse. Random fluctuation of wild-type mtDNA can be seen in the population with regular bursts and decay of mutant mtDNA. Furthermore, it is interesting to observe that despite the significant cell-to-cell variability of mutation load being large (Figure S1), the average accumulation of mtDNA mutation in tissue remains linear after birth (Figure 3A). Also, the variance due to the natural aging process remains roughly constant during the mouse life span, indicating that the variability among individuals is inherited at birth.
However, for comparison with data derived from RMC assay, a second source of variability has to be considered due to the intrinsic statistical properties of the assay protocol. This is because, the determination of point mutation burden by the RMC assay involves drawing a random sample of mtDNA copies (∼840,000) from tissue homogenates [12]. This sampling procedure introduces additional variability that becomes significant due to the low overall count of total mtDNA mutations. This statistical feature of the RMC protocol can be described as sampling from a hypergeometric distribution [27]. (8)where m denotes the number of mutant mtDNA molecules present in a random sample of mtDNA of size n (n = 840,000 mtDNA molecules in this case). Thus, for low mutation frequencies and sample sizes, the RMC protocol introduces significant additional variability in the data. For example, in heart tissue homogenate containing 1010 molecules of mtDNA with a mutation frequency of 10−6/bp (a total of 4×105 mutant mtDNA), samples of 840,000 mtDNA drawn from the same homogenate will have a mean value of 3.36 mutants with a standard deviation of 1.83 molecules or 54.6% coefficient of variation from the RMC sampling alone.
The compounded effect of the two sources of variabilities (intrinsic aging related and RMC assay) can be expressed by,(9)where denotes the underlying probability distribution of mtDNA mutations predicted by the mouse model simulations and is the overall probability function of measured mtDNA mutations. Importantly, the additional variability associated with the sampling of mtDNA in the RMC protocol causes the mutation frequency variance to increase as a function of the average mutation frequency (Figure 3C), a result expected from a hypergeometric distribution. This is particularly relevant here because of the age-dependent increase in mean mutation burden and the fact that the distribution describing the mutation process is long-tailed (Figure 3A, 3B). When this underlying mutation dynamics is sampled using the RMC assay, the resulting data will exhibit an age-dependent increase in variance. Due to low number of replicates (typically n<5 per age group), it is highly probable to obtain data that are best approximated by a non-linear, possibly exponential model (Figure 3C). However, this apparent exponential increase is not actually a feature of the underlying mutation dynamics, which may be in fact, linear (Figure 3A). This has important implications for the interpretation of the available experimental data.
In accordance with the interpretation reached in the original experimental work [12], the variance in the in silico data as well as the experimental data for low n-values appears to suggest an exponential dynamics supporting the ‘vicious cycle’ theory [3],[13]. However, on careful consideration (Figure 3), the apparent exponential increase of the mutational burden is actually an artifact of: (a) intrinsic stochasticity of aging process (Figure 3A, 3B), coupled with (b) the random sampling variability introduced by the statistical properties of the RMC protocol (Figure 3C, 3D). Experimentally, it is not possible to carry out 100 s or 1000 s of repeats and it is therefore difficult to distinguish between a truly exponential and a linear increase of age dependent point mutation burden. In summary, while the RMC assay is able to quantify extremely low levels of mutations, its discrete nature (in terms of mutant mtDNA count) introduces significant challenges in data analysis and interpretation. The interpretation of the data can be flawed if the statistical properties of the RMC assay are not considered. Taking both processes into consideration, the fundamental mtDNA maintenance processes modeled by our in silico mice are in excellent agreement with the published data (Figure 3C). However, the last data point of mutation burden from an old mouse (980 days) deviated from in silico mouse population (p-value = 0.064), suggesting that other processes not predicted by our model may be involved during the last months of life (e.g., inflammation or other disorders that can accelerate oxidative DNA damage [58]).
Transgenic mouse studies on POLG mutator mouse have recently shed some light on the role of mtDNA in aging [8],[9],[12]. However with these mutator models, many open questions still remain about the role of mtDNA mutation in aging. For example, only the homozygous mutator mice exhibited accelerated human-aging-like phenotypes (e.g., anemia, alopecia, kyphosis) and shortened lifespan, while the heterozygous mice have no obvious aging phenotypes, despite significantly elevated mutation burdens [9].
After successfully validating the in silico mouse model against wild-type mouse data, we further simulated 1,100 hetero- and homozygous POLG mouse heart and liver tissues by elevating the baseline POLG error rate to 200 times that of wild-type [28],[43]. We found an excellent agreement of our in silico results with the reported mutation burdens from two different laboratories [9],[12] (Figure 5 and Figure S4). As with the wild-type mice, the point mutation increase was linear with age (Figure S4). Again, mitochondrial turnover and de novo point mutations alone were sufficient to explain the accumulation of mtDNA point mutations. These results indicate that even at the elevated levels of point mutations ROS-mediated acceleration of point mutations with age is not necessary to explain the data presented in [8],[9]. This is consistent with additional experimental observation suggesting that the levels of ROS in POLG mice are not significantly elevated in the mutator mice [8]. Crucially, no modification of mtDNA maintenance rate constants was required to reproduce the experimental data [8],[9]. That is, one does not have to resort to assumptions such as the existence of a vicious cycle or other possible feedback mechanism [59],[60].
The stage in an organism's life from which the accumulation of mtDNA mutations starts to become functionally significant (if at all) is unclear. During development, mtDNA replication is tied to the cellular division, and as a consequence, initial mutations may arise as soon as mtDNA replication begins. In fact, the total number of replications during development is comparable to that during the entire adult life. In mice, the heart tissue develops in about 20 days [39]. Considering the degradation rate described in Table 1 and the mouse heart to contain ∼2.5×107 cardiomyocytes [21],[61] arising from 22 cell divisions (6 progenitor cells), the total number of mtDNA replications needed to maintain homeostatic value of mtDNA (Table 1) [21] per cell should exceed 9×1010 times during the development. On the other hand, based on the degradation rate of mtDNA in postnatal stages (Table 1) [45], the number of mtDNA replications events over the three years lifespan of mice is about 1.3×1011. Thus depending on their source (ROS, POLG errors), the development period may carry comparable contributions in de novo mtDNA mutations as does the entire adult life.
POLG errors have been postulated to be the main cause of de novo point mutations in murine embryonic fibroblast [28],[41]. Therefore, the POLG baseline error rate was used as mutation rate during development. Generally, our in silico mouse data highlight that mutations occurring in the early embryonic cells have a strong impact on the mutation load at birth (Figure 6) and that the variability among individuals is set during development (Figure S2 and Figure S4). Since the mtDNA replication is several folds higher than the degradation during development, de-novo point mutations generated during the early cell divisions can accumulate very quickly, resulting in a high mutation load at birth in some individuals (Figure 6). These results highlight that the stochastic drift of mutation dynamics during the early developmental cell divisions may be a deciding factor of the organism's mutation trajectory, and also a major contributor of the mutation variability in a population, including isogenetic individuals [19]. The variability generated during development is conserved throughout the organism's life (see Figure 2A and Figure S4A, S4B).
In postmitotic tissues, like heart, mtDNA are continuously turned over independent of cellular division [26]. Although the turnover rate of mtDNA is lower during the postnatal stage than during development, the higher mutation rate due to oxidative damage (Table 1) can lead to 2–3 fold increase in the mutation load between birth and old age in wild-type mice (see Figure 2A and Figure 7). The in silico POLG mice however differ from the wild-type because in these mice, the POLG error is the dominant contributor of de novo point mutations, both during embryonic and postmitotic stages (Supplementary Table S2 and S3). Due to faster mtDNA replication (tied to cell division), most of the mutations in mutator mice therefore arise during development (Figure 6B, 6C and Figure 7). This is consistent with the experimental data which shows clearly that mutator mice are born with significantly elevated mutation burden [9],[62]. However, during their adult life, the accumulation is relatively lower compared to their development, due to the slow turnover of mtDNA [45].
Furthermore, the above observation leads to an interesting insight, largely unappreciated in the original work [8],[9],[63], regarding the point mutation load in tissues that remain mitotic (epidermal, stem cells, spleen). Since in POLG mice the point mutation burden of mtDNA is dominated by POLG errors, mutation accumulation in fast dividing cells is expected to be several fold faster than in postmitotic tissues such as heart. This is consistent with the experimental observation in POLG mutator mice, where some of the most prominent pathologies associated with the fast dividing tissues manifest in the form of alopecia, spleen enlargement and anemia. However it should be appreciated that such mechanistic hypothesis is speculative, because we have not included the simulation of mtDNA turnover of any fast dividing tissues in the present work. Treatment of cell division and selection pressure for mitochondrial turnover might be a promising area of investigation for the future work.
By thinking carefully about the different sources of stochasticity in each process from early development all the way to experimental sampling, we have identified the RMC assay procedure as a major contributor to the overall uncertainty. In contrast to the original interpretation of the data, our analysis reveals that the existence of an exponential dynamics in point mutations cannot be inferred with certainty, and thus no contradiction between the observed point mutation dynamics and the apparent absence of evidence for elevated oxidative stress exists. A detailed, quantitative understanding of the relevant sources of noise also allows optimization of experimental designs, thereby opening avenues for maximizing information return and minimizing cost, time and animal use.
The fact that the reproduction of the POLG mouse data requires no modifications to the wild type model, other than the obvious decrease of the polymerase fidelity, suggests that elevation of the point mutation burden does not trigger fundamentally new processes. In particular, neither mutant replicative advantage nor the elevation of the ROS dynamics resulting from the increase of the point mutation burden is required to explain the POLG data. This is consistent with our current view on the mFRTA [4], showing little evidence for the existence of vicious cycle mechanism. Two further observations related to the POLG mice that have originally been seen as somewhat surprising, can also be explained. The first is the observation that dividing tissues seem to be more severely affected in POLG mice than postmitotic tissues [9],[63]. The second is the fact that most mtDNA mutations in the POLG mice are already present at birth with comparatively little further accumulation during adult life, when compared to its development [9],[62]. Quantitative analysis shows both of these observations to be consequences of the low turnover of mtDNA in postmitotic tissues of adult mice.
Finally, our in silico analysis reveals the importance of early development in determining the trajectory of mtDNA mutation burden. This is in sharp contrast to the common assumption that health and diseases are determined predominantly by the genome interacting with the environment. Here, we have demonstrated that in silico modeling can contribute significantly to analysis and understanding of experimental data as well as potentially help to design more effective methodology. We believe that this approach of “Computer Aided Thought” can contribute towards a fundamentally improved understanding of intrinsically challenging biological problems such as aging.
|
10.1371/journal.pntd.0003512 | Evidence of Reversible Bradycardia and Arrhythmias Caused by Immunogenic Proteins Secreted by T. cruzi in Isolated Rat Hearts | Chagas cardiomyopathy, caused by the protozoan Trypanosoma cruzi, is characterized by alterations in intracellular ion, heart failure and arrhythmias. Arrhythmias have been related to sudden death, even in asymptomatic patients, and their molecular mechanisms have not been fully elucidated.
The aim of this study is to demonstrate the effect of proteins secreted by T. cruzi on healthy, isolated beating rat heart model under a non-damage-inducing protocol.
We established a non-damage-inducing recirculation-reoxygenation model where ultrafiltrate fractions of conditioned medium control or conditioned infected medium were perfused at a standard flow rate and under partial oxygenation. Western blotting with chagasic patient serum was performed to determine the antigenicity of the conditioned infected medium fractions. We observed bradycardia, ventricular fibrillation and complete atrioventricular block in hearts during perfusion with >50 kDa conditioned infected culture medium. The preincubation of conditioned infected medium with chagasic serum abolished the bradycardia and arrhythmias. The proteins present in the conditioned infected culture medium of >50 kDa fractions were recognized by the chagasic patient sera associated with arrhythmias.
These results suggest that proteins secreted by T. cruzi are involved in Chagas disease arrhythmias and may be a potential biomarker in chagasic patients.
| Chagas disease, caused by the parasite Trypanosoma cruzi, is an endemic disease of Latin-American countries, affecting an estimated 8 million people in 21 countries. It is spread by the bite of triatomine reduvid bug. Due to immigration towards non-endemic regions, the disease can spread and affect people around the world via blood transfusions. Infection usually occurs in childhood, and some patients may develop acute myocarditis; however, most remain asymptomatic for many years before chronic cardiac and/or gastrointestinal manifestations appear. Chagas disease is characterized by an acute phase, which is generally asymptomatic, or oligosymptomatic, an indeterminate phase, which may persist for several years, and a chronic phase in which dilated cardiomyopathy and arrhythmias are primarily observed and sudden death may occur. Once heart failure develops, death usually occurs within several years. In this work, we demonstrate the pathophysiological role of proteins secreted by T. cruzi on cardiac arrhythmias. The antigenicity of these fractions was tested by an immunological test using chagasic patients’ sera associated with arrhythmias. We showed that perfusion of the proteins secreted by T. cruzi, in an isolated beating rat heart model, induced cardiac arrhythmias such as bradycardia and complete atrioventricular block.
| Chagas disease is an important public health problem in Latin America currently affecting an estimated 8 million people in 21 countries and spreading by human migration to a number of non-endemic regions [1]. The protozoan Trypanosoma cruzi is the etiologic agent of the disease in mammals; the parasite is transmitted by blood-sucking triatomine bugs, blood transfusions or trans-placentally [2]. This illness is characterized by an acute phase, which is generally asymptomatic, or oligosymptomatic, an indeterminate phase, which may persist for several years, and a chronic phase in which dilated cardiomyopathy and arrhythmias are primarily observed. Chagas cardiomyopathy has been attributed to an alteration in intracellular ions, an imbalance between adrenergic and cholinergic innervations, to cellular and humoral autoimmunity and to parasitic effects or micro-ischemic disturbances [3].
Cardiac arrhythmias are one of the most important alterations in Chagas heart disease and may be associated with sudden death [4]. The principal disorders reported are atrial, ventricular extrasystoles, intraventricular and/or AV conduction disturbances and primary ST-T wave changes [5]. Classically, arrhythmias have been linked to autonomic dysfunction [6], anti-adrenergic and anti-cholinergic autoantibodies [7] and to wall motion abnormalities [8]. Although, Chagas patients may present with arrhythmias and sudden death in the absence of ventricular dysfunction (known as the arrhythmogenic form) [9], the causes associated with nonstructural arrhythmias are poorly understood. Notably, ST and T abnormalities, ventricular and supraventricular arrhythmias and low voltage QRS have been reported in a recent acute outbreak characterized by high parasitemia [10], which suggests that the secreted proteins of the parasite may be involved in arrhythmia generation.
The interaction between the parasite and the host cell has gained attention in the pathophysiology of Chagas disease. Parasite surface proteins, such as mucins, trans-sialidases and mucin-associated proteins (MAPs), are adhesion factors involved in the invasion of the host cell [11]. Additionally, these proteins are able to increase the intracellular calcium concentration to facilitate the entry of the parasite [12,13]. Interestingly, a recent report described a calcium overload in the ventricular myocytes of chagasic patients [14] that may be related to parasite signaling and could be responsible for the arrhythmias observed in Chagas disease. However, there are few reports that have linked protein secretion by T. cruzi with arrhythmias [15]. Consequently, the aim this study is to demonstrate that proteins present in T. cruzi-conditioned medium are able to produce arrhythmias in an isolated beating rat heart model.
Written consent from all patients involved in this study was obtained prior to processing the samples. The collection of serum samples from adult Chagas disease patients’ was approved by the Bioethics Committee for Human Research of the Universidad Centro Occidental Lisandro Alvarado, Barquisimeto, estado Lara, Venezuela (IVIC-DIR-0480/09). Data on human subjects was analyzed anonymously and clinical investigations have been conducted according to the Declaration of Helsinki.
For animal experimentation the project was also approved by the COBIANIM (IVIC-DIR-0376/1509/2014). COBIANIM is an advisory body of IVIC, with national reach, in regards to the ethical use of animals in research, in accordance with national and international standards. This committee oversees all research activities at IVIC, requiring the use of animals and wildlife to meet with Venezuelan law and universal ethical values. The Commission assessed the methodological, bioethical and legal aspects of this project (by resolution IVIC/Nro. 127/November, 4, 2009, according to the Código de Bioética y Bioseguridad, Ministerio del Poder Popular para Ciencia, Tecnología e Industrias Intermedias Fondo Nacional de Ciencia, Tecnología e Innovación, 2008 and, Guide Laboratory Animals for care and use, Eighth Edition, www.nap.edu, see details in http://www.ivic.gob.ve/cobianim/?mod=proyectos.php).
Sprague Dawley IVIC female rats (300–400 g.) were obtained from the Bioterio Services at IVIC. The animals were allowed to acclimate for 2 weeks prior to the study. They were housed in a clean wire mesh cages (10 rats per cage) and maintained under standard laboratory condition of 12 hours natural light and 12 hours darkness at ambient room temperature. The rats were fed on pellets and water was made available ad libitum.
Vero (African green monkey kidney cells, ATCC CCL-81; American Type Culture Collection, Rockville, Md.) were maintained at 37°C and 5% CO2 in complete Minimum Essential Medium (Mediatech, Herndon, Va.) containing 10% heat-inactivated fetal bovine serum (FBS; Gibco-BRL, Gaithersburg, Md), 20 mM HEPES, 2 mM L-glutamine, 1 mM sodium pyruvate, and 50 μg of gentamicin/ml (all Sigma-Aldrich, St Louis, MO). Confluent Vero cultures plated in a 75 cm2 Easy Flask were infected with 2x105 EP strain trypomastigotes/ml. EP human strain at a rate of 2 parasites per cell. The EP strain of T cruzi was isolated from a fatal human case in 1967 as describe by Contreras et al [16]. Free parasites were removed after 24 hours and the complete medium was changed at this point to medium FBS-free. The fifth or sixth day post-infection the conditioned serum-free medium was collected. The criteria for harvesting of the conditioned infected medium (CMi) were that a minimum of 75% of the Vero cells should remain adhered and that at least 2.5 x 106 trypomastigotes/ml should be present in the supernatant. The medium was centrifuged at 3000 x g for 10 minutes to separate the parasites, and the supernatant was subsequently filtered using a 0.2 μm membrane filter (Millipore, Billerica, MA, USA) before being stored at -20°C until use. Control medium (CMc) was collected from uninfected Vero cells cultured under the same conditions.
Cells were enzymatically detached with a 1:1 mixture of trypsin (0.25% v/v) and EDTA (0.25%), were lysed at 4°C with ten rounds of sonication for 20 s at power level 3 with a 550 Sonic Desmembrator (Fisher Scientific, Pittsburgh PA, USA). The suspension was centrifuged at 3000 x g for 10 minutes at 4°C to separate the cell debris, and the supernatant was subsequently filtered using a 0.2 μm membrane filter (Millipore, Billerica, MA, USA) before being stored at -20°C until use.
The supernatant fluid which had been decanted was concentrated by using low-protein binding membrane Diaflo (Millipore, AmiconCorp., Cambridge, Massachusetts) ultrafiltration cell operated in a cold room (4°C) at 50 psi. The Diaflo Model 50 ultrafiltration cell is provided with internal stirring and has a capacity of 40 ml. Supernatant from T. cruzi-infected cells (henceforth CMi) and those from uninfected Vero cells (henceforth CMc) SFV-free were passed through a ultrafiltration membranes. In each round, the fraction was concentrated to a volume of 10 ml (4x) and washed three times with milli-Q water to remove the lower molecular weight proteins. The pore size cutoff used were 300, 50, and 10 Kda, and the ultrafiltrates obtained were a) >300 Kda, b) <300 and > 50 Kda and c) <50 Kda [17].
After wash, each fraction was reconstituted to a final protein concentration of 5 μg/ml using Minimum Essential Medium (MEM; 1.8 mM CaCl2, 0.81 mM MgSO4, 5.33 mM KCl, 117.24 mM NaCl, 1.0 mM NaH2PO4, 5.56 mM D-Glucose, with L-Glutamine, Phenol Red and essential amino acids). The osmolalities of the reconstituted medium were measured using an osmometer (model Osmette A, Precision Instruments, Sudbury, MA) and were adjusted at 295–300 mOsm and pH to 7.4 with HCl. Finally, three concentrated fractions, corresponding to three molecular weight ranges, were obtained for the CMc and the CMi as follows: unfiltered, > 50 kDa and < 50 kDa.
For beating heart experiments, hearts were removed from adult female Sprague Dawley rats (weighing 300–400 g) previously anesthetized via intraperitoneal injection of pentobarbital (40 mg/kg). The isolated hearts were placed in cold Tyrode’s solution (25 mM sodium bicarbonate, 10 mM glucose, 116 mM NaCl, 3.3 mM KCl, 2.5 mM Ca2Cl, 1 mM MgSO4. and cannulated through the aorta. The hearts were perfused in a retrograde manner with warm Tyrode’s solution (37°C) for 30 minutes. The heart rate data were collected using the isolated beating heart system (PowerLab ADInstruments, Sydney, NSW, Australia). The presence of a sinus rhythm, a heart rate (HR) greater than 160 bpm, a perfusion pressure higher than 30 mmHg, and a flow rate of 2–6 ml/min were considered to be stable values for all experiments (Fig. 1, panel A).
The isolated hearts were perfused with conditioned media in accordance with a protocol consisting of three consecutive recirculation-reoxygenation cycles (Fig. 1, panel B). During the first recirculation stage, 5 mL of deoxygenated MEM was perfused in a closed circuit at a flow rate of 6 mL/min. To avoid cardiac cell damage due to anoxia, a 2 mL/min parallel flow of oxygenated Tyrode’s solution was added, whereby a standard flow rate of 6 mL/min was achieved with partial oxygenation. During the second recirculation stage, the hearts were perfused with: 1) Vero cell lysate or >50 kDa fraction (5 μg/mL) of 2) CMc (n = 5), 3) CMi (n = 8) and 4) CMi preincubated with serum samples of chagasic patients (n = 4), henceforth CMi+S. During the third stage, the hearts were perfused with MEM. Hearts in treatment received stabilization perfusion by one reoxygenation stage. The reoxygenation stage consisted of perfusing the heart with oxygenated Tyrode’s solution at a rate of 6 mL/min. All experiments were carried out under pression and perfusion controlled condition.
In this study, we used a Bolztmann equation for fitting to HR kinetics during recirculation-reoxygenation protocol as previously described [15] to model the I/R process.
A panel of Venezuelan Chagas patients’ samples of a sera collection compiled from 2002 to 2006 by the Laboratory of Parasites Physiology and the Cardiology Service of Instituto de Medicina Tropical, Universidad Central de Venezuela, was used for physiological assays. Inclusion criteria were Chagas disease diagnosed by two distinct Chagas tests, serological diagnosis was made with Cruzi ELISA kit as recommended by the manufacturer [18] and polymerase chain reaction (PCR) using a high conservation of DNA kinetoplast sequences (S35 5’-AAA TAA TGT ACG GG(T/G) GAG ATG CAT GA-32and S36 5’-GGG TTC GAT TGG GGT TGG TGT-3′) in T. cruzi allows detection of the parasite by means of an amplicon of 330 bp [19]. In this study, we used sera from Chagasic patients’ classified as Class II according to the New York Heart Association (NYHA) functional classification system: A serum of Class II Chagasic patients’ without arrhythmias was used as control in western blot analysis.
To demonstrate preservation of myocardial function ex vivo during recirculation-oxygenation process, we determined aspartate aminotransferase (AST) activity by sampling the effusate. These activities were carried out in 10 μl aliquots of coronary effluent [20] in order to provide indication of cardiac damage. The samples were collected at the end of each recirculation-reoxygenation cycle. Each sample was immediately stored at -20°C until the analysis was performed. The enzyme activity was measured using a commercially available ELISA kit (Invelab, Caracas, Venezuela) in accordance with the manufacturer’s protocol.
SDS—PAGE was performed in 6–18% polyacrylamide gradient gels according to the method of Laemmli [21]. Proteins separated by SDS—PAGE were transferred to Immobilon-P filters (Millipore, USA). Fifteen micrograms of protein was loaded into each lane. Membranes were stained with Ponceau red to verify the transfer of the proteins. The membrane was incubated one hour with chagasic sera at a dilution 1/100 and washed three for five minutes in PBS and 0.05% Tween 20. The secondary antibody incubation was performed with peroxidase-conjugated goat anti-human immunoglobulin, diluted 1:5000. Immunoblots were developed by using diaminobenzidine (Sigma-Aldrich). Proteins were quantitatively assayed by the Lowry’s method as modified by Schacterle et al [22] with bovine serum albumin as standard.
In order to quantify the relationships between continuous variables, QT, PR intervals and heart rate at the different experimental conditions, a Canonical Analysis of populations (CAP) were performed, then we performed a projection into a maximum information subspace which is known as canonical biplot analysis. CAP calculates the highest possible correlation between linear combinations of the values of studied variables, as within as between the a priori generated groups of individuals. Mean squared error (MSE), confidence interval and parameters functions was estimate on the parametric bootstrap methods for bias correction in linear mixed model (n = 250) [23]. Bootstraps were performed using InfoStat professional version (http://www.infostat.com.ar/) and the canonical biplot were carried out with MULTBIPLOT software [24].
ECG recordings from hearts perfused with the different media are shown in Fig. 2. The panels A, B, C, D and E represents the ECG recordings of five hearts, basal recording corresponds to the internal control condition of each ECG heart. The hearts were subjected to three-recirculation and reoxygenation cycles (see fig. 1B). The recordings corresponding to the perfusion of hearts with either Vero cell lysate (Fig. 2A) or CMc (Fig 2B) did not show significant change at any stage of perfusion. In hearts perfused with CMi (Fig. 2C and 2D), a complete AV block with prolonged asystole was observed, together with an episode of ventricular fibrillation. In Fig. 2, panel E is a representative experiment of four co-incubations tested independently, of T. cruzi infected Vero cells conditioned media plus chagasic patients’ sera (CMi+S). Notably, it should be noted that chagasic patient’s sera preincubation abolished the observed effects when the hearts were perfused with >50 kDa proteins Cmi. Also, a partial reversion was observed in the 3rd cycle, and a total reversion was achieved with Tyrode perfusion (Fig. 2E). We did not observe any abnormal cardiac conduction effects in the ECG recordings taken from hearts perfused with the <50 kDa fractions. It is known that the parasites secrete several bioactive lipids and glycolipids, thus lipid compounds were removed from conditioned medium and tested, and did not observe any abnormal cardiac conductions effect.
Fig. 3A shows the percentage of heart rate in three frame times in relation to the recirculation-reperfusion cycles. Three groups were defined by the heart rate in different frame times: Group 1, from 0 to 170 s, Group 2, from 171 to 280 s and, Group 3, higher than 280 s. The adjusted model represents the effect on heart rate kinetics for the three experimental conditions that resulted from perfusing hearts with the different fractions >50 kDa corresponding to CMc, CMi and CMi+S (S2 Table Rawdata Fig. 3A).
The adjusted linear mixed models (LMMs) shows that significant differences between conditions, times and its interaction (p≤0.05). The lowest heart rate percentage was found for the CMi and the highest for CMc. The average heart rate corresponds to 94.21 ± 1.16; 135.77 ± 1.18 and 150.61 ± 0.78 percent’s, for Groups 1, 2 and 3, respectively (see Fig. 3A). The comparison from CMi curve with Boltzmann model shows, in group 1, a significant decrease of initial HR parameter. These alterations, especially bradycardia, may be related to AV blockade observed in EGC (Fig. 2C). Remarkably, these changes were reversed during the reoxygenation stage of the same heart (Fig 2C and 2D) and/or by incubation of CMi+S (Figs. 2E and 3A).
Additionally, the Boltzmann analysis shows discrete changes of inflection times in hearts perfused with CMi and CMi+S, suggesting a HR recovery delayed effect during the early re-oxygenation stage probably related to immunogenic proteins (Fig 3, Groups 2 and 3). Also, we found difference between the initial HR values in the recirculation cycle in Cmi, this result suggests an involvement of secreted proteins in bradycardia. This bradycardia was independent of the QT interval over 450 s (Figs. 3A and 3B).
The Figs. 3B and 3C show the QT and PR segments estimated from EGC of isolated hearts perfused with CMc, CMi and CMi+S. This estimation was carried out over 2700 s. In both, QT and PR segments by LMMs adjusted analysis showed that condition CMi had the most variability observed compared with control condition (p≤0.05). The lowest QT interval was found for the CMi+S and the highest for CMi. The highest PR interval was observed was found for the CMi. This model demonstrates a sinusoidal behavior that cannot be associated with a tendency towards an increased PR interval (Fig. 3C).
To evaluate cardiac cellular damage during the recirculation-reoxygenation protocol, we collected the media after each recirculation-reoxygenation cycle and measured AST activity (Fig. 4). The arrows identify the times at which the samples were taken. Rc1: recirculation 1; Rc2: recirculation 2; Rc3: recirculation 3; Ro1: reoxygenation 1; Ro2: reoxygenation 2; Ro3: reoxygenation 3. We observed that AST activity remained at basal levels during the entire protocol in hearts perfused with proteins >50 kDa CMi and CMc, which suggests the absence of cardiac cellular damage.
Fig. 5 describes the projection at different times of the variables and, the three conditions studied such as CMc, CMi and CMi+S, whose correlation was determined by Mahalanobis distance. The sum of percentages of two axis selected in the canonical biplot explain almost 100% of the variance with a very high quality of representation of the groups CMc (91.6%) and CMi (99.3%) in the canonical axis 1 and of CMi+S (56.3%) in the canonical axis 2 (S1 Table Rawdata canonic biplot).
The goodness of fit of the variables into the canonical subspace demonstrate that 16 of 55 variables were represented with quality over 80% (HR170:93.02, HR120: 91.2, HR180: 90.37, HR130: 89.79, HR160: 89.55, HR60: 88.96, HR150: 88.69, PR2700: 88.23, HR70: 86.77, HR50: 86.16, PR900: 85.41, HR30: 83.48, HR40: 83.43, HR80: 83.39, PR1200: 83.1, HR140: 82.18) 81.25% of the variables are related to the heart rate, and the rest are related to PR. Besides, 48 of 55 (87.27%) variables has a quality of representation over 50%, and 100% of the individuals were has a quality of representation above 85% in the canonical subspace.
To initiate characterization of the proteins in conditioned media that could be responsible for the observed effects on cardiac conduction, we performed western blotting with serum from chagasic patients’. As shown in Fig. 6, chagasic serum recognized various ultrafiltrate proteins. More antigenic proteins are in the range of molecular weight 50–200 kDa. In all cases, proteins in CMc were non-antigenic. Notably, chagasic serum obtained from patient without arrythmias did not reveal any bands.
In developing countries a proportion of chagasic patients die undiagnosed. Developing models to understand the pathophysiology of this disease and test new therapies are necessaries. The role of the secretome of the T. cruzi has been gaining attention. Traditional use of secretome protein in serodiagnosis, i.e. antigenic proteins detection by ELISA test called Trypomastigote Excreted-Secreted Antigens (TESA). There are currently no studies which associate T. cruzi secretome proteins as biomarkers for cardiac arrhythmias in Chagas cardiomyopathy. Recently, Wen et al (2012) [25] resolved the proteome signature of high and low abundance serum proteins in chagasic patients demonstrating the serum oxidative and inflammatory response profile, and serum detection of cardiac proteins parallels the pathologic events contributing to Chagas disease development. In this manuscript, we established a reproducible model using a recirculation-reoxygenation protocol that resembles the early interactions that occur between parasite secretome with cardiac cells. We found that CMi induces a biological effect on healthy, isolated beating rat heart model similar to those observed in Chagas patients. These effects may be related to the direct interaction of proteins into CM on heart cells. Some authors have demonstrated the persistence of the parasite in chronic lesions in patients [26], which reinforces the hypothesis that tissue damage is related to cellular parasitism in vivo. The detection of significant neuronal cell loss in the sympathetic and parasympathetic nervous systems of Chagas cases, in the absence of T. cruzi in situ, is the basis for the hypothesis of factors released from the parasite nest hidden somewhere in the host body, producing cell damage. [2]. Additionally, there is good in vitro and in vivo evidence for autoantibodies against neuroreceptors (beta-adrenergic and muscarinic) in Chagas disease [7,27]. In this study, we evaluated the role of the secretome fractions of T. cruzi co-cultured with Vero cells on cardiac arrhythmias using an isolated beating rat heart model. We observed bradycardia, ventricular fibrillation and complete atrioventricular block in hearts during perfusion with >50 kDa CMi. The antigenicity of the secreted proteins was tested by Western blotting using chagasic patient’s sera. The effects observed in this in vitro heart model are different from the results observed in autoimmunity studies [2], since we detected that immunogenic T. cruzi-secreted proteins was able alter cardiac function independent of a systemic immune response. It should be noted that chagasic patient’s sera preincubation abolished the observed effects when the hearts were perfused with >50 kDa proteins CMi, confirming the relationship between cardiovascular alterations, the immunogenic T. cruzi-secreted proteins and its correlation with arrhythmias in Chagas disease.
We have previously [15] demonstrated that cardiac damage can be estimated based on the amount of AST released into the coronary effluent. In the present study, we carried out serial AST enzymatic measurements as an internal control. The enzymatic activity remained stable through the recirculation-reoxygenation protocol, indicating lack of induced myocardial tissue damage.
The T. cruzi invasion has been linked to the interaction of parasite membrane glycoproteins with cellular ligands and their associated signaling pathways [28]. Several members of the T. cruzi-mucin family (TcMuc) have been linked to the process of invasion and to the increase in intracellular calcium concentration in particular. The secretion of MASP52, a member of the mucin associated protein (MASPs) family, has been associated with parasite attachment and with parasite invasion of Vero cells [29]. A recent report characterized the T. cruzi secretome obtained from medium conditioned by culturing the epimastigote and metacyclic trypomastigote forms under axenic conditions [30], in this study, it was demonstrated that proteins are shed in vesicles and that 3.8% of the secreted proteins are involved in parasite-cell interactions. The study identified the surface glycoprotein GP90, MASP52, trans-sialidase, and an 82 kDa glycoprotein, along with additional proteins, as being secreted by T. Cruzi (Table 1). However, these reports did not evaluate any pathophysiological role of this group of proteins.
Ventricular arrhythmias in Chagas patients’ are related to calcium overload [14]. Accordingly the relationship between proteins secreted by the parasite and the regulation of intracellular calcium levels could provide insight into the arrhythmias observed in isolated beating hearts perfused with high molecular weight T. cruzi proteins. Previously, T. cruzi GP82, [13] GP90 and GP35/50 proteins [31] have been described as involved in both the modulation of calcium increase in the host cell and in determining the invasiveness of the parasite strain. Our group [15], demonstrated that T. cruzi-conditioned medium was able to increase the frequency of occurrence of tachyarrhythmia and cause a decrease in the heart rate during post-ischemic recovery. A novel approach recently reported by Elliott et al [32] showed that Trypanosoma brucei cathepsin-L supernatant disturbs the heart electrical activity, leading ventricular premature complex (which cause palpitations) and triggers arrhythmias in whole rat heart. In the present work, we observed a reversible ventricular fibrillation and a total AV block associated with bradycardia in a non-damage-inducing protocol, the effect was reversed by incubating the infected conditioned medium with chagasic patient’s serum, confirming that a direct interaction between the parasite secreted proteins and cardiomyocytes exist in the pathophysiology of Chagas cardiomyopathy. It is plausible that pro-arrhythmogenic proteins secreted or released by T. cruzi could act as enhancers causing the cardiac conduction system to cross an arrhythmic threshold in Chagas patients. This is the first report that implicates proteins secreted by T. cruzi with arrhythmias in an ex-vivo model.
We obtained a reproducible pattern of antigenic recognition of T. cruzi-secreted proteins by patient’s sera suggesting that immunogenic T. cruzi-secreted proteins are implicated with arrhythmias in Chagas disease. T. cruzi proteins could be used as virulence markers in the prognosis of the arrhythmias in Chagas patients.
To support the secretome proteins interaction a statistical analysis was carried out in order to quantify the relationships between continuous variables, QT and PR intervals and heart rates at the different experimental conditions (CMc, CMi and CMi+S). We decided to perform a Canonical Analysis of Populations (CAP). This methodology is used to project in a biplot simultaneously the structure of the a priori generated groups and the variables responsible for the separation between them. The CAP is used to obtain canonical axis that reflects the maximum separation between groups, not between individuals.
This statistical method has been used in other fields such as econometrics, and recently it has been used in biological systems was reported [33]. In this type of analysis there are two assumptions that should be tested, 1) the mean vectors of the groups must be significantly different. This assumption is evaluated by the study of global contrast based on Wilk’s Lambda (L) which turned out to be statistically significant 132.0474 with a p-value of 4.3475e-31. 2) the variances-covariances of the variables of the groups must be equal. The three variances-covariances matrices turned out to be singular which means that the variables have linear relation between them so it is possible to reduce the dimensionality of the system by eliminating variables. This was corroborated by the fact that with two dimensions we can explain almost 100% of the system variance, but it is important to remark that despite of the high quantity of redundant variables, they were ignored because of the use of Mahalanobis distance.
Canonical analysis of populations proved that 87.27% of the variables were represented with a quality above 50% and, 100% of the individuals were represented with a quality above 85%. This methodology permits us to elucidate two important aspects, first, we demonstrate that with just 16 variables we can explain over the 80% of the information and 13 of them are heart rate variables and the other three are PR variables. This means that the infection phenomena relative to the secretome proteins generated by the host-pathogen interaction could be successfully followed in time just by studying the heart rate. Besides not all time-points are necessary, just the mentioned above. Second, the group projections in the biplot analysis are in agreement with the results shown in Figs, 2 and 3, mainly in Fig. 2A which we can see the heart rate kinetics. We can infer that there is a significant recovery from bradycardia when the conditioned infected medium were mixed with chagasic patients serum, in other words the secretome proteins could be responsible for the heart dysfunction observed. We are already working on identifying the proteins of the secretome and how are their relationships with the dysfunction of the heart.
The contribution of this study was to evaluate, in an isolated heart model, the arrhythmogenic role of the parasite secretome proteins. This reproducible recirculation-reoxygenation model can be an useful system to investigate new drugs for the treatment of Chagas disease arrhythmias. The dissection and simplification of a complex system as constituting Chagas infection becomes necessary to understand and control the disease, in this sense, we wanted to contribute with a simplistic model which evaluates arrhythmogenic factors. This work represents a heuristic contribution to the study of arrhythmias produced by immunogenic proteins secreted by Trypanosoma cruzi in vitro. The findings of this study provide a glimpse into the role of this parasite’s secretome in the pathogenesis of the Chagas’ disease.
|
10.1371/journal.pcbi.1004137 | The Neuroglial Potassium Cycle during Neurotransmission: Role of Kir4.1 Channels | Neuronal excitability relies on inward sodium and outward potassium fluxes during action potentials. To prevent neuronal hyperexcitability, potassium ions have to be taken up quickly. However, the dynamics of the activity-dependent potassium fluxes and the molecular pathways underlying extracellular potassium homeostasis remain elusive. To decipher the specific and acute contribution of astroglial Kir4.1 channels in controlling potassium homeostasis and the moment to moment neurotransmission, we built a tri-compartment model accounting for potassium dynamics between neurons, astrocytes and the extracellular space. We here demonstrate that astroglial Kir4.1 channels are sufficient to account for the slow membrane depolarization of hippocampal astrocytes and crucially contribute to extracellular potassium clearance during basal and high activity. By quantifying the dynamics of potassium levels in neuron-glia-extracellular space compartments, we show that astrocytes buffer within 6 to 9 seconds more than 80% of the potassium released by neurons in response to basal, repetitive and tetanic stimulations. Astroglial Kir4.1 channels directly lead to recovery of basal extracellular potassium levels and neuronal excitability, especially during repetitive stimulation, thereby preventing the generation of epileptiform activity. Remarkably, we also show that Kir4.1 channels strongly regulate neuronal excitability for slow 3 to 10 Hz rhythmic activity resulting from probabilistic firing activity induced by sub-firing stimulation coupled to Brownian noise. Altogether, these data suggest that astroglial Kir4.1 channels are crucially involved in extracellular potassium homeostasis regulating theta rhythmic activity.
| Neural excitability relies on precise inward and outward ionic fluxes. In particular, potassium ions, released by neurons during activity, have to be taken up efficiently to prevent hyperexcitability. Astrocytes, the third element of the synapse, play a prominent role in extracellular potassium homeostasis. Thus unraveling the dynamics of the neuroglial potassium cycle during neurotransmission and the underlying molecular pathways is a key issue. Here, we have developed a tri-compartment model accounting for potassium dynamics between neurons, astrocytes and the extracellular space to quantify the specific and acute contribution of astroglial Kir4.1 channels to extracellular potassium levels and to the moment-to-moment neurotransmission. We demonstrate that astroglial Kir4.1 channels are sufficient to account for the slow membrane depolarization of astrocytes and crucially contribute to extracellular potassium clearance during basal and high activity. We also show that astrocytes buffer in less than 10 seconds more than 80% of the potassium released by neurons, leading to recovery of basal extracellular potassium levels and neuronal excitability. Remarkably, we found that Kir4.1 channels also prominently regulate slow 3 to 10 Hz rhythmic firing activity. Altogether, these data show that Kir4.1 channels acutely regulate extracellular potassium and neuronal excitability during specific patterns of activity.
| Astrocytic processes enwrap more than half of CA1 hippocampal synapses to form tripartite synapses [1,2]. Perisynaptic astroglial processes are enriched in ionic channels, neurotransmitter receptors and transporters, enabling astrocytes to detect neuronal activity via calcium signaling [3] and ionic currents with various components, such as glutamate and GABA transporter [4–7] or potassium (K+) [8–10]. Thus astrocytes regulate neuronal activity through multiple mechanisms, involving signaling or homeostasis of extracellular space volume, glutamate, GABA or K+ levels [11]. Interestingly, membrane depolarization was the first activity-dependent signal identified in glial cells and was attributed to K+ entry across their membrane [10]. Such K+ entry was suggested to contribute to K+ spatial buffering, consisting in glial uptake of excess extracellular K+ ([K+]o), redistribution via gap-junction astroglial networks and subsequent release at sites of low [K+]o [12].
Modeling studies have mostly investigated astroglial regulation of [K+]o during pathological conditions to clarify its impact on aberrant neuronal activity. In particular astrocytes, by regulating [K+]o, have been shown to contribute to initiation and maintenance of epileptic seizures [13–15], as well as to the severity of ischemia following stroke, with a neuroprotective or neurotoxic role, depending on [K+]o [16,17]. In addition, experimental data suggest that several K+ channels or transporters contribute to astroglial K+ clearance, such as inward rectifier 4.1 and two pore K+ channels (Kir4.1 and K2P, respectively) or Na/K ATPases [18,19]. Remarkably, recent work suggest that Kir4.1 channels play a prominent role in astroglial regulation of [K+]o [20–23]. However, the mouse model used to draw these conclusions, i.e. conditional Kir4.1 knockout mice directed to glial cells (GFAP-Cre-Kir4.1fl/fl mice, Kir4.1-/-), exhibits several limitations: 1) Kir4.1 channels are not specifically deleted in astrocytes, but also in other glial cells such as oligodendrocytes or retinal Müller cells [22]; 2) astrocytes are severely depolarized [21,22]; 3) Kir4.1-/- mice die prematurely (~3 weeks) and display ataxia, seizures, hindleg paralysis, visual placing deficiency, white matter vacuolization and growth retardation [22], highlighting that chronic deletion of Kir4.1 channels induces multiple brain alterations and possibly compensations. Thus, the specific and acute contribution of astroglial Kir4.1 channels to [K+]o and to the moment to moment neurotransmission is still unclear. To decipher the acute role of astrocytes in controlling K+ homeostasis and neuronal activity, we built a tri-compartment model accounting for K+ dynamics between neurons, astrocytes and the extracellular space. We quantified K+ neuroglial interactions during basal and high activity, and found that Kir4.1 channels play a crucial role in K+ clearance and astroglial and neuronal membrane potential dynamics, especially during repetitive stimulations, and prominently regulate neuronal excitability for 3 to 10 Hz rhythmic activity.
To model K+ ions dynamics during neuronal activity, we built a biophysical model that includes three compartments: the neuron, the astrocyte and the extracellular space (Fig. 1A). As performed in several studies [13,16,24], the neuron is approximated by a single compartment conductance-based neuron containing Na+ and K+ voltage-gated channels, enabling action potential discharge. The associated neuronal membrane potential is coupled with the dynamics of intracellular and extracellular Na+ and K+ levels via the dependence of the neuronal currents to the Nernst equation. The ion concentrations depend also on the activity of neuronal and astroglial Na/K ATPases, which maintain resting [K+]i by balancing K+ and Na+ fluxes. Similarly, the astrocyte is approximated by a single compartment conductance-based astrocyte containing Kir4.1 channels, which are inward rectifier K+ channels strongly expressed in astrocytes that generate dynamic K+ currents [25]. In the model, neurons and astrocytes are separated by a homogenous extracellular space compartment. The model is based on balancing ionic fluxes between the three compartments (Fig. 1B). The model starts with the induction of a synaptic current (Iapp, see Materials and Methods). This current is the initial input of a classical Hodgkin-Huxley model, which describes the neuronal membrane potential dynamics (entry of Na+ and exit of K+). Released extracellular K+ is taken up by astrocytes through Kir4.1 channels and Na/K ATPases (Fig. 1B and Materials and Methods). Because Kir4.1 channels are strongly involved in K+ uptake [22], we fitted the I-V curve of K+ ions through Kir4.1 channels using equation 22 (see materials and methods) and predicted the I-V curve at various values of [K+]o (Fig. 1C). We obtain that K+ fluxes through Kir4.1 channels vanish around astrocytic resting membrane potential (~-80 mV) and are outward during astrocytic depolarization for a fixed [K+]o (2.5 mM, Fig. 1C). However, they become inward when [K+]o increases (5–10 mM, Fig. 1C). Using this model, we shall investigate quantitatively the contribution of Kir4.1 channels to K+ uptake in relation to neuronal activity associated with different [K+]o.
To validate our tri-compartment model, we compared simulation results with electrophysiological recordings. To account for the synaptic properties of CA1 pyramidal neurons, we generated a synaptic current (Iapp) using the depression-facilitation model (equation 1) (see Materials and Methods with input f(t) = δ(t)) (Fig. 2A,E,I). We first investigated responses to single stimulation. Using the Hodgkin-Huxley model, this synaptic current induces a firing activity (S1A Fig.), resulting in a ~ 0.9 mM increase of [K+]o within 300 milliseconds, which slowly decayed back to baseline levels during 10 seconds (S1B Fig.). The extracellular K+ dynamics was associated in our model with a small astrocytic depolarization of ∆V = −1.35 mV (equations 22, 23, 25) (Fig. 2C). Using electrophysiological recordings of evoked field excitatory postsynaptic potential (fEPSP) by a single stimulation of Schaffer collaterals in acute hippocampal slices (Fig. 2B), we measured astroglial membrane potential depolarization and found that it reached ~ 1.3 mV (1.3 ± 0.2 mV, n = 6) (Fig. 2C), confirming the result of our simulation. After validating the responses of the tri-compartment model to basal stimulation, we investigated the impact of trains of stimulations on the dynamics of astroglial membrane potential. During tetanic stimulation (100 Hz for 1 second), variations in neuronal membrane potential described by the Hodgkin-Huxley equation show a bursting activity during ~ 1 second (S1C Fig.). This is associated with a depolarization of astrocytic membrane potential of ~ 5 mV, which lasts ~ 6 seconds (Fig. 2G,H) and an increase in [K+]o that reaches a peak value of 4.4 mM (S1D Fig.).
For repetitive stimulations (10 Hz for 30 seconds), the neuron exhibited firing activity during the whole stimulation (S1E Fig.). This was associated with an astroglial depolarization of ~ 12 mV (Fig. 2K) and an increase in [K+]o peaking at 6.9 mM after 17.5 seconds of stimulation (S1F Fig.). Although the stimulation lasted 30 seconds, the astrocytic depolarization started to decay after 17 seconds (Fig. 2K).
The kinetics of astroglial membrane potential dynamics obtained with the numerical simulations are comparable to the results obtained with electrophysiological recordings performed in individual astrocytes during single stimulation (rise time: 48.4 ms for numerical stimulation, 42 ± 19 ms n = 6 for experiments; time of peak: 740 ms for numerical simulation, 730 ± 60 ms n = 6 for experiments; decay time: 3.67 s for numerical simulation, 4.50 s ± 0.2 n = 6 for experiments, Fig. 2D), tetanic stimulation (rise time: 610 ms for numerical simulation, 491 ± 122 ms n = 5 for experiments; time of peak: 1.07 s for numerical simulation, 1.05 s ± 0.25 n = 5 for experiments; decay time: 4.18 s for numerical simulation, 4.55 s ± 0.45 n = 5 for experiments, Fig. 2H) and repetitive stimulation (rise time: 1.5 s for numerical simulation, 1.27 s ± 0.18 n = 5 for experiments; time of peak: 6.8 s for numerical simulation, 5.2 s ± 0.9 n = 5 for experiments; decay time: 7.95 s for numerical simulation, 8.3 s ± 0.4 n = 5 for experiments, Fig. 2L). These data show that the dynamics of astroglial membrane potential changes obtained from numerical simulations and from electrophysiological recordings are similar. Thus our model captures the key players sufficient to mimic the evoked astroglial membrane potential dynamics observed experimentally in different regimes of activity.
We investigated the dynamics of the K+ cycle between neurons, extracellular space and astrocytes induced by neuronal activity to decipher the time needed to restore basal extracellular and intra-neuronal K+ levels. We studied K+ redistribution induced by single, tetanic (100 Hz, 1 s) and repetitive (10 Hz, 30 s) stimulations, and found that the general behavior of K+ dynamics was divided into three phases (phases 0, 1 and 2; Fig. 3).
During phase 0 (t = 0 to t1), neuronal K+ is released in the extracellular space (peak [K+]o during phase 0: 0.9 mM at 300 ms for single stimulation; 1.9 mM at 1.3 s for tetanic stimulation; 4.4 mM at 30 s for repetitive stimulation, Fig. 3A,D,G). Compared to basal K+ levels in each compartment (at t = 0), the relative transient evoked increase in K+ concentration is prominent only in the extracellular space (~+37% for single stimulation, +76% for tetanic stimulation and +168% for repetitive stimulation, Fig. 3B,E,H).
During phases 0 and 1, released K+ is then mostly buffered by astrocytes (~80 to 99% at the end of phase 1) during the different regimes of activity (time t2 (at the end of phase 1) for single stimulation: 8.2 s; tetanic stimulation: 8.7 s; repetitive stimulation: 34.2 s, Fig. 3C,F,I). The astroglial net K+ uptake increases with the activity-dependent [K+]o transient rises (S2A–C Fig.) evoked by the different regimes (S2D Fig.). Neurons slowly re-uptake only ~5–10% of their released K+ at the end of phase 1 (Fig. 3C,F,I). Remarkably, although [K+]o increases with the strength of stimulation (from 0.9 to 4.4 mM, Fig. 3A,D,G and S2A–C Fig.), the time needed for astrocytes to buffer the released K+ is not proportional to [K+]o rises (Fig. 3C,F,I), as shown by the phase diagram illustrating astroglial K+ uptake as a dynamic function of activity-dependent changes in [K+]o evoked by the different stimulations (S2D Fig.), but is to the square root of [K+]o (equation 22). In addition, at the end of phase 1, [K+]o is almost back to baseline levels, whereas intra-astroglial K+ levels reach their peak value (Fig. 3C,F,I).
Finally during phase 2 (t2 to end), astroglial buffered K+ is slowly redistributed back to neurons, which ends the K+ cycle. The long-lasting phase 2 is marked by an inversion of K+ fluxes in astrocytes, suggesting moderate K+ release by astrocytes over time. Indeed, K+ redistribution to neurons depends on K+ release through Kir4.1 channels, which is limited by the low outward rectification of these channels (Fig. 1C). Altogether, these data suggest a slow, but dynamic and efficient astroglial clearance capacity for the different regimes of activity.
To study quantitatively the acute and selective role of astroglial Kir4.1 channels in neuroglial K+ dynamics, we inhibited the Kir4.1 current in our tri-compartment model. Because Kir4.1-/- mice display altered synaptic plasticity compared to wild type mice [22,26], we recalibrated the synaptic current (Iapp) parameters τrec and τinact in equations 1,2 (see Table 1) for the facilitation-depression model to get an optimal fit to the recorded postsynaptic responses [26]. Another change in the model consisted in setting at zero both the Kir4.1 current and the leak term. In addition, to compensate for the loss of K+ fluxes through astroglial Kir4.1 channels, we added in equation 27 a constant K+ flux to maintain [K+]o at an equilibrium value of 2.5 mM. Consequently, the astrocytic membrane potential displayed no change during stimulation, in agreement with electrophysiological recordings [21,22].
The numerical simulations show that inhibition of astroglial Kir4.1 channels leads to higher transient peak increase in [K+]o during repetitive and tetanic stimulation compared to control conditions (Fig. 4E,F,I,J), while no difference is observed for single stimulation (Fig. 4A,B). In addition, for all regimes of activity, the rise and decay times of the [K+]o were increased when Kir4.1 channels were inhibited (single stimulation, control: rise time 136 ms, decay time 3.4 s; Kir4.1 inhibition: rise time 232 ms, decay time 4.2 s; tetanic stimulation, control: rise time 638 ms, decay time 4 s; Kir4.1 inhibition: rise time 753 ms, decay time 6 s; repetitive stimulation, control: rise time 6.8 s; Kir4.1 inhibition: rise time 20.2 s, Fig. 4B,F,J).
Finally, Kir4.1 channel inhibition only slightly increased neuronal firing induced by single stimulation (Fig. 4C,D) and tetanic stimulation (Fig. 4G,H), while it had major effect on neuronal excitability during repetitive stimulation (Fig. 4K,L). Indeed, although firing frequency was only slightly increased during the first 8 seconds of repetitive stimulation when [K+]o reached 10 mM (Fig. 4I), action potential amplitude and firing rate then progressively decreased due to neuronal depolarization (from-33 mV to-19 mV after 14 and 30 seconds of stimulation, respectively), suppressing neuronal firing after 14 seconds of stimulation (Fig. 4K). Altogether, these data show that astroglial Kir4.1 channels are prominently involved in K+ buffering during high level of activity, and thereby have a major impact on neuronal resting membrane potential controlling firing during trains of stimulations.
To investigate the effect of astroglial Kir4.1 channels on endogenous physiological rhythmic activity, we generated probabilistic firing induced by sub-firing stimulation coupled to neuronal Brownian noise (Fig. 5A,B). To simulate the firing activity, we generated a sub-firing periodic stimulation (5 ms squared stimulus), which defines the applied synaptic intensity in our tripartite compartment model (Fig. 5A), and added a Brownian noise in the neuronal membrane potential (equation 21, Fig. 5B). Such stimulation induces an increase in [K+]o (Fig. 5C), and thus firing over time (Fig. 5D-E). We found that astroglial Kir4.1 channels had no effect on the firing probability (computed over 100 simulations) for basal (0.1 Hz, Fig. 5F), low (1 Hz, Fig. 5G) and high (50 Hz, Fig. 5K) frequency stimulations. However, Kir4.1 channels directly regulate the firing probability for 3 and 5 Hz stimulations after 7 and 12 s of sub-firing stimulation, respectively (Fig. 5H,I). In contrast, Kir4.1 channels regulate only transiently the firing probability induced by 10 Hz stimulation (Fig. 5J). These data suggest a prominent and specific involvement of astroglial Kir4.1 channels in regulation of firing during theta rhythmic activity.
[K+]o modulates neuronal membrane potential, excitability, release probability and synaptic efficacy [27–32]. To unravel the acute role of astrocytes in extracellular K+ homeostasis and neuronal activity, we used electrophysiological recordings with a tri-compartment model accounting for K+ dynamics between neurons, astrocytes and the extracellular space. We found that Kir4.1 channels play a key role in extracellular K+ clearance, astroglial and neuronal membrane potential dynamics, especially during trains of stimulation, and strongly regulate neuronal excitability for slow rhythmic activity (3–10 Hz).
Several models have investigated extracellular K+ regulation of neuronal activity, including glial uptake mechanisms [13–17,24,33,34].
To study seizure discharges and spreading depression, a first tri-compartment model including the neurons, astrocytes and extracellular space was proposed [24], although the astrocytic membrane potential was not taken into account, and K+ accumulation in the interstitial volume was controlled by a first-order buffering scheme that simulated an effective glial K+ uptake system. With such model, after evoked firing, it took ~17 s for the neuronal membrane potential to return to resting values, via activation of Na/K ATPases. The model also predicted that elevated [K+]o have a key role in the initiation and maintenance of epileptiform activity. In our study, we accounted for the astroglial modulation of K+ buffering capacity regulated by its membrane potential, and found that the biophysical properties of astrocytic membranes including Kir4.1 channels were sufficient to account for the long-lasting clearance of extracellular K+. Interestingly, we confirm that alteration in K+ clearance leading to an extracellular K+ accumulation induces epileptiform activity, and show specifically that Kir4.1 channel acute inhibition leads to such pathological bursting activity during repetitive stimulation.
A similar tri-compartment model has been simplified as a one-dimensional two-layer network model to study how neuronal networks can switch to a persistent state of activity, as well as the stability of the persistent state to perturbations [13]. In this model, Na+ and K+ affect neuronal excitability, seizure frequency, and stability of activity persistent states. In particular, the quantitative contribution of intrinsic neuronal currents, Na/K ATPases, glia, and extracellular Na+ and K+ diffusion to slow and large-amplitude oscillations in extracellular and neuronal Na+ and K+ levels was revealed. In the model, the estimated [K+]o during epileptiform activity are comparable to the ones observed experimentally [35,36]. Although this model does not account for astroglial Kir4.1 channels, it shows that a local persistent network activity not only needs balanced excitation and inhibition, but also glial regulation of [K+]o [15].
Finally, a model accounting for the extracellular space and astroglial compartments has quantified the involvement of several astroglial ionic channels and transporters (Na/K ATPase, NKCC1, NBC, Na+, K+, and aquaporin channels) in the regulation of firing activity [34].
To account for K+ dynamics between neurons, astrocytes and the extracellular space, we built for the first time a tri-compartment model, where we included neuronal voltage-gated channels, Na/K pumps and astrocytic Kir4.1 channels according to their biophysical properties, as well as membrane potential of astrocytes. Because functional expression of voltage-gated calcium channels on hippocampal mature astrocytes in situ in physiological conditions and its impact on astrocytic functions is still a matter of debate [37], such channels were not included in our model. However, many other astroglial K+ channels (such as two pore domain K+ channels (K2P) (TWIK-1, TREK-1, TREK-2 and TASK-1), inward rectifier K+ channels (Kir2.1, 2.2, 2.3, 3.1, 6.1, 6.2), delayed rectifier K+ channels (Kv1.1, 1.2, 1.5, 1.6), rapidly inactivating A-type K+ channels (Kv1.4), calcium-dependent K+ channels (KCa3.1)), but also other channels, transporters or exchangers (such as Cx hemichannels, Na+/K+/Cl- co-transporter (NKCC1) K+/Cl- exchanger, glutamate transporters) [16,38,39] could also play a role in the regulation of activity-dependent changes in [K+]i or [K+]o. Functional evidence of the contribution of these channels, transporters or exchangers in astroglial K+ clearance is actually scarce, although K2P channels have been suggested to participate in astroglial K+ buffering [40], while NKCC1 were recently shown in hippocampal slices not to be involved in activity-dependent K+ clearance [41]. Similarly, adding slower timescale K+ dependent conductances in the neuron model could modulate the slow redistribution of K+ to neurons, and thus the duration of the neuroglial potassium cycle, and is of interest to implement in future development of the model. In our study, the aim was to simplify the system to capture in the model the minimal set of astroglial channels and pumps accounting for our experimental data related to activity-dependent changes in astroglial membrane potential. In addition our tri-compartment model, as most existing models, did not account for the complex multiscale geometry of astrocytes and neurons. Incorporating in our current model additional astroglial and neuronal channels, as well as complex cell geometry is of particular interest to identify modulatory effects of other specific channels and of microdomain geometry on the neuroglial potassium cycle.
In accordance with previous studies, where Kir4.1 channels were chronically deleted genetically in glial cells [20,21,23], we found that acute inhibition of Kir4.1 channels leads to altered regulation of extracellular K+ excess and affects the kinetics of [K+]o (Fig. 4I,J). However, in contrast to these studies, we found that Kir4.1 channel inhibition also alters significantly [K+]o peak amplitudes during repetitive stimulation, suggesting that Kir4.1-/- mice may display some compensatory mechanisms attempting to maintain extracellular K+ homeostasis. In addition, our model reveals that specific and acute inhibition of Kir4.1 channels slows down, but does not abolish, astroglial uptake of excess K+ during single, tetanic and repetitive stimulations, confirming that astroglial Na/K ATPases, included in our model, also contribute to K+ clearance [41].
Contrary to action potentials, characterized by a very fast dynamics in the order of a few milliseconds, astroglial K+ buffering lasts tens of seconds. As shown in the present study, most of extracellular K+ released by neurons is first cleared by astrocytes through Kir4.1 channels. To determine the factors controlling the slow timescale of astroglial K+ clearance, we focused on Kir4.1 channels. Because the astroglial leak conductance (equation 23) is six times smaller than the Kir4.1 channel conductance, we neglected it.
The dynamics of astrocytic membrane potential VA is described by equation 23, where the membrane capacitance is CA ≈ 15 pF and the maximal Kir4.1 channel conductivity isGKir≈60pS. In that case, using equation 23, the time constant of Kir4.1 channel-mediated return to equilibrium of astroglial membrane potential τA is defined as
We obtain the following approximation τA ≈ 0.6s using equation 23 and the parameters of table 1. This time constant is consistent with the fitted exponential decay time obtained in our simulations and experiments for a single stimulation where we obtained τ ≈ 0.7s. However, simulations for stronger stimulations indicate an increase of τ to approximatively 4 seconds (tetanic stimulation) and 9 seconds (repetitive stimulation). This increase in clearance duration is due to the dependence of the Kir4.1 current to [K+]o, as illustrated by the IV relation (Fig. 1C) and described in equation 22. The Nernst potential VKA increases for strong stimulations (tetanic and repetitive), which slow down the kinetics of astrocytic membrane potential VA through the term 1+exp(VA−VKA−V2AV3A) in equation 22. We conclude that the slow time scale of K+ clearance is in part due to the availability of Kir4.1 channels at low and high [K+]o. This clearance timescale is much longer than the glutamate clearance rate of τglu ≈ 15 ms that we previously reported [42]. Moreover, the redistribution of K+ released by neurons during the different regimes of activity shows that the higher the activity, the lower the proportion of released K+ remains transiently in the extracellular space. This suggests that Kir4.1 channels have a strong uptake capacity, especially for high regimes of activity ([K+]o up to 5–6 mM).
Remarkably, our model reveals that astroglial Kir4.1 channels strongly regulate neuronal firing induced by high stimulation regime such as repetitive stimulation. Kir4.1channels are crucially involved in regulation of [K+]o during this regime of activity, most likely because such stimulation triggered long-lasting neuronal release of K+ (20 mM over 30 seconds, Fig. 3G) resulting in a sustained, but moderate increase in [K+]o (>6 mM for ~22 s, Fig. 4I,J), compared to the neuronal release. These data suggest that during repetitive stimulation, astrocytes can buffer up to ~14 mM of [K+]o and thereby preserve neuronal firing. However, astroglial Kir4.1 channels slightly impact neuronal firing induced by single and tetanic stimulations, probably because these stimulations only triggered transient neuronal K+ release (0.9 mM over 300 ms (Fig. 3A) and 1.9 mM over 1.3 s (Fig. 3D), respectively), resulting in a short and small increase in [K+]o (>2.7 mM for ~450 ms for single stimulation (Fig. 4A,B), and >3.5 mM for 1.5 s for tetanic stimulation (Fig. 4E,F)). Nevertheless, we show a prominent and specific involvement of astroglial Kir4.1 channels in probabilistic firing activity induced by 3 to 10 Hz sub-firing stimulations (Fig. 5), suggesting a key role of these channels in sustained theta rhythmic activity. Interestingly, these data imply that Kir4.1 channels can contribute to fine tuning of neuronal spiking involving low, but long-lasting, increase in [K+]o. Thus besides gliotransmission, regulation of [K+]o by Kir4.1 channel provides astrocytes with an alternative active and efficient mechanism to regulate neuronal activity. Several studies have reported decreased Kir4.1 protein levels and Kir functional currents in sclerotic hippocampus from epileptic patients [43–46]. Whether these changes are the cause or the consequence of epilepsy is still an open question. However, Kir4.1-/- mice display an epileptic phenotype [22,47] and missense mutations in KCNJ10, the gene encoding Kir4.1, have been associated with epilepsy in humans [48,49]. These data thus suggest that impairment in Kir4.1 function leading to alterations in [K+]o dynamics, as shown in our study, may cause epilepsy. Remarkably, dysfunction of [K+]o regulation by Kir4.1 channels is likely involved in other pathologies, since it contributed to neuronal dysfunction in a mouse model of Huntington’s disease [50] and the presence of antibodies against Kir4.1 channels in glial cells was recently found in almost 50% of multiple sclerosis patients [51]. Thus astroglial Kir4.1 channels may well represent an alternative therapeutic target for several diseases.
Experiments were carried out according to the guidelines of European Community Council Directives of January 1st 2013 (2010/63/EU) and our local animal committee (Center for Interdisciplinary Research in Biology in Collège de France). All efforts were made to minimize the number of used animals and their suffering. Experiments were performed on the hippocampus of wild type mice (C57BL6). For all analyses, mice of both genders and littermates were used (PN19–PN25).
Acute transverse hippocampal slices (400 μm) were prepared as previously described [42,52–54] from 19–25 days-old wild type mice. Slices were kept at room temperature (21–23°C) in a chamber filled with an artificial cerebrospinal fluid (ACSF) composed of (in mM): 119 NaCl, 2.5 KCl, 2.5 CaCl2, 1.3 MgSO4, 1 NaH2PO4, 26.2 NaHCO3 and 11 glucose, saturated with 95% O2 and 5% CO2, prior to recording. Acute slices were placed in a recording chamber mounted on a microscope including infra-red differential interference (IR-DIC) equipment, and were bathed in ACSF perfused at 1.5 ml/min. ACSF contained picrotoxin (100 μM), and connections between CA1 and CA3 regions were cut to avoid epileptic-like activity propagation. Extracellular field and whole-cell patch-clamp recordings were obtained using glass pipettes made of borosilicate. Astroglial and postsynaptic responses were evoked by Schaffer collateral stimulation (0.05Hz) in the CA1 stratum radiatum region with glass pipettes filled with ACSF (300–700 kΩ). Astrocytes from stratum radiatum were recognized by their small soma size (5–10 μm), very low membrane resistance and hyperpolarized resting membrane potentials (≈- 80 mV), passive properties of their membrane (linear I-V), absence of action potentials, and large coupling through gap junctions. Field excitatory postsynaptic potentials (fEPSPs) were obtained in 400 μm slices using pipettes (4–6 MΩ) located in the stratum radiatum region. Stimulus artifacts were suppressed in representative traces. Whole-cell recordings were obtained from CA1 astrocytes, using 4–6 MΩ glass pipettes containing (in mM): 105 K-Gluconate, 30 KCl, 10 HEPES, 10 Phosphocreatine, 4 ATP-Mg, 0.3 GTP-Tris, 0.3 EGTA (pH 7.4, 280 mOsm). Prolonged repetitive stimulation was performed for 30 s at 10 Hz. Post-tetanic potentiation was evoked by stimulation at 100 Hz for 1 s in the presence of 10 μM CPP ((Rs)-3-(2-Carboxypiperazin-4-yl-)propyl-1-phosphonic acid). Recordings were performed with Axopatch-1D amplifiers (Molecular Devices, USA), at 10 kHz, filtered at 2 kHz, and analyzed using Clampex (Molecular Devices, USA), and Matlab (MathWorks, USA) softwares. The data represent mean ± SEM. Picrotoxin was from Sigma and CPP from Tocris.
We present here the biophysical model we have built to describe K+ dynamics during neuronal activity and specifically the role of astroglial Kir4.1 channels. After Schaffer collateral stimulation, excitatory synapses release glutamate molecules that activate postsynaptic neurons. We modeled this step by classical facilitation/depression model [55]. The resulting postsynaptic activity triggers ionic release in the extracellular space and a change in the astrocytic membrane potential through ion uptake. We used the average neuronal potential and mass conservation equations for ionic concentrations to model changes in astrocytes. We have built a tri-compartment model, which accounts for: 1) the neuron, 2) the astrocyte and 3) the extracellular space. We included voltage gated channels, Na/K pumps and astrocytic Kir4.1 channels.
To account for the stimulation of Schaffer collaterals that induce a postsynaptic response in the CA1 stratum radiatum region, we used a facilitation-depression model [55–57].
where f is the input function. For a single stimulation generated at time tstim, f(t) = δ(t-tstim).
A stimulation instantaneously activates a fraction Use of synaptic resources r, which then inactivates with a time constant τinac and recovers with a time constant τrec In the simulations, at time t = tstim, r and e respectively decreases and increases by the value User. The synaptic current Iapp is proportional to the fraction of synaptic resources in the effective state e and is given by Iapp = Asee (the parameter Ase is defined in table 1). We used the following definitions for the input function f:
The dynamics of the neuronal membrane potential, VN, follows the classic Hodgkin Huxley (HH) equations [58].
with rate equations
αn(VN)=0.01(VN+10)exp(0.1(VN+10))−1
(12)
βn(VN)=0.125exp(VN/80)
(13)
αm(VN)=0.1(VN+25)exp(0.1(VN+25))−1
(14)
βm(VN)=4exp(VN/18)
(15)
αh(VN)=0.07exp(VN/20)
(16)
βh(VN)=1exp(0.1(VN+ 30))+1
(17)
Vrest is the resting membrane potential and VKN and VNaN are respectively the K+ and Na+ equilibrium potentials and are given by the Nernst equations
VNaN=RTFln(Na0NaN)
(18)
VKN=RTFln(K0KN)
(19)
where Na0 and NaN are respectively the extracellular and neuronal sodium concentrations, and K0 and KN are respectively the extracellular and neuronal K+ concentrations that may vary as we shall describe below. We complete the description of all the neuronal currents with a leak current
IlN=glN(VN−VlN)
(20)
which stabilizes the membrane potential at its resting value. Finally, the neuronal membrane potential satisfies the equation
CNdVNdt=−(INa+IK+IlN+Iapp)
(21)
where Iapp is the synaptic current derived from equation 1.
To account for the K+ dynamics in astrocytes, we modeled the Kir4.1 channel according to its biophysical properties [59] and I-V curve [60]. The total astroglial current IKir depends on the membrane potential, the extracellular (K0) and the astrocytic (KA) K+ concentrations, and is approximated by
IKir=GKir(VA−VKA−VA1)(K01+exp(VA−VKA−VA2VA3))
(22)
where VKA is the Nernst astrocyte K+ potential, VA, the astrocyte membrane potential, K0 is the extracellular K+ concentration and VA1 (an equilibrium parameter, which sets Kir current to 0 at-80 mV), VA2 and VA3 are constant parameters calibrated by the I-V curve (Fig. 1C, [60]), as detailed below. The second term of equation 22 describes the dependence of IKir to the square root of K0 [60–64] and to the steady state open/close partition function of Kir4.1 channels according to the Boltzmann distribution [59], which includes dynamic variations of potassium Nernst potential during neuronal activity.
Adding a leak current IlA = glA(VA—VlA), which stabilizes the astrocyte membrane potential at—80 mV, the astrocyte membrane potential VA satisfies the equation
CAdVAdt=−(IKir+IlA)
(23)
where IKir is defined by relation 22.
We fitted the Kir4.1 channel I-V curve (equation 22) using the experimental recordings for the Kir4.1 channel (3 mM [K+] (Fig. 4 in [60,65]). We first obtained that VA1 = (VrestA − 26ln(3/145)) = −14.83 mV where VrestA = −80mV (potential for which the current is zero). We then used the Matlab fitting procedure for a single exponential with formula 22 changed to (V−VA1−26ln(3/145))3I with (V from-100 to 20 mV) to get that VA2 = 34 mV and VA3 = 19.23 mV (table 1). Varying [K+]o by 0.5 mM did not affect significantly the Kir4.1 channel I-V curve, confirming its robustness.
The K+ resting concentrations in neurons and astrocytes are maintained by Na/K pumps that balance the outward K+ and inward Na+ fluxes. The associated pump currents ipump,k (index k = N for the neuron, k = A for the astrocyte) depend on the extracellular K+ K0 and intracellular Na+ concentrations (NaN for the neuron and NaA for the astrocyte) and follow the same equation as [66],
ipump,k=imaxk(1+7.3K0)−2(1+10Nak)−3 for k=N,A
(24)
where imaxk is a constant (table 1).
We converted the different electrogenic neuronal and astrocytic channel currents into ionic fluxes [13]. A current I across a membrane induces a flow of charge i equals to δQ = I per unit of time. The corresponding change in extracellular concentration is given by I/(qNAVol0), where q = 1.6 * 10–19C is the charge of an electron, NA the Avogadro number and VolN, VolA andVol0 are the neuronal, astrocytic and extracellular volume respectively. To model the ionic concentration dynamics, we converted the currents INa, IK and IKir to the corresponding ionic fluxes iNa, iK and iKir We describe in the following paragraphs the equations for the ionic concentrations in the three compartments (neuron, extracellular space and astrocyte).
To determine the system of equations for the K+ fluxes, we use the mass conservation law for the extracellular K0, the neuronal KN and the astrocytic KA K+ concentrations. The extracellular K+ K0 increases with the neuronal current IK (see equation 8), which is here converted to iK (ion flux), but it is also uptaken back into neurons with a flux 2 ipumpN (the factor 2 is described in [67] and into astrocytes as the sum of the two fluxes 2 ipumpA plus iKir. Similarly, we obtain the equations for the neuronal and astrocytic K+ to balance the various fluxes. Finally, we get
To study quantitatively the acute and selective role of astroglial Kir4.1 channels in neuroglial K+ dynamics, we inhibited the Kir4.1 current in our tri-compartment model. We thus set at zero both the Kir4.1 current and the leak term. To compensate for the loss of K+ fluxes through astroglial Kir4.1 channels, we added in equation 27 a constant K+ flux to maintain [K+]o at an equilibrium value of 2.5 mM. This constant K+ flux in astrocytes could be mediated by various channels or transporters such as two pore domain potassium channels (K2P such as TWIK-1, TREK-1, TREK-2 and TASK-1), delayed rectifier potassium channels (Kv1.1, 1.2, 1.5 and 1.6), rapidly inactivating A-type potassium channels (Kv1.4), glutamate transporters or connexin43 hemichannels. However, since TASK-1 [68] and Cx43 hemichannels [69] are thought to be active in basal conditions, they are more likely to mediate such flux.
Similarly to the K+ dynamics, the equations for the Na+ fluxes are derived using the balance between the neuronal, astrocytic and extracellular concentrations. However, the main differences are that the pump exchanges 2 K+ for 3 Na+ ions, leading to the coefficient 3 in front of the pump term. In addition, to stabilize the sodium concentrations, we added two constant leak terms iNalA and iNalN (values given in table 1), as classically used [24],
dNa0dt=iNa+ iNalN+3ipumpN+3ipumpA+ iNalA
(28)
dNaNdt=(−iNa−3ipumpN− iNalN) Volo VolN
(29)
dNaAdt=(−iNalA−3ipumpA) Volo VolA
(30)
Numerical simulations. Simulations, numerical integrations and fitting computations were performed in Matlab. We used Runge Kunta fourth order method for the simulations, which were numerically stable. We used a time step of ∆t = 0.1 ms (simulations were repeated with smaller time step to check whether numerical accuracy was affecting results). The leak currents parameters were adjusted to stabilize the model at the resting membrane potentials (- 70 mV and - 80 mV for neurons and astrocytes respectively) and resting concentrations (neuronal [K+] and [Na+]: 135 mM and 12 mM, respectively; extracellular [K+] and [Na+]: 2.5 mM and 116 mM, respectively; astrocytic [K+] and [Na+]: 135 mM and 12 mM, respectively). The parameters for the Hodgkin Huxley equations were also adjusted to these concentrations.
Approximation of time constants. Time constants τ of simulated extracellular K+ transients were fitted to curves using a single exponential (e−tτ) (Fig. 4B,F,J). For all the fits obtained on the numerical simulation curves, we obtained an error estimation R-square ≥ 0.97. Time constants τ of experimental and simulated astroglial membrane potentials were calculated by computing the rise and decay times between 20% and 80% of the maximal peak amplitude responses (Fig. 2D,H,L).
All time constants τ were fitted to curves using a single exponential(e−tτ). For all the fits obtained on the numerical simulation curves, we obtained an error estimation R-square ≥ 0.97.
Approximation of facilitation/depression model parameters. To account for the synaptic properties of CA1 pyramidal neurons following single, tetanic and repetitive stimulations, we generated a synaptic current using the depression-facilitation model (equation 1) where Iapp depends on the input functions fs(t) (equation 4), fTT(t) (equation 5) and fRs(t) (equation 6), respectively (Fig. 2A,E,I). The synaptic current parameters were fitted to experimental recordings [26] by matching the time of maximal peak amplitude of fEPSP with the one of Iapp in control conditions (τ = 300 ms, τinact = 200 ms). The parameters for the Kir4.1 inhibition condition in the model were extracted from our experimental results on Kir4.1 glial conditional knockout mice [26] and are given by τrec = 500 ms, τinact = 160 ms. When Kir4.1 channels are inhibited (model) or knocked-out (experiment), the maximal peak amplitudes of the applied synaptic currents in the model (Iapp) and fEPSPs recorded experimentally are increased compared to control conditions [26].
We imposed an initial input at various frequencies (0.1, 1, 3, 5, 10, 50 Hz). Each input is generated by a sub-firing square current lasting 5 ms (Iapp). In addition, we added a Brownian noise of amplitude σ = 0.68 pA2 ms-1 to induce neuronal membrane potential fluctuation (equation 21), which amplitude (1 mV) was chosen to induce a probabilistic firing of 0.2, matching the CA1 pyramidal cells synaptic release probability p = 0.2 (probability to induce a postsynaptic response in equation 1) [70]. Using the tri-compartment model, we simulated at various frequencies a quantity that we called the observed firing probability defined empirically at time t as the time dependent ratio of the number of spikes observed at time t to the total number of simulations.
|
10.1371/journal.pgen.1003478 | Aconitase Causes Iron Toxicity in Drosophila pink1 Mutants | The PTEN-induced kinase 1 (PINK1) is a mitochondrial kinase, and pink1 mutations cause early onset Parkinson's disease (PD) in humans. Loss of pink1 in Drosophila leads to defects in mitochondrial function, and genetic data suggest that another PD-related gene product, Parkin, acts with pink1 to regulate the clearance of dysfunctional mitochondria (mitophagy). Consequently, pink1 mutants show an accumulation of morphologically abnormal mitochondria, but it is unclear if other factors are involved in pink1 function in vivo and contribute to the mitochondrial morphological defects seen in specific cell types in pink1 mutants. To explore the molecular mechanisms of pink1 function, we performed a genetic modifier screen in Drosophila and identified aconitase (acon) as a dominant suppressor of pink1. Acon localizes to mitochondria and harbors a labile iron-sulfur [4Fe-4S] cluster that can scavenge superoxide to release hydrogen peroxide and iron that combine to produce hydroxyl radicals. Using Acon enzymatic mutants, and expression of mitoferritin that scavenges free iron, we show that [4Fe-4S] cluster inactivation, as a result of increased superoxide in pink1 mutants, results in oxidative stress and mitochondrial swelling. We show that [4Fe-4S] inactivation acts downstream of pink1 in a pathway that affects mitochondrial morphology, but acts independently of parkin. Thus our data indicate that superoxide-dependent [4Fe-4S] inactivation defines a potential pathogenic cascade that acts independent of mitophagy and links iron toxicity to mitochondrial failure in a PD–relevant model.
| In this work we provide mechanistic insight linking together two of the earliest observations in Parkinson's disease: the excessive build-up of iron in diseased substantia nigra neurons and mitochondrial dysfunction particularly increased reactive oxygen species production at the level of Complex I. We identify aconitase mutants as strong genetic suppressors of Parkinson-related pink1 mutant phenotypes, both at the organismal and at the cellular/mitochondrial level. We show that the mitochondrial dysfunction in pink1 mutants that includes Complex I dysfunction results in superoxide-dependent inactivation of the Aconitase iron-sulfur cluster, leading to the release of iron and peroxide that combine to produce hydroxyl radicals and mitochondrial failure. Consequently, scavenging free iron using expression of mitoferritin or decreasing the levels of aconitase both rescue pink1 mutants; while increased wild-type Aconitase, but not a mutant that does not harbor an iron-sulfur cluster, results in severe mitochondrial defects. Given that reduced electron transport chain activity, increased oxidative stress, and natural iron build-up in the substantia nigra are common factors in sporadic and familial forms of Parkinson's disease, we believe that oxidative inactivation of Aconitase may represent an important pathogenic cascade underlying neuronal dysfunction in Parkinson's disease.
| Parkinson's disease (PD) is the most frequent neurodegenerative movement disorder, but the pathways that explain disease pathology remain poorly understood [1], [2]. While the most recognized pathological feature of PD is the preferential loss of dopaminergic (DA) neurons, one of the earliest observations in post mortem PD brains was the accumulation of iron in the substantia nigra (SN) [3], [4]. Iron-mediated toxicity may thus contribute to DA neuron dysfunction but the mechanism has not been established.
Mitochondrial dysfunction is thought to be an important aspect of PD progression. Mitochondrial toxins have been linked to sporadic forms of the disease and mitochondrial defects have been described in many cell types, also in SN mitochondria of PD patients [5], [6]. Likewise some of the genetic factors linked to the disease also point to a role for mitochondria. PD-associated mutations in pink1 and parkin, both affect mitochondrial function in genetic model organisms and in mammalian cells [7], [8], but how mitochondrial dysfunction and iron toxicity are linked remains elusive.
Pink1 and Parkin have been implicated in the clearance of dysfunctional mitochondria, a process dubbed mitophagy. In support, loss of parkin or pink1 in different cell types in flies, results in the accumulation of swollen and clumped mitochondria [9], [10], believed to be the result of defective mitophagy [11]. Furthermore, expression of factors that promote mitochondrial fission and, as a consequence, also indirectly promote mitophagy (gain of drp1 or loss of opa1 or mfn) partially rescue defects seen in pink1 and parkin mutants [12]–[14]. Further studies indicate that mitochondrial depolarization triggers the recruitment Parkin to mitochondria in a Pink1-dependent manner, facilitating mitophagy [15]. In line with this idea, over expression of Parkin in pink1 mutants, alleviates pink1-associated defects [9]–[16]. Hence, Pink1 acts with Parkin to regulate mitophagy.
In parallel, pink1 may also harbor supplementary roles. Expression of Parkin or Drp1, or loss of opa1 or marf only partially rescue pink1-associated defects, suggesting additional pathways are contributing to the phenotype. Furthermore, loss of pink1 function causes defects in the electron transport chain in fly and mouse cells [17], [18] that are not [19] or only partially [20] rescued by expression of Drp1. Finally, bypassing Complex I dysfunction, by expressing a yeast Complex I equivalent protein Ndi1 partially rescues the defects in pink1 mutants, but not those seen in parkin mutants [19]. Hence, Pink1 may play multiple roles in mitochondria, but the relative contribution of these different pathways to the pink1-dependent phenotypes, including the accumulation of swollen, clumped mitochondria remains to be determined.
In an unbiased genetic screen for heterozygous suppressors of Drosophila pink1 [21] we identified mitochondrial aconitase (acon) that encodes an enzyme catalyzing the first step of the Krebs Cycle [22]. Acon harbors an iron-sulfur [4Fe-4S] cluster [23] and we show that oxidative inactivation of this cluster in pink1 mutants is a major cause of iron toxicity that contributes to mitochondrial swelling and clumping in pink1 mutants. Our data are most consistent with acon acting downstream of pink1 and affecting mitochondrial morphology independently of parkin-mediated mitophagy. Thus oxidative inactivation of Aconitase is a source of iron toxicity that leads to mitochondrial defects in pink1 mutants and we propose a model where different pathways controlled by Pink1, including mitophagy and the maintenance of ETC activity can contribute to mitochondrial failure in specific cell types.
Pink1 mutants show a severe defect to fly caused by mitochondrial dysfunction [19], [21]. To identify genetic modifiers of pink1, we have tested a collection of 193 chemically induced (EMS) recessive lethal mutants that have been pre-selected for defects in mitochondrial function and neuronal communication [24]–[26], for their ability to modify the pink1 null mutant flight defect. At the 1% significance level we isolated 5 suppressors (p<0.01) [21] and to reveal mechanisms by which the modifiers affect Pink1, we mapped one of these recessive lethal suppressors to aconitase (acon) and named it acon1. This mutant fails to complement a deletion that uncovers acon as well as a lethal transposon insertion in acon that we named acon2 (Figure S1A). Sequence analysis of acon1 reveals a nonsense mutation in exon 2 (Figure S1A). In addition, semi-quantitative RT-PCR and Western blot analysis indicates severely reduced mRNA and protein levels in animals that are homozygous for either acon allele (Figure S1C and S1D), indicating that both are loss of function alleles. Moreover we can rescue the lethality and phenotypes associated with acon1/acon2 using a 20 kb genomic fragment encompassing the acon locus, yielding normal adult flies that do not show obvious behavioral abnormalities (Figure S1A, S1B). Likewise ubiquitous expression of acon cDNA is also able to rescue acon1/acon2-associated lethality (Figure S1B). Thus, one of the suppressors of pink1 harbors a lethal lesion in acon and the lethality in the mutants is solely due to disruption of acon.
Heterozygosity of acon significantly suppresses the flight defect associated with pink1B9 mutants (Figure 1A, 1B). The extent of rescue we obtained by removing acon, is similar to previously reported conditions that suppress pink1 flight defects, including adding a copy of drp1 (drp1+) that facilitates mitochondrial fission, removing a copy of opa1 (opa1S3), reducing mitochondrial fusion (Figure S2A, S2B), expression of Parkin, expression of yeast NDI1 that bypasses Complex I of the electron transport chain (ETC), or feeding pink1 mutants ubiquinone or menaquinone that boost ETC function [9], [10], [13], [19], [21]. To test if the rescue that we observe is solely due to partial loss of acon (and not due to second site interactors on the chromosome), we determined flight but also ATP levels of pink1 mutants with one copy of a mutant acon allele. While heterozygous acon1 and acon2 mutants alone do not show defects (Figure 1C), we find that one copy of either acon1 or acon2 significantly rescue the reduced ATP levels in pink1 mutants (Figure 1D). This effect in pink1 mutants is specific to loss of acon as introduction of a genomic copy encompassing wild type acon in pink1B9;acon2/+ flies completely reverses both the flight and ATP level phenotypes to pink1B9 mutant levels (Figure 1B and 1D). Thus, pink1 mutant phenotypes are specifically rescued by partial loss of acon expression.
To further quantify the effect of acon on pink1 mutant phenotypes we also analyzed mitochondrial morphology in adult indirect flight muscles using transmission electron microscopy. As previously described [9], [10], the flight muscles of pink1 mutants exhibit swollen mitochondria with disorganized and fragmented cristae when compared to flight muscles from control flies or when compared to heterozygous acon mutants that do not show mitochondrial morphological defects (Figure 1E and Figure S1F). Partial loss of acon in pink1 mutant results in a substantial rescue of the mitochondrial morphological defects in flight muscles, displaying substantially more intact cristae and less swollen mitochondria compared to pink1 mutants (Figure 1E). Hence, also at the ultrastructural level, partial loss of acon significantly alleviates mitochondrial morphological defects in pink1 mutant muscles.
Pink1 mutants also show swollen and clumped mitochondria in dopaminergic neurons in the adult brain [9], [10]. To test if loss of acon can also rescue this defect, we expressed mitoGFP in pink1 mutant flies and in pink1 mutant animals heterozygous for acon. In line with the electron microscopy data of muscles, mitochondria in muscles labeled by mitoGFP (expressed using the ubiquitous da-GAL4) are spherical and aggregated in pink1 mutants and this defect is significantly rescued by partial loss of acon (Figure 1F and 1G). Next, we expressed mitoGFP in dopaminergic neurons using ple-Gal4 (also called TH-Gal4). While mitochondria are organized in a tubular network in wild type dopaminergic neurons, pink1 mutant mitochondria appear mostly as fragmented spherical aggregates in all dopaminergic neuron clusters analyzed (Figure S1H) [9], [10]. We quantified size and number of mitochondrial aggregates in the PPM3 cluster (Figure S1H and Methods). While heterozygous acon1 and acon2 mutants do not show defects compared to controls (Figure 1H, 1I), we find that both one copy of either acon1 or acon2 significantly rescue the increased size and number of mitochondrial aggregates in pink1 mutants (Figure 1H, 1I and Figure S2A). This rescue in pink1 mutants is specific to the partial loss of acon as introduction of a genomic copy encompassing wild type acon in pink1B9; acon1/+ flies reverses these phenotypes back to pink1B9 mutant levels (Figure 1F–1I). Furthermore, we confirm that protein levels are reduced by about 50% in pink1B9; acon1or2/+ compared to pink1 mutants and are restored in flies expressing a genomic copy of wild type acon (Figure S1E). Thus, together our data indicate that morphological defects of mitochondria in pink1 mutants are significantly rescued by partial loss of acon expression and the mitochondrial morphological defects in pink1 mutants are dependent on acon expression.
acon is predicted to encode mitochondrial Aconitase (Acon), an iron sulfur cluster containing protein, that catalyzes the formation of isocitrate in the first step of the Krebs cycle [22]. To assess whether Acon localizes to mitochondria we fractionated fly tissue in cytoplasmic and mitochondrially enriched fraction and performed Western blotting using anti-Acon antibodies. Acon is enriched in the mitochondrial fraction (Figure S1G).
Acon harbors a single unligated iron atom in its [4Fe-4S]2+ cluster, and the enzyme is in this respect unique in mitochondria. Such an unligated iron atom is particularly sensitive to superoxide (O2−)-dependent oxidation [27]–[29] that results in cluster instability. Oxidation is followed by the release of Fe2+ and H2O2 that may contribute to oxidative damage and mitochondrial morphological defects through the formation of the potent hydroxyl radical (OH.) by the Fenton reaction [30]. Thus, the specific configuration of the Acon [4Fe-4S]2+ cluster in combination with its proximity to mitochondrially generated superoxide place Acon as a major mediator of oxidative stress in mitochondria. We therefore wondered if O2− leaking from defective pink1 mutant mitochondria could be a source of Acon inactivation resulting in morphological defects. To test if also in fruit flies the loss of pink1 function results in increased O2− production, we incubated mitochondrial preparations from pink1 mutant flies and controls with Complex I substrates (pyruvate/malate) and used the fluorogenic probe dihydroethidium (DHE) to measure O2− production [31], [32]. Similar to wild type mitochondria in the presence of AntimycinA, known to induce O2− production (Figure 2A), pink1B9 mitochondria show a significant increase in DHE fluorescence compared to controls (Figure 2A). These data indicate that pink1 loss leads to increased O2− production.
If the increased O2− in pink1 mutants can act via the Acon [4Fe-4S] cluster to cause mitochondrial swelling, we expect (1) that partial loss of acon does not rescue the increased O2− production in pink1 mutants; (2) that Acon enzymatic activity normalized to total Acon protein is reduced in pink1 mutants; (3) that H2O2 and Fe2+ levels are increased in pink1 mutants as a result of Acon inactivation, and (4) that this defect is rescued by partial loss of acon. First we assessed O2− in pink1 mutants heterozygous for acon1 or acon2 that we showed rescues mitochondrial morphological defects in pink1B9. However, in line with our model, heterozygosity for acon does not reduce pink1B9-induced O2− production (Figure 2A), indicating that increased O2− production per se does not induce mitochondrial morphological defects. Next we measured Acon activity in pink1 mutant mitochondria and we find that Acon activity normalized to total Acon protein levels is significantly reduced compared to the controls. These data are in line with increased Acon inactivation in pink1 mutants (Figure 2B), likely as a result of the increased O2−.
Further testing our model, we also measured H2O2 and Fe2+ content. To measure H2O2 and its radical derivatives we incubated fly lysates with the fluorescent probe dichlorofluorescein diacetate (DHCF-DA) [33]. We find a 50% increase in fluorescence in pink1 mutant lysates compared to the control (Figure 2C). Thus, pink1 mutants accumulate H2O2 and/or derivatives thereof. We also measured mitochondrial Fe2+ content by incubating mitochondrial enriched fractions with Rhodamine B-[(1,10-phenanthrolin-5-yl)aminocarbonyl]benzyl ester (RPA) [34]. In the presence of Fe2+, RPA fluorescence quenches and in pink1B9 mitochondria, we observe a significant increase in RPA quenching compared to controls (Figure 2D). These data indicate increased mitochondrial Fe2+ levels in pink1B9 mutants. This effect is specific, as incubating mitochondria of controls and mutants in Rhodamine B 4-[(Phenanthren-9-yl)Aminocarbonyl]benzyl ester (RPAC) that consists of the same fluorophore as RPA but without iron-chelating properties, does not show quenching in pink1B9 or in controls (Figure 2D). Thus, our data indicate that pink1B9 mutants harbor increased levels of Fe2+ and of H2O2 and/or its radical derivatives.
Next we tested if increased mitochondrial Fe2+ and H2O2 accumulation in pink1 mutants is a consequence of Acon[4Fe-4S] inactivation by O2−. We therefore measured Fe2+ and H2O2 and its derivatives levels in mitochondria of pink1 mutants heterozygous for acon1 or acon2. While the increased O2− production in pink1B9 mutants was not reduced by heterozygous acon, as shown above (Figure 2A), we find that compared to pink1B9, mitochondrial Fe2+ and H2O2 levels are significantly lower in pink1B9 heterozygous for acon1 or acon2 (Figure 2C and 2D). Thus, these data are consistent with the possibility that mitochondrial Fe2+ and H2O2 and/or its radical derivatives-accumulation in pink1 mutants is caused by oxidative inactivation of Acon.
Our biochemical data support a model in which oxidative inactivation of Acon and ensuing Fe2+ and H2O2 accumulation contributes to the mitochondrial morphology defects in pink1 mutants. We reasoned that if partial loss of acon protects against mitochondrial stress in pink1 mutants, increased levels of Acon expression may predispose cells to develop mitochondrial morphological defects, provided sufficient O2− is around. We therefore created transgenic animals that overexpress wild type Acon (Figure 3A) resulting in increased Acon activity (Figure 3B). We then determined mitochondrial morphology using mito-GFP and the ple-GAL-4 driver upon expression of Acon in DA neurons. While mitochondria in DA neurons of control flies organize in a long tubular network, mitochondria in DA neurons that overexpress Acon form fragmented spherical aggregates (Figure 3C, 3D and Figure S2A). Hence, in contrast to partial loss of acon that rescues mitochondrial defects in pink1 mutants, overexpression of Acon causes mitochondrial morphological defects and swelling of mitochondria in DA neurons.
Based on the finding that increased expression of Acon causes mitochondrial morphological defects we tested if pink1 mutant flies upregulate Acon expression. We measured acon mRNA and protein levels in pink1 mutants, but in contrast to our expectation, we find a significant downregulation of both acon mRNA and Acon protein levels in pink1 flies (Figure S2C, S2D) suggesting that an adaptive mechanism already acts in pink1 mutants to down regulate Acon expression. Thus, the pink1B9-induced stress response results in lower Acon levels and, as shown above, further reducing Acon expression (using heterozygous acon mutants) is protective against mitochondrial defects in pink1 mutants. Taken together, the data are consistent with Acon being a dosage sensitive modifier of morphological defects in mitochondria.
To test if the mitochondrial morphological defects in DA neurons following Acon overexpression are induced by increased Acon catalytic activity or by the presence of an [4Fe-4S] cluster we generated transgenic flies that either overexpress a catalytic inactive Acon (AconS677A) that still harbors its [4Fe-4S] cluster, or flies that overexpress an Acon without its [4Fe-4S] cluster (AconC459S) and is thus also catalytically inactive [22], [35].Western blotting indeed indicates overexpression of the mutant Acon proteins (Figure 3A), and as expected, Acon enzymatic activity measured in fly head lysates is only increased when wild type Acon is expressed, and not when AconS677A or AconC459S are expressed (Figure 3B). While overexpression AconS677A in DA neurons results in obvious mitochondrial morphological defects similar to the overexpression of wild type Acon, overexpression of AconC459S is inert (Figure 3C, 3D and Figure S2A). Hence, the Acon [4Fe-4S] cluster predisposes DA neurons to mitochondrial morphological defects.
Our data are in line with a model where oxidative inactivation of the Acon [4Fe-4S] cluster by O2− contributes to mitochondrial morphological defects. To find further evidence for this idea we expressed Drosophila mitochondrial Ferritin (Fer3HCH) [36] in DA neurons of pink1B9, using the ple-GAL4 driver and assessed mitochondrial morphology using mito-GFP. We find that expression of Fer3HCH significantly rescues defects in mitochondrial morphology in pink1B9 mutants (Figure 4E, Figure 2F, and Figure S2A), suggesting that iron toxicity causes mitochondrial defects in pink1 mutants. Consistent with this model, expression of Fer3HCH in flies that over express Acon also results in a significant rescue of the mitochondrial morphological defects in the DA neurons (Figure S2E, S2F). Hence, the mitochondrial swelling as a result of Acon overexpression is at least in part mediated by iron. Together these data indicate that Acon is a critical source of Fe2+-mediated mitochondrial toxicity.
Mitochondrial dynamics and mitophagy are critical processes in maintaining a healthy population of mitochondria. Pink1 has been implicated to regulate mitochondrial homeostasis via several mechanisms. Deregulation of these pathways may be a source of O2−, responsible for Acon inactivation. While Pink1 has been found to maintain the activity of Complex I in the ETC [17]–[20], the protein has also been linked to mitophagy in a pathway involving Drp1 and Parkin [11]–[15], [20], [37]–[40]. Dysfunctional mitochondrial parts may be segregated by the fission factor Drp1 [41], [42]. Pink1 stabilized on depolarized mitochondria then mediates Parkin recruitment causing the ubiquitination of mitochondrial proteins and activation of the autophagic machinery [41], [42]. To test if enlarged and swollen mitochondria upon Acon over expression are a consequence of defective remodeling or mitophagy we co-overexpressed Parkin, a protein that ubiquitinates mitochondrial targets, or Drp1, a mitochondrial fission factor, two conditions thought to facilitate mitophagy. While over expression of Parkin or Drp1 -as expected- result in fragmentation of mitochondria, these conditions do not rescue the defect in mitochondrial swelling and clumping induced by expression of Acon or AconS677A (Figure 3E, 3F and Figure S2A). Hence, our data suggest that the defects in mitochondrial morphology induced by Acon expression are at least in part caused independently from defects in remodeling and mitophagy.
Given that pink1 mutants display reduced Complex I activity [17]–[20] and this feature may also be a source of increased O2− we tested if mitochondrial swelling and clumping seen in animals where we downregulated an evolutionary conserved Complex I component, NDUFA8, can be rescued by partial loss of acon. First, we confirm increased O2− production and find a concomitant inactivation of Acon activity upon RNAi-mediated downregulation of NDUFA8 (Figure 4A and 4B). Second, we believe that this O2− is produced at least partly independently from defects in mitochondrial remodeling because expression of Drp1 in DA neurons with reduced NDUFA8 function does not fully rescue the mitochondrial swelling and clumping phenotypes in PPM3 DA neurons (Figure 4C, 4D and Figure S2A). Next, we tested the ability of heterozygous acon to modulate the mitochondrial morphological defect induced by NDUFA8 RNAi and find that heterozygous acon is more effective than expression of Drp1 in rescuing the mitochondrial deficits in DA neurons (Figure 4C, 4D and Figure S2A). Likewise, and in line with our model, expression of mitoferritin (Fer3HCH) also alleviates mitochondrial defects in animals that express RNAi to NDUFA8 in DA neurons (Figure 4E, 4F and Figure S2A). Hence, our data suggest that Acon is inactivated by ETC-derived O2− causing oxidative stress.
Our work suggests that mitochondrial morphological defects in pink1 mutant DA cells can be of different origin: both O2−-dependent Acon inactivation or loss of Parkin-dependent mitophagy yield swollen and clumped mitochondria. Alleviating the defects induced by either pathway using heterozygous acon or expressing Drp1 or Parkin both rescue the mitochondrial morphological defects in pink1 mutants (this work; [12]–[14], [16]). To further support this notion, we first assessed if mitochondrial defects in parkin mutants can be rescued by partially removing acon function. parkin mutants display enlarged and swollen mitochondria in muscles and DA neurons, many of the flies also fail to fly and animals harbor lower ATP levels. In contrast to removing acon function in pink1 mutants, heterozygosity for acon fails to rescue the inability of parkin mutants to fly, their reduced ATP levels and their defects in mitochondrial morphology (Figure 5A–5D). Hence, our data suggest that acon acts independently from defects in Parkin-dependent mitophagy. Finally if our model is correct, we reasoned that the combination of Drp1 expression and acon heterozygosity in pink1 mutants should yield additive ‘super rescue’. We therefore tested the ability of these flies to fly and find that they fly significantly better than pink1 mutants or than pink1 mutants partially rescued by either Drp1 expression or by heterozygous acon (Figure 5E). Hence, these data are in line with Pink1 controlling different mitochondrial pathways that can be targeted largely independently. We speculate that increased O2− derived from a defective Complex I in pink1 mutants is an important contributor to Acon inactivation, but other sources of O2− may contribute to mitochondrial failing as well.
Iron accumulation in the substantia nigra, systemic mitochondrial dysfunction and oxidative stress have all been implicated in PD pathology; however, a link between these factors remains elusive. Here we show that oxidative inactivation of Acon generates iron-mediated oxidative stress that contributes to mitochondrial swelling in Drosophila pink1 mutants (Figure 6). Inactivation of Acon[4Fe-4S] clusters could contribute to mediating O2− toxicity by simultaneous release of Fe2+ and H2O2 [43] that combine in the Fenton reaction to generate highly toxic hydroxyl radicals [30], [44] (Figure 6). Hydroxyl radicals can induce mitochondrial permeability transition and swelling [45]–[47], in line with electron microscopic analyses of pink1 mutants where mitochondria appear swollen and show disorganized cristae [9], [10] (Figure 1). Four major findings support that this iron-mediated toxic mechanism is an additional important aspect of mitochondrial dysfunction in pink1 mutants. First, we find increased O2− production, increased Acon inactivation and more Fe2+ and H2O2 accumulation in pink1 mutants (Figure 2). Second, partial loss of Acon reduces Fe2+ and H2O2 accumulation and alleviates pink1-associated phenotypes including mitochondrial morphological defects in muscle and DA neurons (Figure 1). Third, overexpression of wild type Acon in dopaminergic neurons produces a mitochondrial morphological defect and this effect is completely dependent on the presence of the [4Fe-4S] cluster in Acon (Figure 3). These data also indicate mitochondrial integrity is sensitive to Acon [4Fe-4S] cluster dosage. Finally, chelating iron by expressing mitochondrial Ferritin is sufficient to rescue pink1 mitochondrial morphological defects (Figure 4). Thus, our data suggest that inactivation of Acon and iron accumulation might be a pathogenic mechanism triggered by loss of pink1 and increased superoxide, linking iron accumulation and mitochondrial failure.
Acon inactivation is dependent on O2− that, amongst other sources (see below), may be produced in defective mitochondria. While various mitochondrial insults can result in increased O2− production, our work is most consistent with Parkin-dependent mitophagy being not the major source of Acon inactivation in pink1 mutants. The mitochondrial morphological defects induced by Acon overexpression were not strongly rescued by expressing Drp1, a condition that indirectly promotes mitophagy and parkin mutants were not majorly rescued by partial loss of acon (Figure 3 and Figure 5). In contrast, mitochondrial morphological defects in DA neurons of flies with reduced Complex I activity are significantly rescued when acon is heterozygous (Figure 4). Hence, Acon seems to act in a Pink1-dependent pathway that can operate largely independently of mitophagy (Figure 6).
Defects at the level of Complex I are often associated with increased leaking of the toxic O2− [48], [49], and likewise, systemic inhibition of Complex I mimics features of PD in animal models [50]–[53]. Previous work in flies or mice has indicated reduced ETC function [17]–[20] in pink1 mutants, and we show that this condition results in mitochondrial morphological defects in an Acon-dependent manner. Similar to pink1 mutants, RNAi-mediated knock down of an evolutionary conserved Complex I component, NDUFA8, also results in an increased production of superoxide as well as in Acon inactivation. We show that these biochemical changes correlate with mitochondrial morphology defects in dopaminergic neurons that can be rescued by partial loss of acon or by over-expression of mitoferritin that scavenges the released Fe2+ [36] (Figure 4).
It is interesting to note that increased O2− production per se is not sufficient to generate mitochondrial morphological defects, and that the presence of sufficient amounts of Acon is required. Indeed, our data indicate that pink1 mutants heterozygous for acon show increased levels of O2− but normal mitochondrial morphology. Our data also indicate that upstream events in pink1 mutants that result in increased O2− production contribute to mitochondrial morphological defects because of oxidative inactivation of Acon. In line with this, overexpression of the mitochondrial superoxide dismutase 2 (SOD2) that scavenges O2−, successfully rescues mitochondrial swelling phenotype of pink1 in DA neurons [54]. Given that both genetic forms of PD as well as sporadic cases of PD show ETC defects [5], [6], [17]–[19], our work may be relevant for idiopathic cases that suffer from mitochondrial dysfunction as well. Acon inactivation and iron-mediated toxicity might thus have a more general role in the pathogenesis of PD.
While pink1 loss affects numerous cell types, our data also start to provide insight as to why DA neurons in the substantia nigra are more vulnerable in PD. While overexpression of Acon or downregulation of Complex I produces mitochondrial morphological defects in DA neurons, in Drosophila flight muscles mitochondria appear morphologically largely normal (data not shown). These data suggest a tissue-specific response in that Acon inactivation has a stronger impact in DA neurons than in muscle cells. Each cell type is exposed to various sources of O2−, but DA neurons in particular are exposed to dopamine-induced oxidative stress that is a source of O2− [55]–[57]. Furthermore, the substantia nigra in humans is naturally rich in iron [58] and this feature may lower the threshold for hydroxyl radical production in the Fenton reaction that is facilitated by Acon inactivation. Pink1 mutations or environmental factors in some sporadic cases of PD already result in increased levels of O2−, but we hypothesize that in DA neurons, additional dopamine-induced oxidative stress may facilitate Acon inactivation and hydroxyl radical production providing insight into one of the pathways underlying mitochondrial failure in pink1 mutants.
Flies were raised on standard cornmeal and molasses medium at 25°C. w1118; UAS-mitoGFP, w1118; daGal, w1118; pleGal4, w; UAS-4EBP and w1118; Mi{ET1}AconMB09176/SM6a (acon2) and were obtained from Bloomington stock center (Indiana, USA). w1118 pink1B9 and w1118 pink1RV, parkin1 and parkinRV [59] were provided by Jongkyeong Chung (Advanced Institute of Science and Technology, Korea) [10]. parkinΔ21 mutant flies were a gift from Graeme Mardon (Baylor College of Medicine) [60] and drp1+ genomic rescue constructs were provided by Hugo Bellen (Baylor College of Medicine) [61] w1118; UAS-Fer3HCH was provided by Dr Fanis Missirlis (Qeen Mary University of London, UK). w1118; UAS-CG3683RNAi (w1118; P{GD16787}v46799/CyO) was from the Vienna Drosophila RNAi Center (VDRC) [62].
The genomic clone CH322-18I04 was obtained from BACPAC Resources (Children's Hospital Oakland).
UAS-Aconwt was generated by PCR amplification of BDGP cDNA clone LD24561 using primers: AconcDNA.F (5′ ATGGCTGCGAGATTGATGAACG) and AconcDNA.F (5′ TTACTGGGCCAGCTCCTTCATGC). The S677A and C459S mutations were introduced in the primers and the mutated cDNAs were generated by overlap extension PCR using the following primers: S677A.F (5′GAcgCACCCTCGCCGTAGTTCTCATC) S677A.R (5′AACTACGGCGAGGGTGcgTC), C459.F (5′GGTCCCtCcATTGGACAGTGGGATCG) and C459.R (5′CGATCCCACTGTCCAATgGaGGGACC). All constructs were cloned into the EcoRI and NotI restriction sites of pUAST-attB [63]. Following sequencing, transgenic flies were created at GenetiVision Inc. (Houston, USA) using PhiC31 mediated transgenesis in the VK1 docking site (2R, 59D3) [64].
For quantitative RT-PCR, total RNA was isolated using TRI Reagent (Sigma-Aldrich) according to the manufacturer's protocol. Subsequently, the RNA samples were cleaned up using the RNeasy Mini Kit with the on-column DNAse treatment (Qiagen). 1 µg of total RNA was used as a template for synthesis of oligodT-primed double stranded cDNA using the SuperScriptIII First-Strand Synthesis System (Invitrogen). 20 ng cDNA of each sample was used for acon SYBR Green PCR Master mix (Applied Biosystems) and the following primers were used: aconRT-F (5′ TCGTGCCATTATCGTCAAGTC) and aconRT-F (5′ AGGTTGAGCAGGGAGATTTTG). All experiments were performed in triplicate and run on a Roche LC480 system. The data were normalized utilizing RP-49, a ribosomal gene, using following primers: RP-49-F (5′ ATCGGTTACGGATCGAACAA) and RP-49-R (5′ GACAATCTCCTTGCGCTTCT).
For Western blots, flies were homogenized in cold T-PER buffer (ThermoScientific) with complete protease inhibitor mixture (Roche). Protein concentration was determined by BCA protein quantification kit (Pierce). Samples were diluted in 2-mercaptoethanol 10% SDS loading buffer and boiled for 5 min and 15 µg of proteins were separeted on pre-cast 4–12% NuPage Bis-Tris gels (Invitrogen). Following transfer to nitrocellulose, blots were probed with primary antibodies: 1∶5000 Anti-ACO2 (AbGent), 1∶1000 anti-Tubulin (B5–12, Sigma) and 1∶1000 HRP coupled secondary antibodies (Jackson immunolabs). Blots were developed with Western-Lightning-ECL (PerkinElmer) and imaged. Quantification was performed using gel analyzer tool in ImageJ software from the US National Institute of Health (http://rsb.info.nih.gov/ij/).
Batches of 5 days old male flies were transferred to an empty vial (5 cm D, 10 cm H). Flies were allowed to climb above a marked line at 9 cm height; the vial was gently tapped and visually scored for flying flies. Flies at the bottom were removed and the remaining flies were retested. Flies that fly twice were assigned a score of 1, the others a score of 0.
ATP content was determined as described [10]. 5 days-old flies with abdomen dissected out were homogenized in 50 µl of 6 M guanidine-HCl 100 mM Tris and 4 mM, EDTA, pH 7.8. These homogenates were snap-frozen in liquid nitrogen and then boiled for 3 min. Samples were then centrifuged and the supernatant was diluted (1/50) in extraction buffer, mixed with luminescent solution (ATP Determination Kit, Invitrogen) and luminescence was measured on an EnVision Multilabel Reader (Perkin Elmer). ATP (nmol) was determined using a standard curve and normalized to protein content (mg) measured by BCA assay (Pierce).
Thoraxes were fixed in paraformaldehyde/glutaraldehyde, postfixed in osmium tetroxide, dehydrated and embedded in Epon. Sections 80 nm thick were stained with uranyl acetate and lead citrate and subjected to TEM analysis.
H2O2 was measured as described [33]. 4–5 adult flies were homogenized in 50 µl cold lyses buffer T-PER (Thermo scientific) and the homogenate were cleared by centrifugation at 1000×g for 5 min at 4°C. 140 µL of PBS containing 50 µM of DCFH-DA (molecular probe) were added to 10 µL of lysate in a 96-well plate format and incubated at 25°C for 10 minutes in the dark. DCFH-DA fluorescence (485exc/530em) was measured using Wallac Victor2 1420 (Perkin Elmer). Fluorescence intensity was normalized to the protein amount (BCA, Pierce) and expressed as relative to the control.
Fifty flies were gently crushed in 1 ml chilled mitochondrial isolation medium (Mitosciences) by using a porcelain mortar and pestle, then spun twice at 1,000×g for 5 min at 4°C to remove debris. The supernatant was then spun at 12,000×g, for 15 min at 4°C. The pellet, containing the mitochondria, was washed with 1 ml of isolation medium and resuspended in 40 µl of isolation medium supplemented with complete protease inhibitor mixture without EDTA (Roche).
Mitochondrial Superoxide production was measured as described [32]. 10 µg of mitochondria were incubated in experimental buffer (EB: 125 mM KCl, 10 mM Tris-MOPS, 1 mM KPi, 10 µM EGTA-Tris, pH 7.4, 25°C) supplemented with 1.25 mM Pyruvate/1.25 mM malate and 5 µM DHE (Molecular probe) in a 96-well plate format for 10 min. The fluorescence was measured (485exc/590em) using Wallac Victor2 1420 (Perkin Elmer). Fluorescence intensity was normalized to the initial value and expressed as relative to the control. 10 µM antimycin A was used to induce superoxide production in control mitochondria.
For mitochondrial ferrous iron level measurements, 10 µg of mitochondria were resuspended in isolation buffer (Mitosciences) and incubated with 20 µM of RPA or RPAC (Squarix Biotechnology) in a 96-well plate format at room temperature for 10 min. RPA/RAPC fluorescence (560 exc/600 em) was measured using Wallac Victor2 1420 (Perkin Elmer). Quenching was calculated as percent of initial fluorescence.
Aconitase enzyme activity microplate kit (Mitosciences) was used according to the manufacturer's protocol to measure Aconitase activity. 20 µg of mitochondria were incubated with assay buffer and the activity was measured by following conversion of isocitrate to cis-aconitate as in increased in 240 nm UV absorbance. Measurements were recorded over 30 min. at 1 min intervals and aconitase activity were calculated from the linear increase in absorbance and normalized to the amount of aconitase, determined by western blot, in the same mitochondrial preparation. Values were reported as relative activity to the control.
Brain dissection and whole-mount immunohistochemistry for tyrosine hydroxylase (TH) was performed as described [65]. Primary 1∶100 antibody against TH (Chemicon) and secondary alexa 555 (Invitrogen) were used. Brains were imaged on a Zeiss LSM 510 META confocal microscope using a 63xoil NA 1.4 lens. Mitochondrial tagged GFP (mito-GFP) was visualized using 488 nm laser and 500–530 band pass emission filter. Because mitochondrial morphology is sensitive to environmental conditions, variations did occur from batch to batch. We only compared flies of different genotypes if normal mitochondrial morphology was observed in the control samples (Figure S1H–S1H″) in the same batch. For quantification of mitochondrial aggregates size and numbers, DA neurons of PPM3 cluster (Figure S1H–S1H″) were scored. Quantification of aggregate size was done using “analyzing particles” plugin in ImageJ (http://rsb.info.nih.gov/ij/): rounded particles were automatically detected and the average surface area of aggregates in each neuron was determined as total area occupied by aggregates/number of aggregates.
Adult flies were fixed in PBS with 5% formaldehyde and 0.4% Triton for 3 hours. Thoraxes were dissected in PBS and mounted in vectashield (Vector Laboratories) and were imaged on a Zeiss LSM 510 META confocal microscope using a 63xoil NA 1.4 lens. Mitochondrial tagged GFP (mito-GFP) was visualized using 488 nm laser and 500–530 band pass emission filter. For muscle section with same area were scored and quantification of mitochondrial aggregates was performed as described above.
|
10.1371/journal.pntd.0003241 | Epidemiological Trends of Dengue Disease in Thailand (2000–2011): A Systematic Literature Review | A literature survey and analysis was conducted to describe the epidemiology of dengue disease in Thailand reported between 2000 and 2011. The literature search identified 610 relevant sources, 40 of which fulfilled the inclusion criteria defined in the review protocol. Peaks in the number of cases occurred during the review period in 2001, 2002, 2008 and 2010. A shift in age group predominance towards older ages continued through the review period. Disease incidence and deaths remained highest in children aged ≤15 years and case fatality rates were highest in young children. Heterogeneous geographical patterns were observed with higher incidence rates reported in the Southern region and serotype distribution varied in time and place. Gaps identified in epidemiological knowledge regarding dengue disease in Thailand provide several avenues for future research, in particular studies of seroprevalence.
PROSPERO CRD42012002170
| We conducted this comprehensive systematic review to determine the impact of dengue disease in Thailand for the period 2000–2011, and to identify future research priorities. Well-defined methods were used to search and identify relevant published research, according to predetermined inclusion criteria. In addition to information from studies published in the literature, the review draws largely on surveillance data from the Annual Epidemiological Surveillance Reports published by the Thailand Ministry of Public Health. The pattern of annual number of reported dengue cases over the review period was complicated by epidemic years; consequently, a trend in the number of reported cases could not be identified. It was apparent that despite a shift in age group distribution dengue from younger towards older persons, dengue in Thailand remains a predominantly childhood disease. The seasonality and heterogeneous spatial and temporal nature of the disease were confirmed. It is clear that the nationwide passive surveillance system is a source of consistent data relating to severity, age and serotype. However, several gaps were identified that would benefit the understanding of dengue epidemiology in Thailand, such as seroprevalence data and a record of the proportion of reported cases that are hospitalized.
| Dengue is a global arboviral disease affecting humans. The primary vector is the Aedes aegypti (Linnaeus) mosquito. Dengue is present in the tropical and subtropical regions of the Americas, the eastern Mediterranean, Africa, and the World Health Organization (WHO) Western Pacific and Southeast Asia regions [1]. Countries included within regions designated as Southeast Asia differ according to WHO, political and geographic definitions. Unless otherwise stated, the term Southeast Asia used in this paper refers to the WHO Southeast Asia Region (SEAR).
Globally, more than 2.5 billion people are at risk [1]. The WHO estimates that more than 50 million dengue virus (DENV) infections and 20,000 dengue disease-related deaths occur annually worldwide [2], [3], and a recent disease distribution model estimated there were 390 million DENV infections in 2010, including 96 million apparent infections (i.e., cases that manifest at any level of clinical or subclinical severity). Overall, 70% of these apparent infections occurred in Asia [4].
Thailand observed its first cases of dengue disease in 1949; sporadic cases continued to be reported throughout the 1950s [5], [6] and the first major outbreak of dengue haemorrhagic fever (DHF) was reported in Bangkok in 1958 [7], [8]. There were 2158 cases and 300 deaths in this outbreak [9]. DENV infection is caused by any one of four distinct DENV serotypes (DENV-1, -2, -3 or -4) [1]. Three or possibly four virus types (DENV-1, possibly DENV-2 and two unidentified serotypes) were isolated during the 1958 epidemic [10], and the co-circulation of all four dengue serotypes was demonstrated in the early 1960s in Bangkok [11]. By the late 1970s, the disease was widespread among countries in Southeast Asia and DHF had become a leading cause of hospitalization and death among children in Thailand [12]. There was a major epidemic of dengue disease in 1987, in which 174,285 cases were reported, after which the number of reported cases remained relatively low and stable, with under 100,000 cases reported each year. Two large outbreaks were reported in 1997 and 1998, with 101,689 and 126,348 cases reported, respectively [9], [13]. Before 2004, Thailand reported the highest number of annual dengue disease cases in Southeast Asia, with an average of almost 69,000 cases per year reported between 1985 and 1999 [13]. After 2004, Indonesia reported the highest number of cases from the region, accounting for 57% of the cases reported to the WHO Southeast Asia region in 2006 [13]. The epidemiology of dengue disease in Thailand is characterized by cyclical epidemic activity alternating between years of relatively low and high dengue disease incidence [14], [15].
A reporting system for dengue surveillance in Thailand started in 1958. The national surveillance system for DHF was initiated in 1972 by the Bureau of Epidemiology (BoE), Thai Ministry of Public Health (MoPH) becoming fully operational in 1974 [16]. DF was included in the surveillance system in 1994. Reports for patients diagnosed with dengue disease are collected from hospital in-patients and hospital out-patients from health facilities nationwide, all government hospitals and some private hospitals and clinics (the reporting sites are mostly public hospitals, with a few voluntary reports from private hospitals). It is mandatory to report the confirmed cases, but not for the suspected cases (which is subject to the physicians' willingness to report). The reporting form (Form 506) is used to record demographic data — age, sex, day of onset and the address (locality) where the case occurred, categorised as municipalities (‘cities’ or ‘suburbs’) or ‘other’ (mostly rural) areas [16]. It should be noted that all reported dengue disease cases in Thailand are diagnosed by the trained physician using WHO case definition established since the 1970s, which classify dengue into dengue fever (DF), DHF and dengue shock syndrome (DSS) [17]. Digital or hardcopy reports of dengue disease are transmitted up the system from the local level, initially to provincial health offices and then to the BoE where they are collated and analyzed. Prior to 1999 the reports were sent by post; electronic transmission of reports began in 1999 [16]. For the past 10 years, the Thai surveillance system at the central level has been systematic and relies on electronic-based data although at the local level there is no compulsory electronic proforma, indeed hospitals often generate their own software programs that are compatible with Form 506 for entering and transferring data. Epidemiological data on DF, DHF and DSS in Thailand are disseminated from central departments in the form of weekly newsletters (the BoE Weekly Epidemiological Surveillance Report) and published online on the MoPH website within the Annual epidemiological surveillance reports (AESRs).
Dengue disease laboratory diagnostics in Thailand can be ordered on an individual basis and include dengue virus isolation, viral genome detection by reverse transcription polymerase chain reaction (RT-PCR), four-fold increases of paired sera (haemagglutination inhibition) or IgM >40 U or IgG increasing >100 U. Virological surveillance (virus isolation and serotyping by RT-PCR) is performed by the Department of Medical Science, especially before the outbreak season, which appoints a number of hospitals from around the country to act as sentinel sites. However, only a small proportion of reported cases are tested for DENV infection. Furthermore, the proportion of specimens sent for testing varies between provinces in each region.
Thailand is divided politically into 76 provinces, with the capital, Bangkok, being a special administrative area. A four-region administrative system is used by the MoPH (Figure S1): North (population 11.5 million), Northeast (18.8 million), Central (including Bangkok) (26.3 million), and South (8.9 million) [18]. Thailand has three types of climate, a tropical rain climate in the coastal areas of the east and south, a tropical monsoon climate in the southwestern and southeastern coastal areas, and a tropical wet and dry or savannah climate in the southwest, central and northern regions. Climatic factors such as temperature, rainfall and relative humidity affect the growth and dispersion of the mosquito vector and are known to be associated with dengue outbreaks [19]. In common with other developing tropical and subtropical countries, Thailand has population demographics and socio-economic conditions that favour dengue transmission, such as rapid population growth and rural–urban migration [20], and densely populated areas that provide suitable Aedes mosquito larval habitats [1], [21].
This review describes the epidemiology of dengue disease in Thailand reported in the literature between 2000 and 2011 in the context of the national and regional trends and aims to identify gaps in epidemiological knowledge requiring further research. Incidence (by age and sex), seroprevalence and serotype distribution and other relevant epidemiological data such as geographical distribution are described.
We estimated that a time period of at least 10 years would allow observation of serotype distribution over time and through several epidemics and, in view of the 3–5-year periodicity of dengue outbreaks [7], would also accurately reflect recent changes in dengue disease epidemiology. We set the start date as 1 January 2000, as opposed to an earlier date, to limit the bias that any differences in surveillance practices over time would have on the results. The cut-off for our review period was set as 28 February 2012, the date when we initiated this review.
The overall methodology, search strategy, and inclusion and exclusion criteria for this literature analysis and review are included in a protocol that was developed by a Literature Review Group (LRG). The protocol was based on the preferred reporting items of systematic reviews and meta-analyses (PRISMA) guidelines [22]. The protocol was registered on PROSPERO, an international database of prospectively registered systematic reviews in health and social care managed by the Centre for Reviews and Dissemination, University of York (CRD42012002170: http://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42012002170) on 23 March 2012.
The LRG guided the literature analysis process, defined the search strategy, and prepared the protocol and review documents. Specific search strings for each database were designed with reference to the expanded Medical Subject Headings thesaurus, encompassing the terms ‘dengue’, ‘epidemiology’, and ‘Thailand’. Different search string combinations were used for each electronic database with the aim of increasing the query's sensitivity and specificity. Searches of selected online databases (Table S1) were conducted between 9 February 2012 and 28 February 2012.
As stated in the protocol, studies (as well as conference materials, grey literature and official reports and bulletins) published in either Thai or English between 1 January 2000 and 28 February 2012 were included in the analysis. References not meeting these criteria that were found in databases that did not allow language and/or date limitations were deleted manually at the first review stage. No limits by sex, age and ethnicity of study participants or by study type were imposed, although single-case reports and studies that only reported data for the period before 1 January 2000 were excluded, as were publications of duplicate data sets, unless the articles were reporting different outcome measures. Editorials and reviews of previously published data were also excluded. Additional publications not identified by the search strategy, unpublished reports and grey literature were included if they met the inclusion criteria and were recommended by the LRG.
Sources were reviewed by the LRG to ensure they complied with the search inclusion and exclusion criteria. Following a review of the source titles and abstracts, during which duplicates were removed, the LRG performed a second review of the full text of any published sources selected to make the final selection of relevant sources to include. In an amendment to the original protocol the Literature Review Group sanctioned the extraction of surveillance data for 2011 from the MoPH Bureau of Epidemiology Surveillance Database website on 16 July 2012.
We chose not to exclude articles and other data sources nor formally rank them on the basis of the quality of evidence. Whilst an assessment of study quality may add value to a literature review, we were of the view that given the expected high proportion of surveillance data among the available data sources and the nature of surveillance data (passive reporting of clinically-suspected dengue), such quality assessment would not add value to our review.
The selected data sources were collated and summarized using a data extraction instrument developed as a series of Excel (Microsoft Corp., Redmond, WA) spreadsheets. Data from literature reviews of previously published peer-reviewed studies and pre-2000 data published within the search period were not extracted. The original data sources and the extraction tables were made available to all members of the LRG for review and analysis. In view of the expected heterogeneity of eligible studies in terms of selection and number and classification of cases, a meta-analysis was not conducted; a narrative synthesis of our findings is presented. For the purposes of the analysis, we defined national epidemics as those years in which the number of cases was above the 75th percentile for the period.
This review concentrates on national epidemiological data collated from several sources, including the latest data from the Thailand MoPH (Table S1). The literature searches identified 610 relevant data sources; of these, 40 fulfilled the inclusion criteria for the analysis (Figure 1; Table S2). Most national epidemiological data were derived from the annual surveys or statistical tables produced by the MoPH (12 sources [23]–[34]). Of the remaining 28 articles and reports, the majority were journal articles that mainly described regional epidemiological data derived from surveys and studies conducted in individual regions and provinces (Table S2) and are used to support the national data with regard to regional incidence, serotype and age distribution.
The AESRs published by the MoPH provide a source of country-wide reporting of dengue disease statistics for 2000–2011 [23]–[34], albeit with some missing data owing to reporting variations.
Between 2000 and 2011, more than 860,000 dengue disease cases were reported, corresponding to an annual average of approximately 72,000 cases and 100 deaths, and an average annual incidence of 115 cases per 100,000 population. Peaks in the number of cases (national epidemics) that were above the 75th percentile (102,213) for the period occurred in 2001, 2002 and 2010, when 139,355 cases (incidence rate 224/100,000 population), 114,800 cases (183/100,000), and 116,947 cases (177/100,000) were reported, respectively; another peak of 89,626 cases (142/100,000), which was at the 70th percentile, occurred in 2008 (Figure 2). The lowest incidence occurred in 2000 (30.14/100,000) [23]–[34].
Since 2002, the proportion of the total number of reported cases of DF and DHF reversed, while the proportion of DSS remained relatively stable over the decade, ranging between approximately 1% and 3% (Figure 3) [23]–[34]. The reasons for the change in DF∶DHF ratio change are more probably more related to reporting behaviours than changes to the reporting system. During the 2000–2011 period there were no fundamental changes to the case reporting system. There has however been a change in reporting behaviour over the review period. A DF diagnosis relies on voluntary reports, which reflect to physicians' attention and willingness to report and their sense of importance of this matter. Moreover, in 1999 the King's Project (a large prevention and control programme for dengue) was introduced in which the aim was to increase people's knowledge of the disease through education and television advertisements [35]. Consequently, patients attended hospital earlier resulting in early diagnosis. Physicians cooperated with the programme by reporting DF, whereas previously reporting was focused on severe forms of dengue (DHF and DSS). Improved physician awareness to the disease was assisted by better diagnostic capabilities. In particular, laboratory facilities improved in many areas and the results were reported back to local hospital from the central laboratory faster than before. In addition, following the 2005 avian flu outbreak more PCR laboratories became available at the regional level [26]. Greater diagnostic capabilities (possibly through the wider use of immunoglobulin G (IgG)/IgM test kits and the NS1 antigen test, which may help confirm the diagnosis of mild dengue virus infections) and changes in surveillance methods may also have contributed to the increasing proportion of DF cases detailed in the AESR in recent years.
The number of deaths due to dengue disease reported between 2000 and 2011, and the mortality rate (deaths per 100,000 population), broadly reflect the number and incidence of cases reported. There were 1216 deaths reported from dengue disease between 2000 and 2011, an average of 0.16 deaths per 100,000 population [23]–[34]. The highest mortality rate occurred during the large 2001 epidemic (245 deaths, 0.39 deaths/100,000 population) [23], [24]. Between 2003 and 2011, the average case fatality rate (CFR) reported by the MoPH for DHF was 0.05% (0.03–0.09) and for DSS it was 4.45% (range: 4.04–5.92); the highest CFR for DSS was in 2006 (5.92%) [26]–[34]. There were no deaths attributed to DF over this period.
No clear trend over time in the number of reported dengue disease cases could be discerned as the pattern of the annual number of cases of dengue disease over the review period was complicated by epidemic years. There was an overall decline in case fatality rates reported between 2000 and 2010, reflecting rates reported in most of the dengue disease endemic countries in the Southeast Asia region between 1995 and 2000 [36]. These patterns may be the result of changes in reporting and improvements in case management.
Annual dengue disease case numbers evident at the national level were broadly repeated at the regional level. Consistent with its higher population density, the Central region reported the highest number of dengue cases in most years and the most deaths over the period of the review. The most cases reported was in 2002 from the northeast region (37,191 cases) [23]–[34]. The reported incidence rate was highest in the southern region in 2001, 2002, 2005, 2007 and 2010 (Figure 4, Table S3); the incidence rate in the southern region in 2002 was more than double that reported in the other regions (402.54/100,000 population) [23]–[34]. The highest mortality rate was reported from the southern region in 2001 and 2002 (0.77/100,000 population), as well as in 2010 (0.66/100,000 population) [23]–[34].
Available regional data on the proportion of DF, DHF and DSS cases for the years 2003–2011 [26]–[34] show similar increases in the proportion of reported DF cases to those seen at the national level (Central: from 20% to 39%; North: from 33% to 46%; Northeast: from 31% to 46%; South: from 34% to 48%).
In Thailand, dengue remains a disease of children and young adults, with most cases occurring in individuals aged between 5 years and 24 years, who represent one third of the population (Table S4). However, the age group with the highest incidence changed from those aged 5–9 years to those aged 10–14 years in 2002, and there has been a general shift in age group predominance of dengue disease over the survey period from younger towards older individuals over 15 years of age [31]–[34], [37]–[39] (Figure 5a and 5b), continuing a trend that was first observed in the 1980s [36], . These findings are consistent with a recent publication reporting a significant increase in the age at dengue exposure in December 2010 in Rayong Province, Southeast Thailand [42].
Throughout the review period, relatively more cases of severe dengue (DHF and DSS) were reported in individuals aged between 5 and 14 years compared with those aged 15 years or older. In particular, DSS was less common in individuals aged 15 years or older. Data from cohort studies indicate that many children in the 5–14 years age group may be experiencing a second infection of dengue, which could account for the high incidence of severe disease in this age group [41]. Lower incidence rates of severe disease in older age groups could be due to reduced exposure to infection or reduced severity of disease in individuals experiencing their third or fourth infection [43].
Over the review period, approximately 70% of deaths due to dengue disease reported to the MoPH were in patients younger than 15 years. Typically, the highest CFRs were seen either in young children aged 0–4 years or in older adults aged 55–64 years, a trend that likely reflects the susceptibility of the young and old to more adverse consequences of dengue disease and its clinical management, as well as the risk associated with comorbidities in older adults [44]. However, the number of reported cases in those aged 55 years and above is small compared to the other age groups. Individuals aged over 65 years had the lowest reported incidence rate of dengue overall, and the only case fatalities in this age group were recorded in 2001, 2005 and 2010.
Comparable regional data for age-related distribution of dengue were not recorded in the studies selected for this review. Individual studies suggest a pattern similar to that seen nationally, with younger age groups more likely to contract dengue than adults and the elderly [37]–[39].
Although more females than males were reported to have the disease in 2009 (male∶female ratio 1∶1.6) [32], [45], [46], in general, slightly more males than females were affected by dengue over the survey period, with reported male∶female ratios of between 1.1∶1 and 1.2∶1 [26], [29]–[31], [33]. These differences may be due to differences between the sexes in health-seeking behaviours in Thailand [37].
The available data show a seasonal peak in the numbers of cases (Figure 6) and deaths (data not shown) due to dengue between May and September annually [23]–[34], [47], [48] which is probably due to seasonal changes in climate [14], and the association between the active season of the vectors and the wettest months. Thus the pattern coincides with the rainy season in Thailand, which, although it varies slightly from region to region and is largely dominated by the monsoon, can be classified broadly as May/June to October.
At the time of this review, comprehensive regional DENV serotype data for Thailand from the MoPH AESRs were only available for the period 2005–2010. These serotype data show a broadly similar pattern in each region, with a reduction in the proportion of DENV-1 and an increase in the proportion of DENV-2 isolates over that period (Figure 7) [28]–[33]. DENV-4 peaked during 2005 and 2006 and then declined, but remained in circulation in the Central region throughout this 5-year period, albeit at a decreasing percentage of all dengue disease cases (4.6–6.2% during 2008–2010, compared with 10.4% in 2007 and 46.1% in 2005). By contrast, in 2009 and 2010, DENV-4 was not isolated in samples from the North or Northeast regions and was reported at <2% in the South region. DENV-3 circulated in all regions throughout the whole period.
In general, between 2000 and 2010, DENV-1 and DENV-2 were the most commonly reported serotypes in national and/or regional studies in Thailand [23]–[33], [39], [48]–[53] (Table S5). However, during the years 2000–2002 and 2008–2010, DENV-3 was more commonly identified than during the middle part of the decade [23]–[33]; between 2003 and 2008, reports of DENV-4 were more common than during 2000–2002 and 2009–2010. In 2010, the most commonly identified serotype was DENV-2, representing over half of all those isolated (54.6%), followed by DENV-1 (25.5%), DENV-3 (15.3%) and DENV-4 (4.6%) [33]. Anantapreecha et al. also found similar temporal and spatial changes in the predominant DENV serotype [50], [54], [55].
Whereas most epidemiological reports of dengue in Thailand address the magnitude of clinically apparent infections, a number of studies published during the review period investigated both apparent and inapparent infections. Inapparent infections may have important public health implications in understanding virus transmission and the pathogenesis of dengue disease illness [56].
A variable proportion of inapparent infections relative to clinically apparent infections were reported [48], [56]–[60]. An active case surveillance in 2119 primary school children (median age 9.3 years) in a rural setting in Kamphaeng Phet, North region, by Endy et al. reported an overall incidence of dengue infection for the year 2000 of 2.2%: 0.8% symptomatic infections and 1.4% clinically inapparent infections, a symptomatic to inapparent ratio (S∶I ratio) of 1∶1.75 [48]. A later study reported an overall S∶I ratio of 1∶3 for the period 2004–2008 [56], [57]. Similarly, during a large DF/DHF disease outbreak in Nakhon Pathom province (Central region) in 2001, 8.8% of individuals (age range: 0 years to over 50 years) had an inapparent infection with dengue virus, as determined by IgM positivity, over a 2-month period between March and April. Most of the serologically positive individuals (80.8%) reported no previous fever [58].
Like many surveillance programmes the starting point for reporting in Thailand is a visit to a healthcare provider or hospitalization. As such, the national surveillance data may be incomplete and likely under-reported similar to other Southeast Asian countries [61]. A recent analysis of data from prospectively followed cohorts with laboratory confirmation of dengue cases show that dengue incidence is under-recognized in Thailand and Cambodia by more than eight-fold [62]. Consequently changes in the level of healthcare attendance or in the level of reporting to surveillance system by physicians (as discussed above) may also affect the under- or over-reporting of dengue disease.
Epidemiological knowledge in Thailand benefits from a nationwide surveillance system including virological surveillance, complemented by several local studies including cohort surveys. However, at the time of the review, some gaps in the epidemiological information regarding dengue disease in Thailand were identified such as age-stratified seroprevalence data, and data relating to the proportion of hospitalized cases in the reported cases which are not easily available.
This literature review presents the epidemiology of dengue disease in Thailand over the period 2000–2011. A key strength of this survey and analysis is that it describes the epidemiological data from a national aspect rather than from limited study site data. In addition, the review protocol aimed to minimize potential exclusions of valuable data sources including MoPH data, as well as searching for relevant books, unpublished data, abstracts and dissertations. More than 600 data sources were screened and the selected sources were subjected to a comprehensive data extraction method to capture the data, which adds strength to this review. However, by its very nature, this literature review captured mainly publicly available data and studies and is, therefore, subject to publication bias; the data presented here should be interpreted accordingly.
Another limitation of this review is that much of the peer-reviewed data are drawn from certain regions, which may skew the findings. Use of consistent MoPH data in the analysis for this review has minimized potential bias from studies using different methodologies for collecting information, confirming disease and reporting data, although this does not guarantee a consistent approach. National surveillance systems are subject to the limitations inherent to passive surveillance data, such as under-reporting, misreporting, and reporting biases. The methods and requirements for the surveillance systems in Thailand have also changed over time and the impact of the historical evolution of the systems is unknown.
In a number of papers, associations were proposed between both the burden and severity of disease and the specific DENV serotypes circulating in the population, the sequence of DENV serotypes causing primary and secondary infections or the dengue incidence in the preceding season, which indicate the multifactorial processes that influence dengue disease severity [48], [51], [52], [56], [57], [61], [63]–[66]. For example, DENV-1 has been linked with high morbidity and low mortality [61], and DSS has been associated with secondary infections attributable to DENV-2 [63]. DENV-4, which is generally found at low frequency in Southeast Asia [64], is linked to lower levels of virulence [65] and lower reported incidence [66]. Findings such as these have prompted suggestions that changes in predominant serotypes are associated with changes in disease severity [51] (see Guzman et al., 2013 for full review [41]). While such research papers have contributed to understanding the dengue disease, in the absence of nationwide data it is not clear whether the results are circumstantial (site specific) and thus it is difficult to apply the findings to other parts of the country.
Dengue disease is a public health priority in Southeast Asia, and Thailand contributes substantially to the regional disease burden. Over the review period wide yearly variations in incidence occurred, with regular epidemics in 2001, 2008 and 2010 with dengue disease remaining a highly seasonal disease. Age group distribution of dengue disease shifted during the review period from younger towards older persons even if dengue disease in Thailand remain a childhood disease predominantly with higher severity reported in young children. Heterogeneous geographical patterns of the disease was observed from 2000 to 2011 including higher incidence rates reported in the South and serotype distribution variations in time and place. Passive nationwide surveillance system in Thailand is a source of consistent data including severity, age- and serotype related information. Further information on seroprevalence and on the proportion of hospitalized cases among all reported cases would be beneficial to the description and understanding of dengue epidemiology in Thailand.
|
10.1371/journal.ppat.1001113 | Dynamics of the Multiplicity of Cellular Infection in a Plant Virus | Recombination, complementation and competition profoundly influence virus evolution and epidemiology. Since viruses are intracellular parasites, the basic parameter determining the potential for such interactions is the multiplicity of cellular infection (cellular MOI), i.e. the number of viral genome units that effectively infect a cell. The cellular MOI values that prevail in host organisms have rarely been investigated, and whether they remain constant or change widely during host invasion is totally unknown. Here, we fill this experimental gap by presenting the first detailed analysis of the dynamics of the cellular MOI during colonization of a host plant by a virus. Our results reveal ample variations between different leaf levels during the course of infection, with values starting close to 2 and increasing up to 13 before decreasing to initial levels in the latest infection stages. By revealing wide dynamic changes throughout a single infection, we here illustrate the existence of complex scenarios where the opportunity for recombination, complementation and competition among viral genomes changes greatly at different infection phases and at different locations within a multi-cellular host.
| Viruses are fast evolving organisms for which changes in fitness and virulence are driven by interactions between genomes such as recombination, functional complementation, and competition. Viruses being intra-cellular parasites, one basic parameter determines the potential for such interactions: the cellular multiplicity of infection (cellular MOI), defined as the number of genome units actually penetrating and co-replicating within individual cells of the host. Despite its importance for virus evolution, this trait has scarcely been investigated. For example, there are only three point estimates for eukaryote-infecting viruses while the possibility that the cellular MOI may vary during the infection or across organs of a given host individual has never been conclusively addressed. By monitoring the cellular MOI in plants infected by the Cauliflower mosaic virus we found remarkably ample variations during the development of the infection process in successive leaf levels. Our results reveal that the opportunities for recombination, complementation and competition among viral genomes can greatly change at different infection phases and at different locations within a multi-cellular host.
| Intracellular interactions among co-infecting viral genomes play a central role in viral evolution and ecology as they determine three important phenomena: (i) competition and selection, (ii) re-association with other genetic backgrounds through recombination, and (iii) functional complementation of (or by) other genomes. The overall intensity of these phenomena depends on the probability of encounter of the countless variants of a viral population within the multitude of individual cells composing the host. The basic parameter determining the potential for such encounters is the multiplicity of cellular infection (cellular MOI), i.e. the number of viral genomes (number of genome units) that enter and effectively replicate in individual cells. For example, a cellular MOI above 1 in a given cell corresponds to the co-infection of the same cell by several viral variants, favoring recombination, complementation, and intra-cellular competition; on the contrary, a cellular MOI of 1 will preclude these phenomena. Notably, complementation between viral genomes co-infecting individual cells has been investigated both theoretically and experimentally for the bacteriophage Φ6, and has been demonstrated to be a predominant evolutionary force which directly depends on the MOI, as defined here [1]–[5]. More generally, complementation (shared production of viral polymerase, movement proteins, suppressors of host defenses, structural proteins of the virion, etc.) is undoubtedly frequent in viral populations and is at the basis of collective actions, which largely operate at the intra-cellular level.
Empirical investigations on the cellular MOI are extremely scarce. In fact, the values for this parameter that prevail in nature remain elusive, and their putative dynamic changes during colonization of a host by a virus population have never been conclusively investigated. Formal MOI estimates have been established in only four systems: one bacteriophage [6], [7], one insect virus [8], and two plant viruses [9], [10]. For the bacteriophage and the insect virus, the MOI was considered as a single value calculated at one single time point. For plant viruses, both studies were limited to the initial onset of the host infection. Miyashita and collaborators [10] defined the number of virions infecting individual cells in a local lesion within a leaf immediately following the artificial inoculation of the virus in a single cell. González-Jara and collaborators [9] went a little further by analyzing the MOI both in the artificially inoculated leaf, as well as in the very first leaf where the virus appears through natural systemic movement. These empirical analyses provide important insights into the MOI, but at a very limited spatial and temporal scale during host invasion, thus leaving two remarkable lacunas. First, they cannot inform on whether MOI is constant and homogeneous throughout the entire host and infection process or, on the contrary, subject to ample dynamic changes in time and/or space. Such opposite situations could have totally different implications for viral population genetics (further discussed later). Second, and consequently, the estimated values might not even approximate the average MOI that could be calculated from the entire host across the whole infection process, potentially yielding a totally biased view of the reality.
The present study fills these important gaps by describing the first extensive spatio-temporal monitoring of the cellular MOI of a eukaryotic virus, the Cauliflower mosaic virus (CaMV), from the onset of the systemic invasion until senescence of its host plant. CaMV is an aphid-transmitted double-stranded DNA virus which replicates through reverse transcription of a genomic RNA intermediate, and is thus expected to have a high mutation rate [11], [12]. This virus has been shown to recombine extremely frequently [13], indirectly indicating an elevated cellular MOI. Our analysis at different time points and at different leaf levels demonstrates the occurrence of important dynamic changes of the MOI throughout the infection cycle, starting close to 2 early in infection, peaking at 13, and then decreasing to initial levels. Most importantly, we obtained similar MOI values under different experimental conditions of inoculum doses, plant growth, and inoculation methods - including natural inoculation by aphids - suggesting that our results are robust to experimental conditions, and thus faithfully illustrate what actually happens during a natural CaMV infection cycle.
Our aim was to evaluate the intensity with which the variants in a viral population can interact with each other at the cellular level, or, in other words, to assess how frequently these variants co-exist in individual cells. To this end, we estimated the cellular MOI and its putative dynamic changes during the invasion of turnip plants (Brassica rapa) by CaMV.
Host plants were co-inoculated mechanically with VIT1 and VIT3, two equi-competitive tagged CaMV variants, previously characterized in [14], [15], differing only in a 40-bp non-coding insert that allows their specific identification. In all experiments, the two variants were co-inoculated at the same time and location in order to mimic the situation where a mutant coexists with other genomes from its appearance.
The principle of the procedure in all time-course analyses was as follows (see full details in Materials and Methods). Six plants were inoculated in parallel and sampled at different time points, starting from the development of the first symptoms of systemic infection until flowering and senescence. At each sampling date a single mature leaf was sampled from the same leaf level in all plants. In each individual leaf two parameters were measured: (i) the ratio of the variants VIT1 and VIT3, and (ii) the proportion of cells infected by both variants. From these data, we derived a maximum likelihood estimate of the average number of viral genomes infecting individual cells (i.e., the MOI) at each leaf level.
In fact, assuming that the monitored viral variants infect cells at random, the probability for a given variant to enter a cell directly depends on both its frequency within the corresponding leaf and the total number of viral genomes that enter each cell (MOI). Given the known relative frequency of the variants VIT1/VIT3 within each analyzed leaf, we estimate the average MOI for which the likelihood to lead to the observed proportion of cells co-infected by the two variants is maximum. The full details and formulas for this maximum likelihood framework are given in the Materials and Methods.
Preliminary experiments were designed to define the VIT1/VIT3 ratio to be used in the inoculum in order to obtain an intermediate proportion of cells co-infected by both variants (when all cells contain both variants, it becomes impossible to estimate the MOI). The outcome of these preliminary experiments indicated a very high proportion of cells co-infected by both VIT1 and VIT3, and ample variations of this proportion at different sampling dates. Because variations were also important between repeated plants at each sampling date, these preliminary trials were principally used to adjust and better control our sampling protocol, and are thus fully described in the Materials and Methods, and shown in Figure S1 and Table S1.
In order to remove irrelevant sources of variation as much as possible, we repeated the whole time-course experiment homogenizing parameters during plant growth and leaf sampling (see Materials and Methods). In particular, the exact same leaf levels were collected in all six repeated plants, and all leaves were collected 13 days after their first appearance on the plant (when the leaves were 13 days old).
The results from this controlled repetition of the time-course monitoring of CaMV cellular MOI are shown in Figures 1 and 2 (the full data set is provided in Table S2). The VIT1/VIT3 ratio within infected leaves was close to that in the initial inoculum, and remained nearly constant throughout the experiment (Figure 1 plain line). The slight differences in the VIT1 relative frequency at different time points were not statistically significant (linear mixed-effects model; P = 0.112; F = 2.16; dfnum = 4; dfden = 20). Moreover, the slope of the linear regression of VIT1 relative frequency versus time was not significantly different from 0 (P = 0.078; F = 3.46; dfnum = 1; dfden = 23), consistently with the equi-competitiveness of VIT1 and VIT3 in our experimental condition (see Materials and Methods). In contrast, the proportion of cells infected by both variants varied significantly between leaf levels (Figure 1 dotted line; linear mixed-effects model; P = 3.1×10−4; F = 8.71; dfnum = 4; dfden = 20). In line with the preliminary results presented in Figure S1, we found that the estimated MOI values (Figure 2) followed a bell-shaped curve with a peak at approximately 13 genomes per cell (in leaf level 21), and minima of around 2 at the early symptoms appearance (leaf level 6) and during flowering preceding plant senescence (leaf level 43). Variations between the six repeated plants were lower than in the preliminary experiment mentioned above, and the statistical analysis confirmed both a significant MOI increase from leaf level 6 to 21 (Tukey HSD test; P = 0.027), and a significant decrease from leaf level 21 to 43 (Tukey HSD test; P = 0.048). Because the leaves successively developing on the same plant were all analyzed at the same leaf-age, we conclude that they were infected by CaMV at a significantly different MOI.
This conclusion was further confirmed by an alternative statistical approach where the MOI in each leaf-level was estimated within a full maximum likelihood framework (described in the materials and methods) which results are presented and discussed in detail in the Supporting Online Information (Figure S2).
Our next goal was to test whether our results were specific to the experimental design, in particular to the mechanical inoculation process, which is commonly used in laboratories but does not correspond to the natural mode of inoculation of CaMV. Thus, we investigated how the MOI estimates varied in different experimental conditions (Figure 3, the full dataset is provided as Supporting online Information in Table S3).
These experimental conditions included changes (i) in the plant growing conditions, (ii) in the virus dose inoculated mechanically, and (iii) in the mode of inoculation (including aphid transmission). The experimental design was similar to that in the time-course experiment described above except that, for practical reasons discussed later, only two leaf levels were sampled (leaves 12 and 33). Consequently, we could not investigate the effects of these treatments on the MOI dynamics, but we could nevertheless compare their respective values for these two leaf levels. Figure 3 shows that all conditions yielded values of the same order of magnitude as in the other experiments reported in Figure 2 (and also in Figure S1). A linear mixed-effects model (with leaf level and treatment and their interaction as fixed effects, and plant as a random effect) revealed that treatment did not affect MOI (P = 0.99; F = 0.042; dfnum = 3; dfden = 18), while leaf level did (P = 0.016; F = 7.01; dfnum = 1; dfden = 18). The interaction of leaf level and treatment was marginally significant (P = 0.0488; F = 3.19; dfnum = 3; dfden = 18). The results of the “leaf level” and “leaf level” x “treatment” interaction are driven by the two treatments shown in “red” and “yellow” in Figure 3 (increased viral dose and different growing conditions respectively). They suggest that the MOI dynamics might be shifted to the “left”, i.e. occur faster, under these conditions. While fully testing this possibility would have required more time points in all four treatments, our results strongly suggest that our estimates are robust and most likely representative of MOI values in nature. This important conclusion is particularly supported by the condition where CaMV was inoculated by aphid vectors (shown in green in Figure 3), which is the only mode of transmission reported for this virus in nature.
We here report the first time-course analysis of the cellular multiplicity of infection of a virus invading a eukaryotic host, from the beginning to the end of the host infection process. Our experimental design, monitoring the MOI at different time points and in different locations within the host, was intended to accommodate the likely heterogeneous structure of viral populations in different organs and at different phases of the infection cycle, as suggested by previous studies both in animals [16] and plants [17]. The genetic markers used in CaMV VIT1 and VIT3 are both neutral [15] and highly stable: they are not deleted from the viral genome after at least three successive passages in host plants [14]. These properties enabled the monitoring of the MOI for a very long period (over 80 days), without biases due either to the competitive exclusion of one variant by the other, or to increasing frequency of marker-deleted genomes within the population. The flipside of the use of these markers is that the search for cells co-infected by VIT1 and VIT3 is extremely tedious and time-consuming [18], and this is precisely why we have limited the study of the robustness of our results to the experimental conditions to only two leaf levels (Figure 3). The rationale for choosing VIT1 and VIT3 markers rather than seemingly more amenable markers (such as fluorescent protein genes) allowing high throughput detection in single cells is fully explained in the Materials and Methods. We simply wish to mention here that VIT1 and VIT3 markers can be detected within infected cells for unlimited amounts of time. Upon replication, the CaMV forms characteristic and very stable electron-dense inclusion bodies (“viral factories”), where hundreds or thousands of mature viral particles accumulate and remain sequestered indefinitely [19]. In consequence, once a CaMV variant has entered a cell and replicated, it likely remains detectable by our nested-PCR procedure until cell death.
In preliminary experiments where plants were co-inoculated with both VIT1 and VIT3 at a 1∶1 ratio, we rapidly observed nearly 100% of the cells infected by both variants. This observation is extremely interesting because it indicates that the CaMV variants are not spatially segregated in contrast to most RNA plant viruses [10], [20]–[22]. Together with the equi-competitiveness of VIT1 and VIT3, this observation is consistent with the assumption that CaMV variants infect cells at random within a leaf. Thus, assuming that the number of genomes of a given viral variant entering a cell follows a Poisson distribution, which is at the basis of most statistical methods estimating the MOI [7], [8], [10], appears appropriate in the case of CaMV.
The cellular MOI in a given host/virus association depends a priori on two parameters: the number of viral infectious units available per cell (viral load), and the maximal number of these units that can effectively co-infect the same cell. Variations in these parameters should influence cellular MOI values and explain the dynamics observed in CaMV-infected plants. It is reasonable to imagine that the viral load increases over time in the plant, with a concomitant increase in the multiple infections of cells, as more and more infected leaves develop and shed virus into the phloem. However, the decline in cellular MOI late in infection contradicts this prediction. Since VIT1 and VIT3 are equi-competitive [15], this decline cannot be explained by the dominance of one variant over the other, as confirmed by the unchanged VIT1/VIT3 ratio throughout the experiment (Figure 1 and S1A). One possibility would be a host developmental or physiological effect on the MOI, related to the previously described impairment of virus infection upon flowering [23], or to the onset of a plant defense mechanism [24], [25], with a resulting drop in viral load.
Another explanation of the observed MOI pattern would be a changing balance between benefits and costs of multiple infection of cells. The benefits are basically those derived from recombination [26]–[28], and from cooperation among the genomes co-existing within the same cell (i.e. collective action and mutualistic complementation). The costs of multiple infection arise from the competition for cell resources, and from the evolution of “cheater” genotypes, better adapted to this competition than to host exploitation in single infections [1]. The best studied example of the latter phenomenon is the recurrent observation of defective interfering particles (DIPs) appearing in virus populations [29]–[32]. The CaMV recombines at very high rates in turnip [13], and cooperative behaviours in this virus exist at least during the transmission process [33]–[35] and the suppression of gene silencing [36], [37]. One could thus hypothesize that an increasing cellular MOI could benefit CaMV during the invasion of a host, up to a value (around 13) where the costs would overwhelm the benefits. For example, as indicated above, a high MOI value might increase the proportion of DIPs [7] up to a threshold were functional genomes can no longer sustain the growth of the viral population. The resulting crash of the virus load could therefore explain the MOI drop late in infection. A quantitative monitoring of the virus load within the vasculature of the plants, and an estimate of the frequency of DIPs therein, would support or disqualify these hypotheses.
The MOI values and their dynamic changes reported here cannot be directly compared with the situations in other host/virus associations, because no equivalent information is available. The MOI estimate around 4 for a baculovirus infecting lepidopteran insects possibly represents an average over the complete infection process [8]. For the sake of comparison we calculated the equivalent average MOI for CaMV by compiling the full data sets from time-course experiments shown in Figures S1 and 2, and found values of the same order of magnitude, 7.87±2.03 and 6.67±1.43 (mean±SE) respectively. Whether the value of 4 found in baculovirus-infected caterpillars resulted from a constant MOI throughout the infection cycle or represented the average of ample variations, as is the case here for CaMV, is not known. In two recent studies on plant viruses, the MOI was investigated in the artificially inoculated leaf. Very early after inoculation, the values found for the Tobacco mosaic virus (TMV) infecting Nicotiana benthamiana plants [9], and for the Soil-born wheat mosaic virus (SBMV) infecting Chenopodium quinoa plants [10] were remarkably similar (between 5 and 6). We did not analyze the inoculated leaf in our study on CaMV, because the mechanical inoculation procedure does not reflect any natural process, and how this might or might not bias the viral infection of neighboring cells is hard to evaluate. The study on TMV [9] also reported the analysis of MOI values in the first systemically infected leaf, where the virus enters via its natural route (the plant vascular system). In this leaf, the MOI of TMV was estimated to lie between 1 and 4, very close to our estimate for CaMV in leaf level 6 (mean = 2.73; SE = 1.73) which also represents the first systemically infected leaf level. Interestingly, the same authors assessed a putative time variation of the TMV MOI within this single leaf (a question not tested here on CaMV), and they concluded that the TMV MOI can change through time. However, this conclusion was challenged in the discussion by Miyashita and Kishino [10], thus leaving opened the basic question of a MOI change with time. On this important question, we here definitely demonstrate that dynamic changes of the MOI indeed occur with large amplitudes during the whole host infection by CaMV. Unfortunately, this remarkable phenomenon cannot be compared to the situation with TMV and SBMV, where the viral infection was not monitored in upper leaf-levels being systemically infected.
A dynamic MOI similar to that described here for CaMV likely occurs in other systems, as suggested in HIV by the number of proviruses per cell indicating an elevated MOI [38], and by the fluctuating rates of cell co-infection in cell cultures [39]. However, alternative scenarios are also possible since segregation and isolation of genetic variants in different cells of the same host has been repeatedly observed for several plant viruses [17], [20]–[22], [40]–[42], suggesting more stringent limits to cellular co-infection, and thus to MOI values, at least within some specific cell types, organs, or tissues.
At present, no theoretical predictions are available to fuel a discussion on the potentially different impact that a steady or a variable cellular MOI could have on the evolution of the corresponding viral populations. The few theoretical and experimental studies addressing specifically the role of MOI in the evolution of the phage Φ6 were considering low, intermediate, or high values, but always constant in a given viral line (reviewed in [5]). While we here observe ample MOI variations during host infection by CaMV, we cannot control it, and a comparison with a constant MOI is thus far impossible in this system. In contrast, other virus-host models, like phage systems, would allow the experimental evolution of lines with constant or changing MOI, with various different patterns but similar average value, and the outcome on the evolution of the average fitness in each line would be extremely interesting.
Beyond the within-host scale of virus evolution, a specific pattern of variable cellular MOI might have important implications also at a higher organization level, in a broader ecological context. For instance, in the specific case of CaMV, it is possible that populations evolve under different cellular MOI values depending on the vector species. This virus can indeed be transmitted by several aphid species [43] with different behaviors: colonizing the plant or not, feeding from lower or upper leaves, or from younger or older plants. Given the implications of the MOI for viral evolution and epidemiology, our results urgently call for a broader investigation of this important trait in a wide panel of natural virus/host associations, characterizing the values, their putative dynamic changes and the underlying mechanisms.
The two engineered CaMV variants, VIT1 and VIT3, have been previously characterized in detail [14]. Both are infectious full-length clones of the CaMV Cabb-S isolate [44] harboring a 40-bp DNA insert used as a specific genetic marker that can be quantified in a mixed population [14] and specifically detected within single cells [18]. Such markers were demonstrated to be stably maintained within CaMV genomes over at least three successive passages in turnip host plants [14]. Co-infecting CaMV-VIT1 and -VIT3 proved equi-competitive during turnip plant invasion [15].
The virus particles used in the inoculum were purified from plants infected with each variant individually and quantified as previously described [45]. The inoculum was prepared by mixing purified virus particles and a convenient ratio of 4/1 (VIT1/VIT3) was determined in preliminary experiments (see below). For all time-course analyses of the MOI six healthy plantlets were mechanically inoculated in parallel with 400 ng of virus particles per plantlet as previously described [18], except for conditions with a different viral dose or inoculation by aphids. When symptoms appeared on systemically infected leaves they were harvested and processed as described below.
Unless otherwise indicated turnip plants (Brassica rapa cv. “Just Right”) were maintained in an insect-proof growth chamber under controlled conditions (24/15°C day/night with a photoperiod of 15/9 h day/night).
The actual VIT1/VIT3 ratio in each sampled leaf was estimated from a pool of ∼3000 protoplasts per leaf, using real-time quantitative PCR (PCR conditions and primer sequences are provided in Table S4). A linear mixed-effects model, taking into account the repeated measures within each plant, was used to test for changes in VIT1 frequency between dates (fixed effect) within plants (random effect); it showed that VIT1 frequency was close to that in the mixed inoculum and varied only slightly (if at all) over time (Figure 1 and Supporting online Information Figure S1 and Table S1), confirming previous estimates of marker neutrality [15].
Thirty protoplasts from each sampled leaf were analyzed individually to determine the co-occurrence of VIT1 and VIT3 genomes and thus the frequency of cell infected by both variants. The region of the CaMV genome bearing the genetic markers was amplified from each isolated cell by single-cell nested-PCR, and VIT1 and VIT3 sequences were specifically identified in the amplicons by high resolution melting analysis exactly as described previously [18]. A linear mixed-effects model, taking into account the repeated measures within each plant, was used to test if the proportion of cells infected by both variants varied between leaf-levels (fixed effect) within plants (random effect).
Despite the tediousness of the single-cell detection of such markers [18], we have altogether analyzed over 3400 individual cells (Table S1, S2 and S3). The use of another type of markers, based on the insertion of genes encoding fluorescent proteins such as GFP (green) and RFP (red) into viral genomes, would have provided a straightforward high-throughput approach to visualize their presence within single cells, using for example epifluorescence microscopy (on tissues or extracted protoplasts). However, in contrast to the VIT1 and VIT3 markers used here, such fluorescent markers have a number of drawbacks which limits their usefulness for studies such as that presented in this paper: (i) currently available fluorescent protein genes cannot be introduced in CaMV and in other viruses with an icosahedral shell, because of the limited size of the encapsidated genome [46], [47]; (ii) GFP can diffuse autonomously from cell to cell in plants [48], a phenomenon potentially misleading in identifying cells infected with a GFP-expressing virus; (iii) two Tobacco mosaic virus variants, respectively expressing GFP and RFP, proved differentially competitive in co-infected plants [9], and we observed a similar phenomenon with Turnip mosaic virus (unpublished results); (iv), these GFP or RFP markers are often rapidly deleted from the genomes of plant viruses [49], [50], a phenomenon incompatible with their monitoring throughout the infection process.
The MOI was inferred with a maximum likelihood procedure from (i) the relative proportion of the two variants measured in each sampled leaf, and (ii) in the same leaf, the number of cells infected by both variants among the infected cells.
Assuming that cell infections occur in a random and independent manner for both variants, the number of genomes of a given variant entering a cell follows a Poisson distribution with a parameter equal to the product between the cellular MOI (λ) and the relative frequency of this variant in the sampled leaf (pi, for VIT1). The null class of each Poisson distribution corresponds to the probability of not being infected by the corresponding variant. Thus, in the ith sampled leaf, the probability for a given infected cell to be co-infected by the two variants is , and, among the Ni infected cells observed within this leaf, the number of co-infected cells has a binomial distribution with parameters Ni and pc,i. The corresponding likelihood function is: , where ki is the observed number of cells infected by both variants within the ith sample. The MOI within each sample is then easily derived as the maximum likelihood estimate of λ. A linear mixed-effects model, taking into account the repeated measures within each plant, was used to test if the MOI varied between treatments and between dates (fixed effects), within each plant (random effect). The significance of MOI differences between specific levels of the factors was investigated using Tukey's HSD (honest significant difference) method.
The above-described statistical approach was confronted to an alternative analysis, which consisted in working within a full maximum likelihood framework providing one MOI estimate at each date from all 6 replicates. This full maximum likelihood framework is derived from the likelihood function , with profile-likelihood confidence intervals.
The MOI parameter (λ) was first held constant across all plants and leaf levels, and we used likelihood ratio tests to test whether allowing variation in λ across leaf levels (dates) significantly improved the likelihood of the model. We also similarly tested whether we had a plant effect, though we were much less interested by this factor which should be modeled as a random effect (as indicated above). The outcome of both analyses are shown and discussed in the Supporting Online Information (Figure S2).
All statistical procedures were implemented in the statistical software R [51].
As a first exploratory experiment, the plants were inoculated with a VIT1/VIT3 mixture at a 1/1 ratio and sampled twice, at early and later stages of the infection. The proportion of cells infected by both variants was around 30% in leaves collected 17 days post infection (dpi), and reached nearly 100% in upper leaves collected 60 dpi (not shown). This result interestingly suggested that cell co-infection increased with time, but that it could become frequent enough to “saturate” our experimental system when a 1/1 variant ratio was used in the inoculum: when both variants are detected in nearly all cells it becomes impossible to obtain an accurate MOI estimate with our method.
In a second time-course experiment, we thus decided to use a 4/1 ratio for VIT1 and VIT3. At 21, 42, 60 and 84 dpi, fully expanded leaves were collected near the apex of six plants infected in parallel. The results shown in Figure S1A indicate that the relative ratio of VIT1 and VIT3 was indeed close to 4/1 in infected leaves, and remained approximately constant throughout the experiment. Most interestingly, the average proportion of cells infected by both variants dramatically increased in successive sampling times but remained below saturation, suggesting both that the 4/1 ratio was appropriate and that important changes in the MOI may occur during the invasion of the host. The calculated average MOI values showed a dynamic pattern, starting at lows around 1, sharply increasing up to 13 and then decreasing late in infection (Figure S1B). Unfortunately, important variation between the six replicated plants at each sampling date resulted in too wide confidence intervals, and the statistical analysis failed to confirm the significance of the observed bell-shaped pattern (the full data set is provided in Table S1).
In order to reduce to a minimum the variations between repeated plants, we very precisely adjusted the leaf-sampling protocol during time-course experiments. The development of every new leaf was periodically scrutinized in six plants infected in parallel, to record the dates of their first appearance in the center of the rosette, and to later estimate their respective age at the sampling time. Leaves were numbered so that the first true leaf (above cotyledons) was leaf level 1. The mixture of CaMV VIT1/VIT3 purified virions (ratio 4/1) was inoculated to leaf levels 3 and 4, and the first leaf level showing systemic symptoms homogeneously distributed all over its surface was leaf level 6. The induction of flowering was generally observed around 40 dpi, when leaf 30 appeared. Senescence of individual leaves started when they were approximately 35 days old, whatever the leaf level considered. At each of five time points, one identical leaf level was sampled in the six replicated plants. Selected leaf levels corresponding to the five time points were levels 6, 12, 21, 33 and 43. All leaves were sampled at the same age (13 days after their apparition on the plant) to improve comparison among leaf levels. At this age, all cells within the leaf were likely infected as indicated by the high proportion of CaMV-positive cells found during PCR analysis of individual cells (Table S1 and Table S2 in Supporting online Information). Moreover, 13 days old leaves had already gone through the physiological sink-to-source transition that stops import of photo-assimilates and viruses from the phloem [52].
Finally, to limit interference of the sampling process with plant development and systemic infection, several evenly distributed leaf discs (0.8 cm Ø), amounting solely 20% of the total leaf surface, were collected from each leaf. Protoplasts were extracted from each sampled leaf as previously described [18], [53].
Four treatments were compared for their putative impact on MOI values. To limit potential sources of variation, the experiments were carried out in parallel with the previous experiment on the MOI dynamics and with the same batch of plantlets and inoculum. In all treatments, 6 turnip plants were co-inoculated with VIT1 and VIT3 and the leaves were sampled when they were 28 days old. Sampling was performed exactly as described above except that, for practical reasons, only two leaf levels were sampled (leaf levels 12 and 33). We reasoned that limiting this experiement to two sampling points could provide enough resolution to address the question of a possible MOI difference in different experimental conditions. In three treatments plants were kept in the same growth chamber as for the experiment shown in Figure 1 and 2. The first treatment corresponded to a mechanical inoculation exactly as above, the second to the mechanical inoculation with a 4X dose, and the third to a more natural inoculation by aphid vectors. For the latter, 20 individuals of the aphid Myzus persicae (Sulz.) were fed on a plant co-infected by the two viral variants and then released on the fourth leaf of healthy plantlets as previously described [54]. Finally, in the fourth treatment plants were mechanically inoculated with a 1X dose but maintained in a greenhouse where they were exposed to approximately 16 hours sunlight and higher temperatures. Under these conditions the rate of leaf appearance was nearly identical to that in the growth chamber, but total biomass was multiplied by three and flowering started approximately one week earlier (not shown).
|
10.1371/journal.pbio.1001921 | Carbonic Anhydrase Generates CO2 and H+ That Drive Spider Silk Formation Via Opposite Effects on the Terminal Domains | Spider silk fibers are produced from soluble proteins (spidroins) under ambient conditions in a complex but poorly understood process. Spidroins are highly repetitive in sequence but capped by nonrepetitive N- and C-terminal domains (NT and CT) that are suggested to regulate fiber conversion in similar manners. By using ion selective microelectrodes we found that the pH gradient in the silk gland is much broader than previously known. Surprisingly, the terminal domains respond in opposite ways when pH is decreased from 7 to 5: Urea denaturation and temperature stability assays show that NT dimers get significantly stabilized and then lock the spidroins into multimers, whereas CT on the other hand is destabilized and unfolds into ThT-positive β-sheet amyloid fibrils, which can trigger fiber formation. There is a high carbon dioxide pressure (pCO2) in distal parts of the gland, and a CO2 analogue interacts with buried regions in CT as determined by nuclear magnetic resonance (NMR) spectroscopy. Activity staining of histological sections and inhibition experiments reveal that the pH gradient is created by carbonic anhydrase. Carbonic anhydrase activity emerges in the same region of the gland as the opposite effects on NT and CT stability occur. These synchronous events suggest a novel CO2 and proton-dependent lock and trigger mechanism of spider silk formation.
| The spinning process of spider silk is crucial for making webs or other complex constructions to catch spider's prey. The main components of the silk are spidroins, which are large and repetitive proteins that have conserved nonrepetitive terminal domains (NT and CT). Spiders manage both to store the highly aggregation-prone spidroins in solution at extreme concentrations in the silk glands and then to rapidly convert these spidroins into a solid fiber within fractions of a second as they spin fibres. This process has been extensively studied and is thought to involve a pH gradient, but how this pH gradient is generated and maintained was not resolved. Here, we measured the pH at locations along the ampullate gland and determined that the pH decreases to 5.7 in the middle of the spinning duct. We also observed that the carbon dioxide pressure is simultaneously increased and that its accumulation may affect the stability of CT. We find that active carbonic anhydrase (CA) is crucial to maintain the pH gradient along the gland. Detailed molecular studies of NT and CT under the disparate conditions present along the gland revealed a lock and trigger mechanism whereby in more neutral pH conditions, precocious spidroin aggregation is prevented, and when in more acidic pH conditions, NT dimers firmly interconnect the spidroins and the CT unfolds into β-sheet nuclei that can trigger rapid polymerization of the spidroins. We conclude that this mechanism enables temporal and spatial control of silk formation and may be harnessed in attempts to produce artificial silk replicas.
| Spider silk fibers contain regions of crystalline and noncrystalline β-sheets, which mediate mechanical stability [1]. In contrast, the soluble spidroins (dope) stored in the tail and sac of major and minor ampullate silk glands [2] exhibit unordered and helical conformations [3]. How spiders rapidly convert the dope into a solid fiber at a defined site of the S-shaped duct has been extensively studied [4]–[8], but major questions are unresolved: First, how is the pH gradient in the gland generated and maintained? Second, what is the pH at the phase transition in the duct? The pH in the major ampullate gland has been shown to decrease from 7.2 in the proximal parts of the sac to 6.3 in the beginning of the duct [7], but it has also been proposed that the gradient goes from 6.9 in the sac to 6.3 in the third limb of the duct [6]. Third, how are the terminal domains affected by the conditions in the duct at a molecular level, and in particular, do they, as proposed [4],[5], act in similar manners? Documented pH-dependent effects at a molecular level include that the N-terminal domain (NT) dimerizes at pH 6 [9]–[11], but pH-induced structural changes of the C-terminal domain (CT) have only been observed at pH 2 [4]. Here we address these questions and unravel novel physiological mechanisms for regulated spider silk formation.
By use of concentric ion selective microelectrodes (ISMs) [12] we determined the pH in the major ampullate gland of Nephila clavipes, from the proximal part of the tail to the middle part of the second limb of the duct. Concentrations of CO32− were also determined at locations where pH was high enough to allow reliable measurements, and used to calculate HCO3− concentrations. We found that the pH decreases from 7.6±0.1 (n = 11) in the proximal tail to 5.7±0.0 (n = 6) in the second limb of the duct and that HCO3− concentration increases from 5 mM in the proximal tail to 21 mM in the distal part of the sac (Figure 1 and Table 1). With these values in the Henderson–Hasselbalch equation, the carbon dioxide pressure (pCO2) could be calculated and was found to increase along the gland (Figure 1). We observed that the intraluminal pH at different locations did not change despite superfusion of the gland with an elevated pCO2. This indicates that the epithelium of the major ampullate gland does not allow permeation of CO2, a phenomenon previously described for parietal and chief cells in gastric glands [13]. The concentrations of K+, Na+, and Cl− in the sac were determined to be 6, 192, and 164 mM, respectively, using concentric ISMs (Table 1).
The observation of simultaneously decreasing pH and increasing HCO3− and CO2 concentrations from the proximal to the distal parts of the gland (Figure 1) suggested that carbonic anhydrase (CA) could be involved through catalysing the conversion of H2O + CO2 ↔H++ HCO3−. By use of a histochemical method [14] we could indeed identify abundant CA activity in intracellular vesicles and at the apical cell membrane of the epithelium in the distal part of the major and minor ampullate sacs and ducts, as well as in aggregate gland ducts and tubuliform glands (Figure 2A–E). The site in the major ampullate epithelium where CA was found to emerge (Figure 2A) exactly coincides with the location where the glandular epithelium ceases to produce spidroins [15]. To investigate whether CA is responsible for generating and maintaining the pH gradient, we immersed freshly dissected N. clavipes major ampullate glands in buffers containing methazolamide, a membrane-permeable CA inhibitor [16]. Exposure to methazolamide collapsed the pH gradient, and pH levelled out to approximately 7 in the tail and sac. The gradient could subsequently be restored by removing the methazolamide (Table 2). Thus, the pH gradient in the major ampullate gland is dependent on active CA. Because CA activity was found in the epithelium of the distal major ampullate duct (Figure 2E), where also proton pumps are present [17], the pH may well continue to drop along the entire duct—that is, below pH 5.7 now measured half-way through the duct. This needs to be experimentally verified, as the extremely small inner diameter in the second half of the duct (<20 µm) did not allow measurements with the currently used ISMs.
To address the third unresolved question—that is, how the terminal domains are affected by the conditions in the duct at a molecular level—we first compared the in vitro structural stability of NT and CT in the broad pH gradient now observed. We studied isolated domains, and it may be that these domains behave differently in their natural context of full-length spidroins. However, we have observed that NT followed by five repeats behaves as the isolated domain in terms of pH-dependent dimerization [11]. Urea and temperature denaturation studies at different pH values were performed for recombinant NT and CT (Figures 3 and 4). The stability of NT towards urea remained largely unchanged between pH 7.5 and 6.5, but was significantly increased between 6.0 and 5.0 (Figure 3). We here analyzed a minor ampullate spidroin (MiSp) NT, which has not been studied before, but a similar pH effect was recently shown for a major ampullate spidroin (MaSp) NT [11]. This indicates that the structural effects now observed are applicable to spidroins from major and minor ampullate glands, in concordance with the observation of CA in major and minor ampullate, aggregate, and tubuliform glands (Figure 2). A similar effect as seen for stability towards urea was seen for NT thermal stability; that is, it was increased at lower pH (Figure 4). Dimerization of NT is completed at pH 6 [11], and the subsequent stabilization of NT dimers between pH 6 and 5 (Figure 3) may result in the firm locking of spidroins into multimers in the distal part of the duct (cf., Figure 1). CT, in sharp contrast to NT, was gradually destabilized towards urea (Figure 3) and temperature (Figure 4) when pH was lowered from 7.5 to 5.0. Heteronuclear single quantum coherence (HSQC) nuclear magnetic resonance (NMR) spectra of CT showed a folded structure at pH 6.8, whereas a gradual conversion to an unfolded state was observed at a pH below 5.5, and at pH 5.0, it is completely unfolded (Figure 5). Moreover, we observed that CT irreversibly converted from α-helical to β-sheet structure upon thermal denaturation at pH 5.5, but not at pH 6.5 or 7.5 (Figure 6 and Table S1). The fact that NMR spectroscopy of CT shows an unfolded state at pH 5.0 (Figure 5) whereas circular dichroism (CD) spectroscopy and urea denaturation shows residual structure at pH 5.0 (Figure 3) may be explained by the different CT concentrations (0.3 mM versus 5 µM) and recording times (hours versus minutes) used. It should also be pointed out that unfolded species should have increased NMR intensities (and may thus be overestimated relative to folded species) due to favorable relaxation and dynamic properties and that helical structure (which is observed by CD) may be present in the species that are observed as random coil/unfolded by NMR. Denaturation of NT, in contrast, resulted in mainly unordered structure and was reversible at all three pH values (Figure 6 and Table S1). It may be worth noting that the structural conversion now observed for CT, but not for NT, resembles that seen for the spidroin dope [18], which may be relevant for the trigger mechanism as discussed below.
Next, we used hydrogen-deuterium exchange mass spectrometry (HDX-MS) to study the backbone conformational dynamics of CT at pH 7.5 to 5.5. No major differences in HDX were seen between pH 7.5 and 6.5, but helices 2, 3, and 5 showed increased HDX at pH 5.5 compared to at pH 6.5 (Figure 7), indicating increased structural flexibility at lower pH. Previous studies of CT [4],[19] have identified a strictly conserved salt bridge between an Arg residue in helix 2 and a Glu residue in helix 4. The NMR structure of Araneus ventricosus MiSp CT now studied (Figure 8 and Table S2) is very similar to those of MiSp CT from Nephila antipodiana [19] and MaSp CT from A. diadematus [4] with backbone root-mean-square deviations (RMSDs) of 2.4 Å and 3.4 Å, respectively (over 202 residues from both chains; see Figure 8). Largest differences are observed for the N-terminal helix, which is shorter, and the C-terminal helix, which is kinked near the C-terminus in the A. ventricosus MiSp CT structure. A salt bridge between Arg38 in H2 and Glu82 in H4 is indeed found in A. ventricosus MiSp CT (Figure 8). Computational pKa predictions [20] of the available CT structures uniformly suggested that the Glu residue in H2 (that participates in the saltbridge) has a pKa ≥6, making it possible to protonate in the pH interval now observed in the gland, and mutations interfering with this salt bridge greatly destabilize CT [4],[19]. Our results suggest that protonation of the conserved Glu in H2 is involved in pH-dependent unfolding of CT in spider silk glands, and further experimental studies are warranted to determine exactly what residues are protonated in CT at low pH. Although the NMR structures of several CTs from different spidroins have been solved and their biochemical properties have been studied, the now observed pH responsive behavior of this domain has not been investigated in detail before [4],[5],[19],[21]–[23]. The shared overall fold suggests a conserved function of CT, but the possibility that CT has diverse functions in different silks cannot be excluded and is an important topic for further studies.
The conditions now determined for the distal parts of the gland—that is, low pH combined with increasing HCO3− concentration and low CO2 permeability of the gland—imply that pCO2 is elevated along the sac and duct. For MaSp CT, it has been shown that shear forces induce conformational changes that result in increased exposure of nonpolar surfaces [4], and CO2 interacts mainly with nonpolar regions in proteins [24],[25]. Therefore, we used the CO2 analogue CS2 [24] to identify potential interaction sites in the NMR structure of A. ventricosus MiSp CT. CS2 interacts specifically with a few, mainly hydrophobic, CT residues distributed in helices 2–4, of which many are partly buried (Figure 9A–D). NT on the other hand shows weak interactions with CS2 and only at conditions that favor the monomeric form, at pH 7.2 and 200 mM salt (Figure 10), which is characteristic to parts of the gland where pCO2 is low (Figure 1). In contrast to CT, no specific interactions between NT and CS2 were found at pH 5.5 (Figure 10), suggesting that NT stabilization at low pH (Figure 3 and Figure 4) protects its hydrophobic, buried residues from interacting with CO2.
Amyloid fibrils are β-sheet polymers formed from (partly) unfolded proteins in a nucleation-dependent reaction and are found in tissue deposits associated with disease but also in some functional protein aggregates [26]. Amyloid fibrils share similarities with the β-sheets of spider silk and have been observed in the distal third of the spinning duct by electron microscopy (EM), and it was proposed that the spidroin repetitive parts are responsible for the amyloidogenic behavior [27]. The poly-Ala segments of spidroins need to rapidly form β-sheet structure in silk formation, although Ala is highly prone to form α-helices [28], raising the question, What nucleates this process? We investigated whether CT may convert to amyloid-like fibrils at low pH by measuring Thioflavin T (ThT) fluorescence of CT over time at different pH values. When ThT binds to β-sheet polymers in amyloid-like fibrils, it gives an increased fluorescence [29]. At pH 5.5 and below, CT converted to a ThT-positive state, which was not observed at higher pH, or for NT at any pH tested (Figure 11A). Analysis of the ThT-positive aggregates by transmission EM showed typical amyloid-like fibrils, 5–10 nm thick, elongated and nonbranched (Figure 11B). Only samples of CT incubated at pH 5.5 showed the presence of amyloid-like fibrils. Furthermore, the CT fibrils were positive for Congo red staining and showed green birefringence under polarized light (Figure 11C), another hallmark of an amyloid-like fiber [30].
The spidroins' terminal domains are highly conserved, both between species and between different types of silks [31], which suggest that they play important roles in spider silk formation rather than for the silks' mechanical properties. Further supporting the hypothesis of general polymerization mechanisms between different types of silks, CA is found in the distal parts of several different spider silk glands and occur at the same location as the observed structural changes of NT and CT will take place provided that their behavior in vitro is recapitulated in vivo. NT and CT are unique to spidroins and there are no known homologues. The lock (accomplished by NT) and trigger (accomplished by CT) mechanism proposed herein is therefore likely unique for spider silk formation, in contrast to the previously identified shear-induced polymerization mechanism that also apply to, for example, silk worm silk formation [32]. A detailed understanding of the natural spinning process will be vital for the development of a spinning process capable of generating truly biomimetic spider silk fibers and may provide novel insights into Nature's way of confining amyloid fibril formation to a specific location.
In summary, the spidroin N- and C-terminal domains show synchronous and opposite structural changes in response to the physiological conditions of the spinning duct. CT unfolds into β-sheet nuclei that can trigger rapid polymerization of the spidroins, whereas gradually locked NT dimers alleviate the need for rapid diffusion [11],[33], firmly interconnect the spidroins, and allow for propagation of pulling forces along the peptide chains. These events are driven by CO2 and proton gradients that ensure temporal and spatial confinement of the divergent structural changes of CT and NT. This novel lock and trigger mechanism elegantly explains how silk formation can occur at a very high speed, more than 1 m/s [34], and at the same time be confined to the very distal part of the spinning duct.
Concentric ISMs [12] were used to measure the concentrations of hydrogen, carbonate, sodium, potassium, and chloride ions. Thin-walled borosilicate glass capillaries of two different diameters were used for construction of concentric ISMs. The capillary forming the outer barrel (outer diameter 2.0 mm, inner diameter 1.5 mm, A-M Systems 6185) was pulled to a tip diameter of 2–4 µm using a Flaming/Brown micropipette puller (Sutter Instrument Co. US, Model P87). The tip of the outer barrel was silanized by back-filling with N,N-dimethyltrimethylsilylamine (Fluka 41716), after which the barrel was mounted on a micromanipulator and heated using a hot air gun giving temperatures of 200–300°C for 60 s. Ion-selective cocktails for H+ (Fluka 95291), CO32− (described by Chesler et al.) [35], Na+, K+, and Cl− were sucked into the tip to form a 100 to 200 µm long column, and a backfilling solution (pH electrode, 150 mM NaCl pH 7.4; CO32− electrode, 10 mM NaHCO3, 150 mM NaCl) was added in the middle of the outer barrel. The inner barrel (outer diameter of 1.2 mm and inner diameter of 0.9 mm, A-M Systems 6160) was pulled to a tip diameter of 1 µm and filled with 3 M KCl pH 7.4. The inner barrel was then inserted into and secured in the outer barrel, the inner glass tip being positioned 4–10 µm away from the outer barrel tip. A silver wire was inserted into the inner barrel and connected to an amplifier.
The ISMs were calibrated using pH 6.87 and pH 7.42 buffers, 50, 100, 200, and 400 mM Na+ or Cl−, or using 1, 2, 4, and 8 mM K+, respectively. Carbonate electrodes were calibrated as described [35].
Adult female N. clavipes collected in Florida from September to November were kept in individual containers and fed water. Spiders were anaesthetized with CO2 gas before severing at the pedicle. Dissection of the major ampullate glands was carried out in a modified spider Ringer [36] (with 2 mM MgCl2, 2 mM CaCl2, 3 mM KCl, and 10 mM glucose) buffered with 26 mM bicarbonate and 5% CO2, yielding a pH of 7.4.
Major ampullate glands were mounted in a submersion-style incubation chamber and superfused with HCO3− and CO2-buffered modified spider Ringer at room temperature. ISM measurements were performed in triplicates in different parts of the gland. The difference in potential between the bath and the inside of the gland was recorded on a chart recorder (Zipp and Konnen) and later translated into change in concentration of the ion of interest using the Nernst equation (H+, Na+, K+, Cl−) or a modified Henderson–Hasselbalch equation [35] (CO32−) to get the concentration of HCO3−. Determined pH values and HCO3− concentrations were used to calculate pCO2 according to the Henderson–Hasselbalch equation, assuming equilibrium.
To study the influence of CA activity on the pH gradient, some glands were incubated for 1 h in 0.1 mM methazolamide (M4156, Fluka), a membrane-permeable CA inhibitor, prior to pH measurements, after which the methazolamide was washed away for 30 min and pH measurements repeated.
Some glands were subjected to CO2 permeability studies. Glands were dissected, mounted, and superfused with HCO3− and CO2-buffered spider Ringer at room temperature as described above. A pH electrode was inserted into the gland, after which the surrounding Ringer solution was buffered by 26 mM bicarbonate and 100% CO2. pH measurements were continued up to 1 h to see if intraluminal pH changed in response to the elevated pCO2 surrounding the gland. The Ringer solution was then changed again, being buffered by 26 mM bicarbonate and 5% CO2, yielding a pH of 7.4, after which the pH electrode was removed from the gland and put in the Ringer and pH was recorded. This was made to ensure that the electrode had not been drifting.
Spiders (A. diadematus, N. clavipes, E. australis, and Tegenaria sp.) were anesthetized and sacrificed as described above. Dissection was carried out in 67 mM sodium phosphate buffer at pH 7.2 or in a modified Spider Ringer (see above). Some opisthosomas were fixed and embedded directly after removal of the exoskeleton, whereas others were dissected so that the major and minor ampullate glands could be isolated before fixation.
Tissues for histochemical localization of CA activity were immersion fixed in 2.5% (v/v) glutaraldehyde in 67 mM phosphate buffer, pH 7.2, for 24 h at 4°C and subsequently rinsed with phosphate buffer, pH 7.2. After fixation, tissues were dehydrated using increasing concentrations of ethanol, infiltrated and embedded in a water-soluble glycol methacrylate (Leica Historesin embedding kit).
Historesin embedded major and minor ampullate glands and opisthosomas were sectioned at 2 µm in a microtome (Leica RM 2165) and stained for CA activity using a histochemical method [14]. The method involves incubation of sections in a medium containing NaHCO3, CoSO4, H2SO4, and KH2PO4, whereby carbon dioxide leaves, pH increases, and a cobalt–phosphate–carbonate complex is formed at sites with CA activity. This complex is then converted into a black cobalt–sulphide precipitate. The sections were counterstained with Azure blue. For control of unspecific staining, the CA inhibitor acetazolamide was included in the incubation medium.
A. ventricosus MiSp NT and CT coding gene fragments corresponding to (NT: GSGNSQPIWT NPNAAMTMTN NLVQCASRSG VLTADQMDDM GMMADSVNSQ MQKMGPNPPQ HRLRAMNTAM AAEVAEVVAT SPPQSYSAVL NTIGACLRES MMQATGSVDN AFTNEVMQLV KMLSADSANE VST) and (CT: GSGNSTVAAY GGAGGVATSS SSATASGSRI VTSGGYGYGT SAAAGAGVAA GSYAGAVNRL SSAEAASRVS SNIAAIASGG ASALPSVISN IYSGVVASGV SSNEALIQAL LELLSALVHV LSSASIGNVS SVGVDSTLNV VQDSVGQYVG) were amplified by PCR with the full-length MiSp gene as template [37], cloned into a modified pET vector (resulting in the target proteins being fused to His tag–Thioredoxin–His tag followed by a thrombin cleavage site) and transformed into BL21 (DE3) Escherichia coli. The E. coli were grown at 37°C in LB medium containing 70 mg/l kanamycin until OD600 was about 0.9. The temperature was lowered to 30°C, IPTG was added to a final concentration of 0.3 mM, and the cells were incubated for about 4 h. The E. coli were then harvested by centrifugation at 6,400×g for 20 min at 4°C (Sorvall RC 3BP+, 500 ml flasks), after which the pellet was resuspended in 20 mM Tris pH 8.0, 1 mg/ml lysozyme was added, and the solution was incubated on ice for 30 min. Next, DNase and MgCl2 were added and the mixture was kept on ice for 30 min. The cell lysate was centrifuged (27,000×g) at 4°C for 20 min (centrifuged as above, 50 ml tubes). For purification of CT, the supernatant was loaded on a Ni-NTA column and the fusion protein was eluted with 300 mM imidazole. For purification of NT, which is mainly found in the pellet after lysis, pellets were resuspended in 20 mM Tris pH 8.0 containing 2 M urea, sonicated for 2 min, and the supernatant was treated as for CT. The fusion proteins were then dialyzed against 20 mM Tris pH 8.0 overnight at 4°C, cleaved by 1/1,000 (w/w) thrombin, and run over a Ni-NTA column to remove the fusion tag. This resulted in essentially pure NT or CT (>90% purity as determined by SDS PAGE gel electrophoresis and Coomassie staining).
For NMR structure determination, we initially expressed a 150-amino-acid-residue-long C-terminal part of A. ventricosus MiSp (full-length sequence above). The expressed protein was labeled with 15N, and the NMR spectrum showed that the first 25 residues adopt random coil fold. Therefore, A. ventricosus MiSp CT was truncated and residues 31–150 (marked in bold in the sequence above) were expressed in minimal medium and labeled by 15N and 13C/15N. The NMR sample was prepared by adding 8% (v/v) D2O and 0.02% (w/v) NaN3 to a 1 mM solution of uniformly 13C/15N-labelled protein in 20 mM sodium phosphate buffer (pH 6.8) with 20 mM NaCl.
All NMR experiments were carried out at 298 K on a 600-MHz Varian Unity Inova spectrometer equipped with an HCN triple-resonance pulsed-field-gradient cold probe. The following 2D and 3D spectra were acquired for backbone resonance assignment (number of complex points given in parentheses): [15N-1H]-HSQC (1024×128), HNCA (1024×24×40), CBCA(CO)NH (2048×48×40), HNCO (1024×24×40), HN(CA)CO (1024×24×40), and for side-chain assignment and NOE restraint collection (mixing time given in parentheses): 15N-resolved NOESY-HSQC (1024×38×150, 60 ms), 13C(aliphatic)-resolved NOESY-HSQC (768×52×150, 60 ms), and 13C(aromatic)-resolved NOESY-HSQC (1024×16×150, 60 ms). Additionally, in order to identify intermolecular NOEs, a 13C/15N-filtered 13C(aliphatic)-resolved NOESY-HSQC spectrum (768×34×70, 60 ms) was recorded on a sample containing 50% 13C/15N-labelled and 50% unlabelled proteins [38] that was prepared by mixing equal amounts of labeled and unlabelled proteins in 8 M urea followed by dialysis against the NMR sample buffer. The same sample was afterwards used to probe interactions with CS2. Aliquots of 20% CS2 in DMSO were added in a stepwise manner to the NMR sample of CT, yielding CS2 concentrations of 50 mM, 100 mM, and 200 mM, and a 2D [15N-1H]-HSQC spectrum was recorded each time. To account for perturbations due to DMSO, a reference experiment was performed by adding DMSO only in the same amounts. CS2-induced chemical shift perturbations were calculated by comparing the spectrum at 200 mM CS2 with the spectrum at the end of the reference titration with DMSO, and using the formula () [39]. All spectra were processed with Bruker TopSpin 3.1 and analyzed using CARA [40]. The assigned chemical shifts have been deposited in BioMagResBank (accession number 19579). To probe interactions between NT and CS2, NT from MaSp1 from E. australis was expressed and purified as previously described [10]. Chemical shift perturbations of MaSp NT backbone amides were determined at pH 7.2 and 200 mM NaCl and at pH 5.5 upon addition of CS2 (0 to 200 mM) as described for CT.
For 2D [15N-1H]-HSQC NMR spectra of MiSp CT, samples at pH 5.0, 5.3, and 5.5 were prepared by diluting 50 µl of a concentrated stock solution of MiSp CT in 20 mM sodium phosphate buffer, 20 mM NaCl, 0.03% NaN3, pH 6.8 with 200 µl of 100 mM CD3COOD/CD3COONa, 20 mM NaCl, 0.03% NaN3 buffer, and adding 20 µl of D2O.
Automated peak picking of the three NOESY spectra was performed using UNIO-ATNOS/CANDID 2.0.2 [41]. Distance restraints were obtained from these peak lists using the internal NOE calibration procedure of CYANA 2.1 [42]. Intermolecular contacts were identified by analysis of the 13C,15N-filtered NOESY spectrum and used as distance restraints with an upper limit of 5 Å. No explicit torsion-angle restraints were used in the input. Structure calculations were performed using CYANA 2.1 [42] and involved seven iterations of automated NOE assignment with the routine CANDID [41] followed by a simulated annealing procedure starting in the first cycle from a homology model generated based on the MiSp CT structure from N. antipodiana [19] (PDB accession code 2M0M) that was annealed in 15,000 steps of torsion-angle dynamics. This approach was used to reduce the assignment ambiguity during the first cycles of the automated NOE assignment and resulted in significantly more unambiguous distance restraints in the final cycle of the calculation concomitantly with a lower target function value. The 20 conformers with the lowest residual restraint violations were energy minimized in a water shell using the program CNS 1.2 [43],[44], and their coordinates were deposited in PDB (accession code 2MFZ). Table S2 shows an overview of the restraints used and structural statistics. Ramachandran statistics for structured part (residues 20–120) are 94.2% most favored, 5.8% additionally allowed regions; for all residues including the unstructured N-terminal tail, 88.3% most favored, 11.1% additionally allowed, 0.3% generously allowed, and 0.3% disallowed regions.
For analysis of amyloid fibril formation, 10 µM of A. ventricosus MiSp NT and CT were incubated under quiescent conditions at 25°C with 10 µM ThT in 20 mM sodium phosphate or 50 mM sodium acetate buffer with or without 154 mM NaCl at different pH values between 5.0 and 7.5. ThT fluorescence was recorded on a BMG FLUOstar Galaxy plate reader using bottom optics in 96-well polyethylene glycol-coated black polystyrene plates with a clear bottom (Corning Glass, 3881) using a 440-nm excitation filter and a 490-nm emission filter.
For analysis of amyloid fibrils, 10 µM of A. ventricosus MiSp NT and CT were incubated overnight (12–16 h) under quiescent conditions at 25°C in 20 mM sodium phosphate buffers with or without 154 mM NaCl at pH 7.5, 6.5, and 5.5, respectively. Samples were incubated overnight and 2 µl were adsorbed on copper grids, stained with 2.5% uranyl acetate in 50% ethanol for about 20 s, and examined and photographed with a Hitachi H7100 electron microscope at 75 kV.
Ten µM A. ventricosus MiSp CT was incubated at 37°C with shaking (250 rpm) for 2.5 h at pH 5–7 in 20 mM sodium phosphate and 50 mM sodium acetate buffers, respectively. Samples were centrifuged, supernatant removed, washed with dH2O, and then centrifuged again. The supernatant was removed and 10 µl dH2O was added, the sample was vortexed, and droplets (0.8 µl) were applied to microscopical slides, air dried, and stained with Congo red B [45]. After mounting under coverslips, the materials were examined in a polarization microscope for Congophilia and green birefringence.
A. ventricosus MiSp NT and CT stability between pH 5.0 and 7.5 with and without 154 mM NaCl was determined by urea denaturation. Like in previous denaturation studies of MiSp CT from N. antipodiana [19], we used a two-state model for analyzing our denaturation data. Although a two-state transition is supported by a CD isodichroic point at 203 nm [46] for NT at low pH (Figure 4A), this is not the case for CT at any pH, or NT at pH 7.5 (Figure 4B). To emphasize that we assumed a two-state transition for both NT and CT, the urea concentrations derived from fitting the data to a two-state unfolding model are referred to as apparent half-denaturation ([den]50%). Notably, the main conclusion from these experiments—that NT and CT respond in completely opposite ways to lowered pH—is not dependent on whether a two-state transition applies or not.
For NT, urea-induced denaturation was performed by diluting the protein to 5 µM in 20 mM HEPES/20 mM MES with 0–7 M urea in 0.25 M steps. Tryptophan fluorescence emission spectra were measured on a spectrofluorometer (Tecan Safire 2) using Costar black polystyrene assay plates with 96 flat bottom wells. The samples were excited at 280 nm using a 5 nm bandwidth, and emission spectra were recorded in 1 nm steps between 300 and 400 nm using a 10 nm bandwidth. Spectra were recorded at constant pH values ranging from 5.0 to 7.5 with 0.2–0.5 unit steps. For CT, CD spectroscopy at 222 nm was used to determine [den]50% as a function of pH. The CT samples were diluted to 7.5 µM in 20 mM sodium phosphate buffer and ran with 0–7 M urea in 0.25 M steps. At each pH, the average 222 nm CD ellipticity from three temperature scans for different urea concentrations were obtained with the settings described below (see CD spectroscopy). The ellipticities for each measured pH values ranging from 5.0 to 7.5 with 0.2–0.5 unit steps were plotted against the urea concentration and fitted to a two-state unfolding model in order to determine the [den]50% by KaleidaGraph.
CD spectra were recorded from 260 to 190 nm at 25°C in 0.1 mm path length quartz cuvettes using an Aviv 410 Spectrometer. The wavelength step was 0.5 nm, averaging time 0.3 s, scan speed 20 nm/min, time constant 100 ms, and bandwidth 1 nm. The spectra shown are subtracted for background and averaged over three consecutive scans. The HT voltages were always below 600 V during the entire scans. Spectra of 7.5 µM A. ventricosus MiSp NT (110 µg/ml) or CT (90 µg/ml) in 20 mM sodium phosphate buffer at pH 7.5, pH 6.5, or pH 5.5 were recorded at 25, 45, 65, 85, and 95°C and at 25°C again after cooling. For temperature melting curves, the CD at 222 nm was monitored between 25 and 95°C with 1°C/min increase.
Deuteration buffers were prepared by freeze-drying 200 µl of 20 mM sodium phosphate buffer, pH 5.5 or 6.5, followed by reconstitution in 200 µl D2O (Cambridge Isotopes). A. ventricosus MiSp CT was diluted from 555 µM stock solution, pH 6.5, to 55.5 µM in deuterated phosphate buffer, pH 6.5 or 5.5. We removed 19.5 µl aliquots after 400 s, 50 min, 100 min, 200 min, or 300 min (pH 5.5) or after 40 s, 5 min, 10 min, 20 min, or 30 min (pH 6.5). Aliquots were placed in prechilled 500 µl Eppendorf tubes containing 0.5 µl 5% TFA (Merck), vortexed, and immediately frozen in liquid nitrogen. For a fully deuterated control, CT was incubated in deuterated phosphate buffer, pH 6.5, for 24 h at 25°C. Samples were stored at −80°C until ESI MS analysis.
Samples were thawed and immediately injected into an HPLC system using a chilled 25 µl Hamilton syringe. CT protein was digested in a Porozyme pepsin cartridge (Applied Biosystems), and peptides were trapped and desalted in a Waters Symmetry C18 trap column (Waters). Two 140D solvent delivery systems (Applied Biosystems) were employed, operating at 20 µl/min (for washing with 0.05% TFA) or at 15 µl/min (for elution with 70% acetonitrile, 0.2% formic acid). Digestion and desalting were carried out in a single step for 10 min, and peptides were then eluted in a single step and delivered to the mass spectrometer via a TaperTip emitter (Proxeon). The entire flow system was submerged in an ice bath.
ESI spectra were acquired in positive-ion mode with a Waters Ultima API mass spectrometer (Waters) equipped with a Z-spray source. The source temperature was 80°C, the capillary voltage was 2.5 kV, and the cone and radiofrequency lens 1 potentials were 100 and 38 V, respectively. The mass spectrometer was operated in single-reflector mode to achieve a resolution of 10,000 (full width at half maximum). The mass scale was calibrated using [Glu1]fibrinopeptide B. Peptic peptides were identified based on a map of pepsin-digested undeuterated CT using automated liquid chromatography–tandem mass spectrometry (LC-MS/MS) analysis with a Waters NanoAcquity system (Waters). Peptide sequences were identified by individual analysis of collision-induced dissociation (CID) spectra using the Waters MassLynx and ProteinLynx software packages (Waters).
We analyzed 100 µl of 1 mg/ml A. ventricosus MiSp CT equilibrated 10 min in 20 mM HEPES/MES pH 7.5 or 5.5 using Sephacryl S-100 (GE Healthcare) in the same buffers and at a flow rate of 0.5 ml/min. Molecular mass standards aprotenin (6.5 kDa), ribonuclease (13.7 kDa), CA (29 kDa), ovalbumin (43 kDa), and conalbumin (75 kDa) were used for calibration.
|
10.1371/journal.pbio.1002024 | A Structural Model of the Genome Packaging Process in a Membrane-Containing Double Stranded DNA Virus | Two crucial steps in the virus life cycle are genome encapsidation to form an infective virion and genome exit to infect the next host cell. In most icosahedral double-stranded (ds) DNA viruses, the viral genome enters and exits the capsid through a unique vertex. Internal membrane-containing viruses possess additional complexity as the genome must be translocated through the viral membrane bilayer. Here, we report the structure of the genome packaging complex with a membrane conduit essential for viral genome encapsidation in the tailless icosahedral membrane-containing bacteriophage PRD1. We utilize single particle electron cryo-microscopy (cryo-EM) and symmetry-free image reconstruction to determine structures of PRD1 virion, procapsid, and packaging deficient mutant particles. At the unique vertex of PRD1, the packaging complex replaces the regular 5-fold structure and crosses the lipid bilayer. These structures reveal that the packaging ATPase P9 and the packaging efficiency factor P6 form a dodecameric portal complex external to the membrane moiety, surrounded by ten major capsid protein P3 trimers. The viral transmembrane density at the special vertex is assigned to be a hexamer of heterodimer of proteins P20 and P22. The hexamer functions as a membrane conduit for the DNA and as a nucleating site for the unique vertex assembly. Our structures show a conformational alteration in the lipid membrane after the P9 and P6 are recruited to the virion. The P8-genome complex is then packaged into the procapsid through the unique vertex while the genome terminal protein P8 functions as a valve that closes the channel once the genome is inside. Comparing mature virion, procapsid, and mutant particle structures led us to propose an assembly pathway for the genome packaging apparatus in the PRD1 virion.
| The life cycle of a virus involves serial coordination of viral molecular machines. These machines facilitate functions such as membrane fusion and genome delivery during infection, and capsid formation and genome packaging during replication and shedding. Icosahedral dsDNA viruses use one genome-translocation machine for both genome delivery and packaging. The genome-translocation machine of the membrane-containing bacterial virus PRD1 is composed of four packaging protein species at a unique vertex. Because these proteins do not follow the dominating icosahedral symmetry of the viral capsid, the structure of this vertex has remained elusive. In this study, we localize the unique vertex in the virus from raw electron cryo-microscopy images of the virus. We show that the genome-packaging complex of PRD1 replaces the regular 5-fold structure at the unique vertex and contains a transmembrane conduit as a genome translocation channel. We extend our structural studies to the procapsid—a precursor of the virus—and three packaging mutant particles, allowing us to localize all individual protein species in the complex. Based on these structures, we propose a model of the molecular mechanism of assembly and packaging in the life cycle of the PRD1 virus.
| The functional and structural knowledge of assembly principles of macromolecular complexes, in general, and viruses, in particular, have extended our understanding of viral capsid maturation and genome packaging processes. The model systems used are most often double-stranded DNA (dsDNA) viruses composed of only proteins and nucleic acids. Viruses with lipids possess additional complexity when exploring the mechanistic and structural properties of such fundamental functions.
The common mechanism for the genome encapsidation in icosahedral dsDNA viruses, including head-tailed phages, herpes, pox, and adenoviruses, involves a translocation of the viral DNA into a preformed procapsid by an ATP-driven reaction powered by the packaging complex localized at a single vertex [1]. This single vertex-portal complex operates in both genome delivery and packaging. A dodecameric connector at a 5-fold vertex provides a conduit for nucleic acid entry into the capsid [2]–[5]. It is also an assembly site for the transiently associated packaging NTPase powering DNA translocation [6]. The DNA packaging complex in tailless icosahedral dsDNA viruses with an internal membrane, such as bacteriophage PRD1, operates in a similar manner, but is driven by a virion associated ATPase [7],[8]. In PRD1, the ATPase P9 powers DNA packaging and has, in addition to the Walker A and B motifs, a conserved motif that may contribute to its anchoring to the membrane [7]. P9 also shares sequence similarity with several other putative viral packaging ATPases, implying that this packaging mechanism might be common among the internal membrane-containing viruses [7]. The only structural evidence for the packaging components of a tailless icosahedral virus with a membrane comes from the crystal structure of the archaeal Sulfolobus icosahedral virus 2 (STIV2) packaging ATPase, which shows that these ATPases belong to the FtsK-HerA superfamily of P-loop ATPases, having both cellular and viral members [9],[10]. However, how the packaging complex is connected to the virion and how it provides a conduit through the internal membrane remain unknown.
The discovery that bacterial virus PRD1 and human adenovirus have the same major capsid protein (MCP) fold and virion architecture led to the hypothesis that viruses infecting host cells belonging to different domains of life are related, even though they do not share any detectable sequence similarity [11],[12]. This finding has led to the structure-based classification of viruses, and accordingly it was also proposed that viruses fall into a relatively small number of structure based viral lineages [13],[14]. One of these lineages is represented by PRD1 and includes several other viruses such as adenovirus, bacteriophage PM2, vaccinia virus, Paramecium bursaria chlorella virus 1 (PBCV-1), archaeal Sulfolobus turreted icosahedral virus (STIV), and virophage Sputnik [15]–[21]. In addition, there are also similar viruses with two MCPs instead of one. The relation of these viruses to the double β-barrel MCP containing viruses has been recently discussed [22],[23]. All these viruses are thought to derive from a common ancestor preceding the separation of the three domains of cellular life [13],[24],[25].
Bacteriophage PRD1 is the best-studied viral system, where the virion possesses an internal membrane (Figure S1). The broad structural information on PRD1, down to atomic resolution, has provided insights into assembly principles of complex viruses [26]–[28]. The mature virion (∼66 MDa) is formed of at least 18 protein species of which ∼ten are membrane associated, constituting about half of the membrane mass [29],[30]. The external pseudo-T = 25 icosahedral capsid shell of PRD1 is composed of 720 copies of the MCP P3 (43.1 kDa) cemented together by 60 copies of minor coat protein P30 (9.1 kDa) (Figure S1A) [26],[31]. The MCP P3 has a canonical double jellyroll fold, which is conserved within the lineage of PRD1-like viruses [11],[26]. The viral membrane, which is selectively acquired from the host plasma membrane, has a higher phosphatidylglycerol/phosphatidylethanolamine (PG/PE) ratio than that of its host [32],[33]. In addition, the lipids in the viral membrane are asymmetrically distributed between the leaflets—PE and PG are enriched in the inner and outer leaflets, respectively, most probably due to the high membrane curvature imposed by the capsid [27],[33].
In PRD1, the regular 5-fold vertex (receptor binding vertex) consists of the membrane anchor protein P16 (12.6 kDa), penton base protein P31 (13.7 kDa), receptor recognition protein P2 (63.7 kDa), and spike protein P5 (34.2 kDa) [26],[31],[34]–[37]. Protein P2 initiates infection by attaching to the host cell receptor [35],[38]. However, unlike head-tailed bacteriophages in which the tail hub is used to penetrate the host cell envelope and provide a channel for genome delivery, PRD1 uses its internal membrane that transforms into a tail tube penetrating the capsid through an opening at the unique vertex and crossing the host cell envelope [38]–[40]. The structural transition of the membrane triggers the release of the other vertex complexes leading to the loss of interaction between the capsid and the underlying membrane and allowing the tube to be formed [39].
Among the 12 icosahedral vertices, PRD1 has one unique vertex responsible for the packaging of its linear 14,297 bp-long dsDNA genome, where the covalently 5′ end linked terminal proteins are necessary for genome packaging as well as for replication (dsDNA-P8 complex; P8 is a 29.6 kDa protein) [8],[41],[42]. The unique vertex consists of transmembrane proteins P20 (4.7 kDa) and P22 (5.5 kDa) as well as proteins P6 (17.6 kDa) and P9 (25.8 kDa), which were identified by genetic analyses and immuno electron microscopy (Figure S1B) [7],[43]–[45]. Previous experiments have shown that there are naturally occurring empty procapsids that lack protein P9 and are incompetent to package the genome [8]. The in vitro packaging system applied to different PRD1 packaging mutants showed that while P9 is the packaging ATPase, the packaging efficiency factor P6 participates in the process, most probably by having a role in the incorporation of P9 into the unique vertex [7],[8],[45]. To date, the unique vertex still remains structurally elusive, mainly due to technical difficulties in identifying non-icosahedral features in a highly symmetrical virus particle for cryo-EM structural determination.
In this study, we report the structure of a viral packaging complex with a membrane conduit using cryo-EM reconstruction without icosahedral symmetry imposition at 12 Å resolution. Using virus particles devoid of specific unique vertex protein species allowed us to define the structure of this DNA translocation conduit and propose an assembly pathway for this portal structure crossing both the protein shell and the underlying viral membrane layer.
Resolving non-icosahedrally organized features that are essential functional components in icosahedral viruses remains a challenge. Using algorithms specific to handling icosahedral objects in the multi-path simulated annealing (MPSA) software package [46],[47], several non-icosahedrally symmetric features in icosahedral viruses have been revealed, such as the tail organization in cyanophage P-SSP7 [47], the portal in herpes simplex virus 1 B-capsid [4], and the portal in enteric phage P22 procapsid [5] and mature virion [2],[48].
In order to reveal the unique vertex in tailless mature PRD1 virion, 26,000 out of 50,000 particles were used to reconstruct the final density map at 12 Å resolution based on gold-standard criterion of two independent datasets [49],[50] without icosahedral symmetry imposition (Figures 1A–1C and S2A; Tables 1, 2, and S2; Movie S1). The map showed a unique packaging complex structure at one of its 12 vertices (Figure 1C and 1D) and regular 5-fold structures in the remaining 11 vertices (Figure 1E). The receptor recognition protein P2 and spike protein P5 were not resolved at the regular 5-fold vertices because of their flexible nature [37]. Except for the unique vertex, the overall virion density map revealed a similar capsid organization as in the X-ray structure of the icosahedral PRD1 capsid [26]. The Fourier shell correlation (FSC) calculated between the crystal structure of the MCP P3 (PDB: 1W8X, chain B) and the virion cryo-EM density map indicated that their structures match to 12 Å based on 0.5 FSC criterion (Figure S2B). This quantitative measure is substantiated by their apparent structural match (Figure 1F) and validates the overall accuracy of the image processing protocol.
Crystal structures of the penton protein P31 and the MCP P3 [26] were docked into the cryo-EM density map (Figures 2A and 2B; Movie S2). The docking shows unambiguously that the unique vertex does not have the pentameric protein P31 and the five neighboring MCP P3 trimers (peripentonal MCPs) as do the regular 5-fold vertices (Figure 2A). At unique vertex position, the packaging complex is surrounded by ten MCP P3 trimers (Figure 2B). The segmented unique packaging vertex comprises not only the capsid region that replaces the regular 5-fold structure, but also the transmembrane region that anchors the inner membrane layer interior to the capsid shell (Figure 2C). To understand the interactions between the unique packaging vertex and the capsid shell, the electrostatic inner surface of the ten P3 trimers surrounding the packaging complex was calculated by APBS [51]. The inner surface of the surrounding MCPs had an overall weak negative charge, leading to a hypothesis that the outer surface of the packaging complex is positively charged to allow a stable interaction with the encompassing capsid shell (Figure 2D).
As we have determined the overall structure of the unique packaging complex in the mature virion, the locations of the four packaging protein candidates remain unassigned in the complex. We thus investigated the structures of the procapsid and three other packaging deficient mutant particles in order to localize the four protein species forming the packaging vertex.
Comparison of the mature virion to the procapsid devoid of packaging ATPase P9 and the viral genome (dsDNA-P8 complex) allowed the initial dissection of different protein components of the packaging vertex.
The procapsid density map without icosahedral symmetry imposition at 14 Å gold-standard resolution (Figures 3A–3C and S3A; Tables 1, 2, and S2; Movie S3) revealed that the organization of the MCP and internal lipid membrane was similar to that of the icosahedral map of the procapsid [28]. We noted a sharper fall-off of the FSC plot at low resolution between the two independent maps of the procapsid (Figure S3A) relative to that observed in the mature virion (Figure S2A), which can be attributable to disordering of the lipid membrane in the procapsid (Table 2). Docking of the crystal structure of MCP P3 into the symmetry-free procapsid density map (Figure S3B) revealed that their structures match. The FSC between the P3 crystal structure and the segmented P3 cryo-EM density shows a structural match to 14 Å based on the 0.5 FSC criterion (Figure S3C). On the basis of the difference map calculated between the procapsid and the mature virion at equivalent resolution (Figure S4A and S4B), the unique vertex of the procapsid displayed densities only in the transmembrane conduit but not at the radii of the capsid shell exterior to the membrane (Figure 3B and 3C). No density was observed on either side of the lipid bilayer confirming that protein P9 is part of the unique vertex. Since P9 is considered to reside at the external surface of the virus [8], we could attribute the missing density facing the exterior part of the virus to P9 (Figure 2C).
To localize the packaging efficiency factor P6 in the unique vertex, we utilized packaging deficient mutant Sus621 particles (amber mutation in gene VI), which are devoid of P6 and in which the amount of P9 is reduced to less than half of the wild-type (wt) amount (Table 1). The density map of the Sus621 particle at 19 Å gold-standard resolution (Figures 3D–3F and S5A; Table S2) revealed the transmembrane densities at the unique vertex similar to those seen in the procapsid (Figure 3B and 3C). The maps of the unique vertices in the procapsid and Sus621 particle lacked any density exterior to the membrane (Figure 3C and 3F). The icosahedrally arranged capsid proteins in the procapsid and Sus621 maps were structurally similar. However, the regular vertex penton densities showed higher structural variance in the Sus621 mutant particle than in the procapsid, as shown in their difference maps both compared against the mature virion map (Figure S4A–S4D). These suggest that the regular vertex pentons in the mutant particle are not as rigid as that of the mature and procapsid particles yielding higher variance in the reconstructed densities.
When examining closely at the transmembrane densities at the unique vertices of the procapsid and Sus621 maps (Figure 3C and 3F) and their difference map at the same resolution (Figure S6A), we found that there were extra densities in the center of the transmembrane densities in the procapsid map but not in the Sus621 map. Since protein P6 is present in the procapsid but not in the Sus621 particle, these additional densities may correspond to the region of the P6 anchored to the center of the transmembrane conduit, while the remaining region of the P6 exterior to the membrane is disordered in the absence of P9. Hydrophobicity cluster analysis of the P6 sequence reinforces the presence of hydrophobic domains within protein P6 (Figure S7A and S7B) [52],[53]. To explain these observations, we propose that the density exterior to the membrane at the unique vertex is a composite of P9 and portion of P6. The non-membrane region of protein P6 is disordered in the procapsid lacking P9, and protein P9 is disordered in the Sus621 particle in the absence of P6. When and only when P9 and P6 are both present, such as the case in the mature virion, they become well-ordered and their corresponding densities can be resolved (Figure 2C).
Furthermore, the rest of the membrane density in the Sus621 particle (Figure 3F) appears to be less pronounced than that of the procapsid and the mature virion (Table 2). This suggests that P6 may exert an impact on the membrane structure rigidity. The low resolution fall-off in the FSC curve of the two independent maps in the Sus621 mutant particle (Figure S5A) also supports this interpretation of the membrane disordering.
In order to translocate the genome across the internal membrane of the virus, a transmembrane conduit has been proposed to exist at the unique vertex providing the channel for genome translocation [39]. Secondary structure element predictions by psipred [53] indicate that proteins P20 (4.7 kDa) and P22 (5.4 kDa) both have one long transmembrane helix and one short one (Figure S7C and S7D), implying that they can potentially form a transmembrane conduit at the unique vertex. To assign protein components to the transmembrane region of the unique vertex, two packaging deficient PRD1 mutants (amber mutation in gene XX or XXII) were exploited (Table 1). They are defective in the synthesis of protein P20 or P22, and form unpackaged particles also lacking proteins P6 and P9 (Sus526 and Sus42 particles) (Table 1). Based on biochemical analyses, it is not clear whether P20 and P22 are both simultaneously absent in the mutant particles [44],[54].
In the cryo-EM images of Sus526 and Sus42 particles (Figure 4A and 4D), the membrane showed increased disorder and was unable to maintain a rigid shape. With a new non-icosahedral symmetry particle orientation search approach (details in Methods), we obtained the reconstructed density maps of the mutant particles determined at 22 Å and 18 Å gold-standard resolutions (Figure S5B and S5C; Table S2). This revealed a void density in the capsid region and a disordered density in the membrane region at the unique vertex (Figure 4B, 4C, 4E, and 4F), confirming that the unique vertex consists of proteins P6, P9, P20, and/or P22. Based on the difference map calculated at the same resolution between the Sus526 particle and the procapsid (Figure S6C and S6D) and that between the Sus42 particle and the procapsid (Figure S6E and S6F), the disordered transmembrane densities at the unique vertices of Sus526 and Sus42 particles do not contain the transmembrane conduit seen in the procapsid. In addition, the rest of the membrane density in Sus526 and Sus42 particles appears to be weaker than that of the mature virion and procapsid (Table 2). These observations suggest that proteins P20 and P22 contribute to the membrane density at the unique vertex and are critical to maintaining the integrity of the membrane. Without the presence of P20 and P22, the membrane exhibits additional flexibility.
Following the localization of the four protein species in the unique vertex, we examined the detailed features of the segmented packaging complex from the mature virion (Figure 5A) and the transmembrane conduit from the procapsid (Figure 5D). Exterior to the membrane region at the unique vertex in the mature PRD1, the density is a composite of P6 and P9 and shows an apparent 12-fold symmetry based on rotational correlation curve (Figures 5C and S8A). It is surrounded by 10 MCP P3 trimers (Figure 2B and 2C). On the basis of the secondary structure prediction of P9 by psipred (Figure S7E) [53], the packaging ATPase P9 has a conserved α/β phage portal motif [55], suggesting that it can form a channel for genome translocation.
The density in the membrane region of the procapsid map displays an apparent 6-fold or 12-fold symmetry arrangement (Figure 5F). The volume of the transmembrane densities (excluding the central extra density that could belong to part of protein P6) is estimated to be around 83.6 nm3, which is equivalent to a molecular mass ∼69 kDa based on the previously established volume to mass equation [56]. Six copies of P20 and six copies of P22 add up to 60.6 kDa, which is reasonably close to the observed density. Based on rotational correlation curve (Figures 5F and S8B), the peaks at 6-fold symmetry arrangement were higher than at 12-fold one, also suggesting that the density is organized as a hexamer. Each of these hexameric components, potentially decorated by surrounding lipids, may represent a heterodimer made of one copy of P20 and P22. The central genome delivery channel, formed by P20 and P22, is estimated to be 40–50 Å wide. The assembly of P20/P22 complex may also provide the nucleating site for the packaging vertex assembly.
Interior to the membrane region, the packaging complex in the mature virion has an additional density, part of which probably corresponds to the terminal protein P8 complex with the dsDNA because this density is seen only in the virion (Figure 5G). A more close-up comparison of the density in the membrane region between the mature and procapsid maps showed some differences (Figure 5H), which may be caused by the membrane bilayer itself undergoing a conformational expansion between the states of procapsid and mature virion [28],[57]. The structural change of the membrane may as well be induced by the addition of P9, P6, and DNA-P8 onto the packaging complex.
Many biological processes involve the utilization of ATP as the fuel source. One exemplary illustration of the extensive roles of ATPases is the encapsidation of viral genomic material into a preformed procapsid shell. PRD1 ATPase P9 provides the energy for the viral genome packaging as shown using an in vitro packaging assay [7],[8]. P9 has a dual role. Functionally, it is a powerhouse to fuel the packaging process by hydrolyzing ATP, and structurally, P9 with the packaging efficiency factor P6 form the portal providing the external part of the channel at the unique vertex for DNA translocation. The internal part of the packaging complex (P20/P22) at the unique vertex is embedded in the membrane, and provides the transmembrane conduit. P20/P22 complex also serves as the nucleating site for the whole specific vertex assembly.
The MCP P3 of PRD1 forms an icosahedral shell with a pseudo-T = 25 lattice [11],[26]. At the regular 5-fold vertex, five P31 proteins organized as a penton with a strict 5-fold symmetry (Figure 2A). In the reconstruction without icosahedral symmetry imposition (Figures 1 and 2), there are 705 (720−5×3) P3 and 55 (60−5) P31 molecules forming the PRD1 capsid shell. The special vertex occupied by several protein components does not obey 5-fold symmetry (Figure 2B). Ten MCP P3 trimers wrap around the 12-fold symmetrical P9/P6 complex at the unique vertex (Figures 2B and 5A–5C). Such a symmetry mismatch is a structural hallmark of the head-tailed dsDNA viruses with a portal complex arranged with 12-fold symmetry [3],[4],[47],[48]. In PRD1 mature virion, the internal membrane follows the shape of the icosahedral capsid shell. This is presumably due to the pressure formed by the packaged genome, the presence of various membrane proteins, and the intercalation of the P3 shell into lipid moieties (Figure S1B). However, at the unique vertex, the proteins connected to the P9/P6 complex are membrane proteins P20 and P22, which are organized with 6-fold symmetry (Figure 5F). A 12-fold versus 6-fold symmetry mismatch among different components at the unique portal vertex, seen here at the interface between P20/P22 and P9/P6, is also found in membrane-less head-tailed dsDNA phages [47].
We propose a molecular model for procapsid assembly and genome packaging (Figure 6), which will serve beyond PRD1 and provide one of the first structural clues in understanding the life cycle of the tailless internal membrane-containing icosahedral dsDNA viruses. As the first step, newly-synthesized viral membrane proteins are transported to the cytoplasmic membrane of the host cell (Figure 6A) [58]. The virus-specific membrane patch is then presumably pinched off, resembling the mechanism of the eukaryotic clathrin-coated pits, providing the framework for procapsid assembly (Figure 6B). The correct folding of certain viral structural proteins (e.g., MCP P3) and the formation of the PRD1 procapsid is facilitated by the host GroEL-GroES chaperonin and virus-encoded scaffolding protein P10 and assembly factor P17 (and most probably P33) [59]–[61]. Interestingly, procapsids devoid of the unique vertex can still assemble, which suggests that the membrane and/or other membrane proteins, for example, the membrane associated non-structural protein P10 [31], are functioning as a scaffold for capsid formation without the packaging complex. However, lacking the P20/P22 membrane pore in the unique vertex leads to disordered internal membrane layers as suggested by the weaker intensities in the membrane region of the map (Figure 4; Table 2). Since P20/P22 membrane pore is one of the defining features of the unique vertex, its absence leads to the formation of non-biologically active particles. In these particles, the specific interactions between the capsid shell proteins and the membrane could be altered, which would result in a weaker density in the map.
In the procapsid, in which P9 is absent, and the Sus621 particle lacking P6 and half of P9, the packaging process is deficient, but the presence of the P20/P22 conduit defines the unique vertex and thus allows the stable interactions between the capsid shell and the underlying membrane, making the internal membrane rigid (Figure 3; Table 2). However, without the internal genome pressure, certain flexibility may exist in the membrane envelop of the procapsid and Sus621 particle. Integral membrane proteins P20 and P22, which tend to form hexameric heterodimers (Figure 5F) with potential lipid decorations, assemble to form a transmembrane conduit (Figure 6C). Then, P9 and the packaging efficiency factor P6 form a 12-fold portal complex with P6 positioned atop the transmembrane conduit. P8 is linked to the 5′ end of the linear dsDNA genome and may recruit the genome to the packaging motor by binding to P9. After this complex is formed, the genome and genome-associated P8 begin to be packaged [45].
Once the packaging efficiency factor P6 and packaging ATPase P9 together become ordered in their position in the unique vertex, DNA packaging can begin. ATP hydrolysis by P9 provides the energy for DNA translocation into the procapsid through the unique vertex (Figure 6D). The conduit across the membrane formed by integral membrane proteins P20 and P22 provides the 40–50 Å wide channel for the dsDNA-P8 complex to be transported through the inner membrane underneath the capsid shell. After packaging, the pore in the vertex must be sealed. The terminal protein P8 may play a role as a protein valve similar to the valve of the head-tailed phage P-SSP7 [47].
After packaging the 14.9 kb dsDNA genome, the increased internal pressure leads to the expansion of the membrane, which dissipates the energy and prevents the massive expansion of the capsid shell. The mature PRD1 virion, as observed in this study, has undergone membrane expansion (Figures 1C and 3C; Table 2). The spacing of the lipid bilayer decreases upon the maturation of the particle and the membrane layer gets closer to the capsid shell as seen in our symmetry-free reconstructions as well as in the previous icosahedral maps (Table 2) [57],[62]. In addition, the internal genome pressure and the closer interaction between the membrane and the capsid shell make the membrane envelope most secure in its relative position and thus result in a stronger density in the reconstructed map (Table 2). These observed changes accommodate the packaging process and eventually lead to the maturation of PRD1 procapsid into infectious virion (Figure 6E).
Biochemical and structural analyses of the unique vertices in the head-tailed dsDNA bacteriophages such as T4, T7, ϕ29, P22, epsilon15, P-SSP7, and some eukaryotic viruses have demonstrated that their packaging and assembly processes share similarity both functionally and structurally. One example is the well-studied bacteriophage P22 [2],[5],[48], where the portal proteins function as the nucleating site for the procapsid assembly with the help of scaffolding proteins. Once the procapsid is formed, the DNA is packaged through the channel of the portal powered by the terminase motor with ATPase activity [63]. Virus maturation involves the release of the scaffolding proteins and terminase [64], before the tail is attached at the unique vertex.
In phage ϕ29 [65], the MCPs, connector/portal protein, head fiber proteins, and packaging RNA (pRNA) molecules together form the prohead with the help of the scaffolding proteins. Then the ATPase motor of ϕ29 packages the DNA into the prohead through the channel provided by the portal proteins and the pRNAs at the unique vertex. After that the tail is attached onto the unique vertex completing the assembly of the virion.
The assembly of PRD1 differs from the head-tailed viruses. First, PRD1 does not possess a conventional portal protein like the portal of phage P22 [66] or the connector in ϕ29 [67]. How is the PRD1 procapsid formed? The portal protein complex of P9 and P6 are assembled to the procapsid, providing the channel for DNA translocation. The ATPase activity of P9 provides the energy for DNA packaging [7] analogous to the terminase in P22 [68] or the ATPase in ϕ29 [69]. However, PRD1 P9 does not dissociate from the capsid after the DNA is packaged as in P22 and ϕ29 systems. Second, the packaging efficiency factor P6 of PRD1 serves as a facilitator for the ATPase motor in genome packaging. In contrast, the DNA-packaging motor of bacteriophage ϕ29 is geared by a ring of pRNAs [70]. Third, our study provides the structural insights into the packaging and assembly process in an icosahedral virus with an internal membrane. In this membranous virus, P20 and P22 form a transmembrane nanotube and provide a nucleating site for the recruitment of P9 and P6. For comparison, in the head-tailed bacteriophages like P22, the portal complex is the initiating site for procapsid assembly [5]. Finally, during maturation of the head-tailed dsDNA bacteriophages, such as HK97 [71] or P22 [5], the viral capsid goes through significant conformational changes including capsid expansion and angularization. In contrast, virus maturation in PRD1 mainly involves the membrane expansion and conformational changes at the MCP-membrane interface as well as in the transmembrane densities at the unique vertex without major conformational changes in the viral capsid [57],[62]. Thus, it is the inner membrane in PRD1 that undergoes most of the significant structural re-arrangements during virus maturation, not the viral capsid shell.
The unique vertex of PRD1 resolved here portrays the detailed structural picture to advance our understanding on procapsid assembly and genome packaging in a membrane-containing virus. The number of different PRD1-like icosahedral internal membrane-containing viruses is increasing: these can infect archaea, bacteria, and eukaryotes, covering all domains of life [72]. This is the first time, to our knowledge, that such a packaging-portal complex structure is revealed. Based on sequence data, all PRD1-like viruses encode a packaging ATPase, including the Walker A and B motifs and the P9-specific region [7] like PRD1 P9. However, even within this group of viruses the packaging mechanisms must differ between those with a circular or linear genome. For viruses with a linear genome, the packaging mechanism resembles that of PRD1, but for circular genomes, like in bacteriophage PM2, the mechanism for the packaging/condensation of the genome could be totally different [16].
Wt PRD1 and its packaging deficient mutants (Table 1) were propagated (LB medium at 37°C) on Salmonella enterica serovar Typhimurium LT2 DS88 (wt non-suppressor host) [73] and on S. enterica suppressor strain PSA (supE) [74] or DB7154 (supD10) [75] harboring plasmid pLM2. The suppressor-sensitive mutant phenotypes were verified by an in vivo complementation assay using plasmids carrying the corresponding PRD1 wt genes (Tables 1 and S1).
To reduce the background in mutant virus productions, the infected cells (multiplicity of infection 8) were collected 15 minutes post infection (Sorvall SLA3000 rotor, 5,000 rpm, 10 min, 22°C) and resuspended in pre-warmed fresh medium. Released virus particles were concentrated and purified by polyethylene glycol-NaCl precipitation, rate zonal (5%–20% gradient; Sorvall rotor AH629, 24,000 rpm, 55 min, 15°C), and equilibrium (20%–70% gradient; Sorvall rotor AH629, 24,000 rpm, 16 h, 15°C) centrifugations in sucrose using 20 mM potassium phosphate (pH 7.2), 1 mM MgCl2 buffer [76]. The equilibrated particles were concentrated by differential centrifugation (Sorvall rotor T647.5, 32,000 rpm, 2 h, 5°C) and resuspended in the same buffer. The protein concentrations were measured by Coomassie blue method using bovine serum albumin as a standard [77]. The wt/revertant backgrounds of the purified mutant particles were analyzed by assaying their specific infectivity on suppressor and wt hosts (Table 1). The protein pattern of the purified particles was analyzed by sodium dodecyl sulfate-polyacrylamide (16% acrylamide) gel electrophoresis (SDS-PAGE) [78].
Aliquots of 2.5–3 µl of purified PRD1 particle suspension (Table 1) were applied to 400 mesh R1.2/1.3 Quantifoil grids (Quantifoil Micro Tools GmbH), blotted for 2 s and immediately frozen in liquid ethane using an automated vitrification device: either a Vitrobot MarkIII (FEI) or a Cryo-Plunger 3 (Gatan). Images were taken with a 300 kV JEM3200FSC electron microscope (JEOL) equipped with in-column energy filter. A slit width of 20 eV was used for data collection. The first dataset of the virion and all procapsid data was recorded at 80 K×nominal magnification (1.42 Å/pixel sampling) with a dose of 20 e/Å2 using a Ultrascan 4000 CCD camera (Gatan) with defocus ranging from 0.5 to ∼2 µm (Table S2). The second dataset of virion was collected using a Ultrascan 10000 CCD camera (Gatan) binned by 2 (1.3 Å/pixel sampling) with a defocus range from 1 to 3 µm. All mutant particles were imaged on a 200 kV JEM2010F electron microscope (JEOL) with a dose of 25 e•Å−2 using a Ultrascan 4000 CCD camera (Gatan) at 40–60 k×nominal magnification sampling from 1.81 to 2.18 Å/pixel and defocus ranging from 1.5 to 3 µm (Table S2).
The virus particle images of PRD1 virion and procapsid were picked automatically with program ETHAN [79] and then manually screened using the EMAN2 program e2boxer.py [80]. Contrast transfer function (CTF) parameters of these particles were adjusted and determined using the EMAN program ctfit with detectable signals to ∼1/6 Å−1 in their 1D power spectra. The MPSA package was used to determine the icosahedral orientation of each particle starting from a random spherical model and only considering information below 10 Å to avoid model bias and over-fitting of the noise [46]. To exclude bad or low quality particles, the consistency of alignment parameters (both the orientation and the center) was used as the selection criterion. MPSA determines five orientation parameters simultaneously using Monte Carlo scheme for icosahedral and symmetry free virus reconstructions. If a raw particle is bad or low quality at a given resolution search range, the program would not yield a stable set of alignment parameters if repeating the orientation search multiple times. We used a strict consistency criterion (orientation difference <0.5°, center difference <3 Å) to compute the best possible icosahedral reconstruction. These criteria of particles selection were published in our earlier paper [46] and have been applied for many applications [4],[5],[47]. Using this algorithm, we filtered about 48% of the particle in our mature virion dataset and ∼20% in the procapsid and other mutants dataset. A possible reason of such a high rejection rate in the mature virion dataset could be the structural plasticity of the samples, which would be a more significant issue when reaching for higher resolution. EMAN make3d program was used to reconstruct the 3D map [81].
The algorithm for breaking the icosahedral symmetry and obtaining the asymmetric particle orientation has been previously described (Figure S9) [47]. Briefly, an initial icosahedral orientation was determined and an icosahedral reconstruction was obtained. Using a very low density threshold, a faint feature at the vertex was segmented out and used as a starting point to determine the asymmetric orientation. Through an iterative process, the features of the unique vertex were improved, which further allowed the more accurate assignment of the genuine asymmetric orientations out of the 60 equivalent possible choices (12 vertex locations×5 possible attachments of the symmetry mismatch at a 5-fold).
The selected particles were split to even and odd half-datasets at the beginning of the refinement (Table S2). Therefore, each half dataset was refined independently starting from separate random spherical models. The resolutions of the FSCs were calculated between two independent reconstructions without any masking for each virus particle dataset. The resolutions of the maps at 0.143 criterion were 12 Å for the wt virion (Figure S2A), 14 Å for the procapsid (Figure S3A), and 19 Å for the Sus621 particle (Figure S5A).
For the Sus526 and Sus42 mutant particles, the same algorithm and approach were attempted but no unique vertex complex was seen in either case. In order to find the missing unique vertex, a new algorithm was developed using the 11 regular 5-fold vertices as a reference for asymmetric search. For this approach, the Sus526 and Sus42 particles were oriented with best matches of regular 5-fold vertices and the missing unique vertex, if there is one, will be seen at the one remaining vertex location. This approach allowed us to successfully identify the orientation of the particles without the unique vertex. The independent FSCs resolution assessments were also done for these two maps, revealing the resolutions to be 22 Å for Sus526 particle and 18 Å for Sus42 particle (Figure S5B and S5C).
In order to compare maps of the virion, the procapsid and packaging mutants at various resolutions, difference maps were calculated between the two maps filtered at the same resolution (Figures S4 and S6). In each pair of comparison, the higher resolution map was filtered to the same resolution to the lower resolution map. The difference map between any two maps at same resolution was computed in Chimera with operation: vop map1 subtract map2. The difference map was displayed with surface color.
The even-odd FSC curves of the procapsid and packaging mutant particles (Figures S3A and S5) showed a moderate drop at the low-resolution region, which was not the case in that of the mature virion (Figure S2A). This observation could be caused by the fact that the membranes in the procapsid and packaging mutant particles are not as rigid as that in the mature virion where the genome pushed the membrane to secure its stable shape. Thus, the less rigid membrane could be one of the reasons for the moderate drop at lower resolution in the FSCs of procapsid and packaging mutant particles. When comparing only the density map of the MCPs in the procapsid to the corresponding X-ray structure (Figure S3C), such drop was not present in the FSC.
|
10.1371/journal.pbio.1000285 | A Predominantly Neolithic Origin for European Paternal Lineages | The relative contributions to modern European populations of Paleolithic hunter-gatherers and Neolithic farmers from the Near East have been intensely debated. Haplogroup R1b1b2 (R-M269) is the commonest European Y-chromosomal lineage, increasing in frequency from east to west, and carried by 110 million European men. Previous studies suggested a Paleolithic origin, but here we show that the geographical distribution of its microsatellite diversity is best explained by spread from a single source in the Near East via Anatolia during the Neolithic. Taken with evidence on the origins of other haplogroups, this indicates that most European Y chromosomes originate in the Neolithic expansion. This reinterpretation makes Europe a prime example of how technological and cultural change is linked with the expansion of a Y-chromosomal lineage, and the contrast of this pattern with that shown by maternally inherited mitochondrial DNA suggests a unique role for males in the transition.
| Arguably the most important cultural transition in the history of modern humans was the development of farming, since it heralded the population growth that culminated in our current massive population size. The genetic diversity of modern populations retains the traces of such past events, and can therefore be studied to illuminate the demographic processes involved in past events. Much debate has focused on the origins of agriculture in Europe some 10,000 years ago, and in particular whether its westerly spread from the Near East was driven by farmers themselves migrating, or by the transmission of ideas and technologies to indigenous hunter-gatherers. This study examines the diversity of the paternally inherited Y chromosome, focusing on the commonest lineage in Europe. The distribution of this lineage, the diversity within it, and estimates of its age all suggest that it spread with farming from the Near East. Taken with evidence on the origins of other lineages, this indicates that most European Y chromosomes descend from Near Eastern farmers. In contrast, most maternal lineages descend from hunter-gatherers, suggesting a reproductive advantage for farming males over indigenous hunter-gatherer males during the cultural transition from hunting-gathering to farming.
| Events underlying the distribution of genetic diversity among modern European populations have been the subject of intense debate since the first genetic data became available [1]. Anatomically modern humans, originating in East Africa, colonized Europe from the Near East ∼40 thousand years ago (KYA), then during the last glacial maximum populations retreated into the peninsulas of Iberia, Italy, and the Balkans, followed by northward recolonization from these refugia ∼14 KYA. The most important cultural transition was the adoption of agriculture originating in the Fertile Crescent in the Near East at the start of the Neolithic, ∼10 KYA [2]. It spread rapidly westwards via Anatolia [3] (Figure 1A), reaching Ireland by 6 KYA, accompanied by the development of sedentary populations and demographic expansion. Debate has focused on whether this spread was due to the movement and expansion of Near-Eastern farmers (demic diffusion), or to the transmission of cultural innovation to existing populations (acculturation), who then themselves expanded.
The observation of southeast–northwest frequency clines for “classical” genetic markers [1],[4], autosomal DNA markers [5],[6], and Y-chromosomal markers [7],[8] (though not for mitochondrial DNA [mtDNA] [9]) has been used to support the demic diffusion model. No dates can be automatically attached to these clines, however, and some [1], detected by principal component analysis, may simply reflect isolation by distance [10]. The direction of movement underlying a cline can also be ambiguous: the high-frequency pole could indicate the area of preexisting substrate least affected by a migration originating far away, or the final destination of a wave of migration into thinly populated territory, where expansion and drift have had their greatest effects [11].
The origins of a frequency cline of a lineage can be illuminated by analysing the diversity within it. For Y-chromosomal lineages defined by binary markers (haplogroups), this can be done using multiple microsatellites. This approach has been applied to haplogroups E, J [12], and I [13] within Europe, but the major western European lineage has not yet been focused upon. The frequency of the major western European lineage, haplogroup (hg) R1b1b2, follows a cline from 12% in Eastern Turkey to 85% in Ireland (Figure 1B), and is currently carried by some 110 million European men. Previous studies of lineages approximately equivalent to hgR1b1b2 [7],[8] suggested that it has a Paleolithic origin, based simply on its high frequency in the west. Here, in contrast, we show that the geographical distribution of diversity within the haplogroup is best explained by its spread from a single source from the Near East via Anatolia during the Neolithic. Taken together with the evidence on the origins of many other European haplogroups, this indicates that the great majority of the Y chromosomes of Europeans have their origins in the Neolithic expansion.
To investigate the origins of hgR1b1b2, we assembled a dataset of 840 chromosomes from this haplogroup with associated nine-locus microsatellite haplotypes (Table 1; Table S1). The diversity of the lineage within each population (measured by mean microsatellite variance) should reflect its age: under a hypothesis of recolonization from southern refugia, we expect a gradient of diversity correlating with latitude, whereas Neolithic expansion from Anatolia predicts a correlation primarily with longitude. Figure 1C shows the geographical distribution of mean microsatellite variance, and Figure 2 shows that although there is no evidence for correlation with latitude (R2 = 0.06; p = 0.268), the correlation with longitude is significant (R2 = 0.358; p = 0.004), with greatest diversity in the east (strongly influenced by highly diverse samples within Turkey), thus providing support for the Neolithic colonization hypothesis.
The two hypotheses also make different predictions for the number of sources of diversity within hgR1b1b2: under the postglacial recolonization model, we expect multiple sources, whereas under the Neolithic expansion model, we expect only one. We can test this by examining the phylogenetic relationships among microsatellite haplotypes. A reduced median network of 859 haplotypes (Figure 3) shows a simple star-like structure indicative of expansion from one source: 74 haplotypes (8.6%) lie in its central node, and this node plus its single-step mutational neighbours together comprise 214 haplotypes (24.9%). Haplotypes belonging to populations from all three refugia are present in the core of the network. This pattern seems incompatible with recolonization from differentiated refugial populations, and in terms of the history of hgR1b1b2, the refugia possess no special status. The core of the network also contains haplotypes from Turkey (Anatolia), which is compatible with a subpopulation from this region acting as a source for the westwards-expanding lineage.
Does the time to the most recent common ancestor (TMRCA) of the hgR1b1b2 chromosomes support a Paleolithic origin? Mean estimates for individual populations vary (Table 2), but the oldest value is in Central Turkey (7,989 y [95% confidence interval (CI): 5,661–11,014]), and the youngest in Cornwall (5,460 y [3,764–7,777]). The mean estimate for the entire dataset is 6,512 y (95% CI: 4,577–9,063 years), with a growth rate of 1.95% (1.02%–3.30%). Thus, we see clear evidence of rapid expansion, which cannot have begun before the Neolithic period.
The similarity between the isochron map of Neolithic sites (Figure 1A; [3]) and those of hgR1b1b2 frequency (Figure 1B) and diversity (Figure 1C) is striking. Further support for the association of the expansion of hgR1b1b2 with that of farming comes from a statistical comparison of the variables. The frequency of hgR1b1b2 at different points in Europe is significantly negatively correlated (R2 = 0.390; p = 0.0005) with the dates of local Neolithic sites (Figure 4A). For the local variance of the microsatellite haplotypes within hgR1b1b2, the correlation with Neolithic dates is significantly positive (R2 = 0.331; p = 0.0124; Figure 4B).
Previous observations of the east–west clinal distribution of the common Western European hgR1b1b2 (or its equivalent) [7],[8] considered it to be part of a Paleolithic substrate into which farmers from the Near East had diffused. Later analyses have also considered variance, and have conformed to the Paleolithic explanation [14],[15]. Here, we concur that the cline results from demic diffusion, but our evidence supports a different interpretation: that R1b1b2 was carried as a rapidly expanding lineage from the Near East via Anatolia to the western fringe of Europe during the Neolithic. Such mutations arising at the front of a wave of expansion have a high probability of surviving and being propagated, and can reach high frequencies far from their source [11]. Successive founder effects at the edge of the expansion wave can lead to a reduction in microsatellite diversity, even as the lineage increases in frequency.
The innovations in the Near East also spread along the southern shore of the Mediterranean, reflected in the expansion of hgE1b1b1b (E-M81) [16], which increases in frequency and reduces in diversity from east to west. In sub-Saharan Africa, hgE1b1a (E-M2) underwent a massive expansion associated with the Bantu expansion [17],[18]. In India, the spread of agriculture has been associated with the introduction of several Y lineages [19], and in Japan, lineages within hgO spread with the Yayoi migration [20], which brought wet rice agriculture to the archipelago. On a more recent timescale, the expansion of the Han culture in China has been linked to demic diffusion [21]. In this context, the apparently low contribution of incoming Y chromosomes to the European Neolithic, despite its antiquity and impact, has appeared anomalous. Our interpretation of the history of hgR1b1b2 now makes Europe a prime example of how expansion of a Y-chromosomal lineage tends to accompany technological and cultural change.
Other lineages also show evidence of European Neolithic expansion, hgE1b1b1 (E-M35) and hgJ, in particular [12]. Indeed, hgI is the only major lineage for which a Paleolithic origin is generally accepted, but it comprises only 18% of European Y chromosomes [13]. The Basques contain only 8%–20% of this lineage, but 75%–87% hgR1b1b2 (Table S1); our findings therefore challenge their traditional “Mesolithic relict” status, and in particular, their use as a proxy for a Paleolithic parental population in admixture modelling of European Y-chromosomal prehistory [22].
Is the predominance of Neolithic-expansion lineages among Y chromosomes reflected in other parts of the genome? Mitochondrial DNA diversity certainly presents a different picture: no east–west cline is discernible, most lineages have a Paleolithic TMRCA [23], and hgH [24] and hgV [25] show signatures of postglacial expansion from the Iberian peninsula. Demic diffusion involves both females and males, but the disparity between mtDNA and Y-chromosomal patterns could arise from an increased and transmitted reproductive success for male farmers compared to indigenous hunter-gatherers, without a corresponding difference between females from the two groups. This would lead to the expansion of incoming Y lineages—as suggested by the high growth rate observed for hgR1b1b2. Similar conclusions have been reached for the Bantu expansion (in which the current Bantu-speaking populations carry many mtDNA lineages originating from hunter-gatherers [26]), the introduction of agriculture to India [19] and the Han expansion [21].
Some studies have found evidence of east–west clines for autosomal loci [6],[27]. By contrast, recent genome-wide SNP typing surveys [28]–[30] find a basic south–north division or gradient, including greater diversity in the south, but they provide no indication of the time-depth of the underlying events, which could in principle involve contributions from the original colonization, postglacial Paleolithic recolonization, Neolithic expansion, and later contact between Africa and southern Europe [31].
The distinction between the geographical patterns of variation of the Y chromosome and those of mtDNA suggest sex-specific factors in patterning European diversity, but the rest of the genome has yet to reveal definitive information. Detailed studies of X-chromosomal and autosomal haplotypes promise to further illuminate the roles of males and females in prehistory.
Males were recruited with informed consent, following ethical approval by the Leicestershire Research Ethics Committee and the ethics committees of the Universities of Ferrara, Pavia, and Exeter and Plymouth.
A total of 2,574 DNA samples from European males, assigned to populations based on two generations of residence, were typed for the SNP M269 [17], defining hgR1b1b2. Following PCR amplification using the primers 5′-CTAAAGATCAGAGTATCTCCCTTTG-3′ and 5′-ATTTCTAGGAGTTCACTGTATTAC-3′, the T to C transition was analysed by digestion with BstNI, which cleaves M269-C-allele chromosomes only. Samples from the Iberian peninsula were typed using the SNaPshot (ABI) procedure [31]. Haplotype data were obtained for up to 20 Y-specific microsatellites [32],[33]. Data from the Ysearch database (http://www.ysearch.org) for Germany (GE) and Ireland (IR) were added, together with published data for Turkey, subdivided into East, West, and Central subpopulations based on published sampling information [14]. To avoid a bias from very large samples of hgR1b1b2 (GE and IR), these were randomly subsampled to give sample sizes of 75. This allowed a comparison of nine-locus haplotypes (DYS19, DYS388, DYS389I, DYS389II, DYS390, DYS391, DYS392, DYS393, and DYS439) for 849 hgR1b1b2 chromosomes, subdivided into 23 populations. Greek and Serbian samples were too small for population-based analyses, but were included in Network analysis.
Neolithic dates, frequencies of hgR1b1b2, and local microsatellite variances were displayed using Surfer 8.02 (Golden Software) by the gridding method. Latitudes and longitudes were based on sampling centres.
Intrahaplogroup diversity was assessed for populations with hgR1b1b2 sample size ≥15 as the mean of the individual microsatellite variances [34], as has been done elsewhere (e.g., [35]); this measure is highly correlated (R2 = 0.871; p = 6.72×10−10) with a more conventional measure, average squared distance (ASD) [36]. Regression analyses were carried out in the R statistical package [37] to compare these two measures, and also to compare mean of variance with latitude and longitude.
A reduced median network [38] of microsatellite haplotypes was constructed using Network 4.5 and Network Publisher, using weighting based on the inverse of the microsatellite variances.
TMRCA and population growth rates were estimated using BATWING [39], under a model of exponential population growth and splitting. Whereas standard use of BATWING assumes a random sample from a population, we validated its use to analyse single haplogroups. Justification of this, together with other details, is given in Text S1.
To assess the correlation between the dates of Neolithic sites and the local hgR1b1b2 frequency and variance, we considered 765 sites and their associated calibrated radiocarbon dates [3]. We identified sites lying within a buffer-zone of 150-km radius around each location for which we had frequency or variance data (Figure 1B and 1C). When more than one site was identified in a given buffer-zone, we considered the mean of the dates. Regression analyses were carried out as described above.
|
10.1371/journal.ppat.1004760 | A Plasmodium Phospholipase Is Involved in Disruption of the Liver Stage Parasitophorous Vacuole Membrane | The coordinated exit of intracellular pathogens from host cells is a process critical to the success and spread of an infection. While phospholipases have been shown to play important roles in bacteria host cell egress and virulence, their role in the release of intracellular eukaryotic parasites is largely unknown. We examined a malaria parasite protein with phospholipase activity and found it to be involved in hepatocyte egress. In hepatocytes, Plasmodium parasites are surrounded by a parasitophorous vacuole membrane (PVM), which must be disrupted before parasites are released into the blood. However, on a molecular basis, little is known about how the PVM is ruptured. We show that Plasmodium berghei phospholipase, PbPL, localizes to the PVM in infected hepatocytes. We provide evidence that parasites lacking PbPL undergo completely normal liver stage development until merozoites are produced but have a defect in egress from host hepatocytes. To investigate this further, we established a live-cell imaging-based assay, which enabled us to study the temporal dynamics of PVM rupture on a quantitative basis. Using this assay we could show that PbPL-deficient parasites exhibit impaired PVM rupture, resulting in delayed parasite egress. A wild-type phenotype could be re-established by gene complementation, demonstrating the specificity of the PbPL deletion phenotype. In conclusion, we have identified for the first time a Plasmodium phospholipase that is important for PVM rupture and in turn for parasite exit from the infected hepatocyte and therefore established a key role of a parasite phospholipase in egress.
| Leaving their host cell is a crucial process for intracellular pathogens, allowing successful infection of other cells and thereby spreading of infection. Plasmodium parasites infect hepatocytes and red blood cells, and inside these cells they are contained within a vacuole like many other intracellular pathogens. Before parasites can infect other cells, the surrounding parasitophorous vacuole membrane (PVM) needs to be ruptured. However, little is known about this process on a molecular level and Plasmodium proteins mediating lysis of the PVM during parasite egress have not so far been identified. In this study, we characterize a Plasmodium phospholipase and show that it localizes to the PVM of parasites within hepatocytes. We demonstrate that parasites lacking this protein have a defect in rupture of the PVM and thereby in host cell egress. In conclusion, our study shows for the first time that a phospholipase plays a role in PVM disruption of an intracellular eukaryotic parasite.
| The controlled exit of intracellular pathogens from host cells is an important step in infection and pathogenesis. This process is important for determining an organism’s life-cycle progression and the efficiency of a secondary infection and additionally the route and timing of egress may influence host immune responses [1]. Compared to what is known about the molecular mechanisms pathogens use to invade host cells, the process of host cell exit is much less understood. To escape from host cells, many pathogens have to disrupt two membranes, that of the vacuole they are contained within and the plasma membrane of the host cell. Although different molecules have been identified that play a role in the disruption of membranes, the precise mechanisms of membrane degradation are not well understood. For bacteria and intracellular protozoan parasites, pore-forming proteins (PFPs) have been shown to be involved in promoting vacuole escape. Similarly, bacterial phospholipases have also been identified as playing key roles in the disruption of vacuole membranes (reviewed in [1,2]).
Plasmodium parasites infect hepatocytes and red blood cells (RBCs) and inside these cells reside in a parasitophorous vacuole (PV). The PV membrane (PVM) is formed during invasion by invagination of the host cell plasma membrane [3] and is extensively modulated by the parasite through the insertion of Plasmodium-specific proteins and depletion of host proteins (reviewed in [4]). At the end of their development, parasites disrupt the PVM during the tightly regulated process of egress and are released, which is essential for progression of an infection.
For both blood and liver stage parasites, it has been shown that cysteine proteases play a key role in egress. In both stages, the general cysteine protease inhibitor E64 blocks egress from the PV and members of the Plasmodium serine repeat antigen (SERA) family are cleaved shortly before the release of parasites [5,6,7,8]. For the blood stage, subtilisin-like protease 1 (SUB1) and dipeptidyl peptidase 3 (DPAP3) have been identified to be part of a protease cascade resulting in parasite release [6,9] and recently it could be demonstrated that PbSUB1 also plays an important role for parasite egress at the end of liver stage development [10,11].
In addition to proteases, it has been demonstrated that perforin-like proteins and kinases are involved in parasite egress. Plasmodium parasites express a small, conserved family of proteins encoding perforin-like proteins (PPLPs) containing membrane-attack complex/perforin domains [12]. One of these proteins, PPLP1, has membranolytic activity and localizes to the PVM and RBC membranes of Plasmodium falciparum blood stages just before egress [13] and gametocyte stage parasites deficient in PPLP2 were unable to escape from their host RBC [14,15]. In addition, Plasmodium parasites deficient in either cGMP-dependent protein kinase (PKG) or the calcium-dependent protein kinase 5 (CDPK5) exhibit defects in parasite egress [16,17]. Furthermore, it has been shown that a liver stage-specific protein, LISP1, plays an important role in PVM disruption, but in contrast to the aforementioned proteins, LISP1 has no defined functional domain and its molecular function is unknown [18].
Despite this knowledge regarding the importance of proteases, kinases and perforin-like proteins for Plasmodium egress, the precise ordering of events and cellular mechanisms governing membrane degradation/disintegration remain unknown. Whether proteases have a direct role in membrane disruption by hydrolysis of membrane bound proteins or if they in turn activate other effectors is unclear.
Intracellular bacteria, such as Listeria and Rickettsia, use phospholipases to rupture vacuolar membranes [19,20] and it is reasonable to assume that other pathogens including Plasmodium parasites employ similar mechanisms to be liberated at the end of their development. However, the role of Plasmodium phospholipases in membrane disruption during egress has not been analyzed so far, although several putative phospholipases have been identified in Plasmodium based on sequence or structural similarity to phospholipases from other organisms (GeneDB.org).
Studies using phospholipase-C (PLC) inhibitors demonstrated that PLC activity is involved in multiple processes ranging from gametocyte development and sporozoite motility to egress of merozoites by regulating Ca2+ release [21,22,23]. Attempts to disrupt the gene encoding Plasmodium berghei PI-PLC (phosphoinositide-specific phospholipase C) have been unsuccessful, indicating an essential role for blood stage development [24]. Another protein on the surface of P. berghei sporozoites has been shown to exhibit phospholipase and membrane lytic activity in vitro [25]. This protein, P. berghei phospholipase (PbPL, PBANKA_112810), contains a predicted signal sequence and a carboxyl terminus that is 32% identical to the human lecithin:cholesterol acyltransferase, a secreted phospholipase [25]. Sporozoites deficient in PbPL expression have a reduced capacity to cross epithelial cell layers, indicating a role for PbPL in damaging host cell membranes.
In this study, we analyze the localization and role of PbPL during liver stage development. We show that PbPL is located at the PVM of liver stage parasites and that mutants deficient in PbPL exhibit a delayed egress as a result of impaired rupture of the PVM. Together, this is the first report of a protein with phospholipase activity that is involved in PVM disruption by a protozoan parasite.
PbPL was shown to be expressed on the surface of sporozoites and to have an important role during transmigration of sporozoites through cells [25]. However, it was not known whether PbPL is also expressed during liver stage development. We therefore first analyzed its transcription by RT-PCR, which showed mRNA expression of PbPL throughout liver stage development (Fig. 1A). To determine protein expression and PbPL localization, we generated a mouse antiserum against a hydrophilic fragment of PbPL (amino acid 195 to 312). Performing immunofluorescence assays (IFA) with the anti-PbPL antiserum revealed that PbPL colocalizes with the PVM resident protein exported protein I (ExpI, PBANKA_092670) in infected hepatocytes 30 and 54 hours post-infection (hpi). At 30 hpi, PbPL was also observed in vesicular structures within the parasite cytoplasm, which may be newly synthesized PbPL located in secretory vesicles being transported to the PVM. No signal was observed in liver stages of PbPL-knockout (KO) parasites (see below), confirming the specificity of the antiserum for PbPL (Fig. 1B). We further confirmed the localization by generating parasites expressing a PbPL-GFP fusion protein under the liver stage specific lisp2 (PBANKA_100300) promoter [26], in which PbPL-GFP also localized to the PVM (Fig. 1C). These observations, that PbPL is expressed by parasites in infected hepatocytes and localizes to the PVM, may indicate that PbPL also plays a role in parasite development after sporozoite invasion.
To assess the role of PbPL in liver stage development, we sought to analyze a PbPL-KO parasite line. Unfortunately, the existing PbPL deletion mutant [25] shows insufficient fluorescence to allow liver stage development analysis using our assays. For this reason, we generated a new PbPL-KO parasite line by targeted deletion of the pbpl gene by double crossover homologous recombination (Fig. 2A). A gene-specific plasmoGEM vector [27,28] was transfected into blood stage schizonts of a marker-free P. berghei reporter line (mCherryhsp70) that expresses mCherry at high levels under the control of the hsp70 (PBANKA_071190) regulatory sequences throughout the life cycle, making it particularly useful for fluorescence-based assays (S1 Fig.). Parasites of the mCherryhsp70 line are hereafter referred to as wild-type (WT) parasites. Successful deletion of PbPL was confirmed in two clonal KO parasite lines (KO1 and KO2) by diagnostic PCR (Fig. 2B).
To validate that any potential mutant phenotype is the result of the absence of PbPL, we reintroduced the pbpl gene into KO parasites, taking advantage of the fact that the vector used for generation of the PbPL-KO contains a fusion of the positive drug selectable marker hdhfr (human dihydrofolate reductase) and the negative marker yfcu (yeast cytosine deaminase and uridyl phosphoribosyl transferase) under the control of the P. berghei eef1α promoter. To allow complementation, we first removed the selectable marker by treating KO2 parasites with 5-Fluorocytosine (5-FC), selecting for marker-free PbPL-KO parasites that had undergone homologous recombination between the two 3’dhfr untranslated regions present in the targeting vector flanking the hdhfr::yfcu cassette (Fig. 2A). Successful removal of the selectable marker in a clonal KO parasite line was confirmed by diagnostic PCR (Fig. 2C). In a next step, we complemented these marker-free PbPL-KO parasites by transfection of a plasmid, in which expression of a V5-tagged PbPL is under the control of the endogenous pbpl promoter (Fig. 3A). We confirmed the correct complementation in three clonal lines (CMP1–3) by diagnostic PCR (Fig. 3B). In addition, PbPL expression during liver stage infection in these complemented lines was demonstrated by IFA using either our anti-PbPL antiserum or an anti-V5 antibody (Fig. 3C).
We first checked mosquito development of PbPL-KO parasites and found that they did not differ from WT and complemented parasites with respect to: (i) the kinetics of male gamete egress after in vitro gametocyte activation (S2A Fig.), (ii) formation of male exflagellation centers after in vitro activation (S2B Fig.), (iii) production of oocysts (S2C Fig.) and (iv) the production of sporozoites inside oocysts (midgut sporozoites, S2D Fig.). However, we detected fewer PbPL-KO sporozoites in both the mosquito hemolymph (S2E Fig.) and salivary glands (S2F Fig.) compared to both WT and complemented parasites. The reduced number of hemolymph and salivary gland sporozoites despite the production of normal numbers of sporozoites within oocysts indicates that PbPL-KO sporozoites have a defect in their egress from oocysts. A function of PbPL during sporozoite formation was not completely unexpected, since a previous study has shown that the pbpl promoter is active in oocysts [29].
When we infected C57BL/6 mice intravenously with 1,000 sporozoites of WT, KO or complemented parasites, the first blood stage parasites were detected in all mice 3 days after sporozoite injection. However, parasitemia for PbPL-KO parasites was significantly lower in comparison to both WT and complemented parasites on day 4 and subsequent days, as determined by FACS analysis (Fig. 4). These observations indicate either a reduced sporozoite infectivity, a defect during liver stage development or egress, or a reduced growth rate of blood stage parasites. To exclude that the lower parasitemia of PbPL-KO parasites after sporozoite infection is the result of a reduced growth rate of blood stages, we calculated the blood stage multiplication rate of PbPL-KO parasites from the increase in parasitemia after sporozoite infection (S3A Fig.). In addition, we analyzed blood stage growth in mice after intravenous injection of 1,000 WT, PbPL-KO or complemented blood stage parasites (S3B Fig.). The results of these analyses showed that the blood stage growth of PbPL-KO parasites was comparable to that of WT and complemented parasites, indicating that PbPL does not play a critical role during blood stage development. To analyze whether the reduced parasitemia is the result of a reduced infection of liver cells by sporozoites, we quantified the liver load of mice infected with the same number of WT, PbPL-KO or complemented sporozoites 38 hpi by real-time PCR. In agreement with previous findings [25], we were not able to detect a difference between the parasite lines (S3C Fig.), indicating a normal infectivity of PbPL-KO sporozoites.
In conclusion, the reduced parasitemia of PbPL-KO parasites after sporozoite infection in combination with comparable liver loads and the absence of an obvious blood stage phenotype indicates that PbPL-KO parasites either take longer to emerge from the liver or that fewer infectious merozoites are released.
To better characterize the PbPL-KO liver stage phenotype, we next analyzed in detail the intrahepatic development of PbPL-KO parasites in vitro. To again exclude differences in sporozoite infectivity, we infected HepG2 cells with the same number of WT, KO and complemented sporozoites and counted infected hepatocytes at 48 hpi. KO sporozoites showed a similar infectivity to HepG2 cells in comparison to WT and complemented sporozoites (Fig. 5A), further supporting our in vivo analyses.
To investigate whether PbPL plays a role during the growth of Plasmodium liver stages, we measured the size of intrahepatic parasites in vitro at 48 hpi. No significant differences in size were observed between WT, KO and complemented parasites, suggesting that PbPL is not involved in liver stage development prior to this stage (Figs. 5B and S4A).
We subsequently investigated a potential role of PbPL during the final stages of liver stage schizogony, merozoite formation and egress from host hepatocytes. First, we showed by IFA that expression and localization of the merozoite surface protein 1 (MSP1, PBANKA_083100) and the PVM marker Exp1 in intrahepatic PbPL-KO parasites was the same as in WT parasites (S4B Fig.).
Next, we counted the number of detached cells produced at the end of intrahepatic development, as a marker for the final phase of liver stage development in vitro [5]. Cells detach upon formation of merozoites and rupture of the PVM, followed by the release of merozoites into the hepatocyte cytoplasm, which typically occurs between 55 and 60 hpi [5]. We observed that PbPL-KO parasites produced approximately 60% fewer detached cells compared to WT parasites (Figs. 5C and S4C). Furthermore, a significant proportion of detached cells in the KO population showed an aberrant morphology; merozoites were not released into the hepatocyte cytoplasm but remained clustered together (Fig. 5D). In complemented parasites, the WT phenotype of detached cells was completely rescued (Fig. 5C, D).
The reduced detachment in the case of mutant parasites could result from fewer parasites that successfully form merozoites or from a defect in PVM disruption. In the latter case, an increase in attached hepatocytes containing merozoites may be predicted because daughter cells would be produced normally but due to a defect in PVM disruption, cell detachment would not occur. To distinguish between these possibilities, we quantified attached hepatocytes containing schizont, cytomere and merozoite stages at 54 and 65 hpi. At 54 hpi, before PVM rupture, no significant difference in the number of hepatocytes containing merozoites existed between WT, PbPL-KO and complemented parasites, indicating normal merozoite formation up to this time point (Fig. 6). In contrast, at 65 hpi, a time point where the PVM is disrupted in normally developed WT parasites, a significantly higher number of attached hepatocytes containing merozoites was seen for PbPL-KO parasites compared to WT parasites, indicating impaired merozoite release and a potential defect in PVM disruption (Fig. 6).
We were able to rescue the WT phenotype by complementation, although complemented parasites showed a small but statistically significant increase in merozoite release compared to WT parasites in this assay. This difference in merozoite release could be the result of slightly increased PbPL-expression levels in complemented parasites compared to WT parasites. This may be explained by the process of complementation, where the DNA construct encoding the pbpl gene was introduced into the PbPL-KO genome by single-crossover recombination, possibly resulting in multiple insertions of the DNA construct. This increase in pbpl gene copy number could then lead to higher PbPL expression levels compared to that in WT parasites. Further, even the expression level of a single introduced PbPL expression construct might result in differences in PbPL expression, since we used only 1067 bp upstream of the pbpl start codon as a promoter region in the complementation vector, and consequently may not have captured the entire pbpl promoter, resulting in altered expression levels.
To further analyze impaired merozoite release in PbPL-KO parasites and a potential role of PbPL in PVM disruption, we infected GFP-expressing HepG2 cells with WT and PbPL-KO parasites and analyzed their intrahepatic development from the cytomere stage to cell detachment by live-cell time-lapse microscopy. An intact PVM is impermeable to host cell-derived GFP, whereas PVM rupture leads to a rapid GFP influx into the PV [30]. Analysis of GFP influx by live-cell time-lapse microscopy therefore allows the determination of the percentage of merozoite-forming parasites with a disrupted PVM and quantification of the speed of PVM disintegration. Nearly all WT parasites that developed to the merozoite stage were able to disrupt the PVM, with an average time of 70 minutes (Fig. 7, S1 Movie). In contrast, PbPL-KO parasites that developed to the merozoite stage either did not rupture the PVM at all or this process was significantly delayed (Fig. 7, S2 Movie). Importantly, the WT phenotype was completely rescued in complemented parasites (Fig. 7, S3 Movie). In conclusion this experiment shows that in the absence of PbPL, PVM rupture is compromised, confirming the results of the detached cell (Figs. 5C, and S4C) and the stage quantification assays (Fig. 6) and providing a perfect explanation for the observed delay in development of blood stage parasitemia after sporozoite infection (Fig. 4).
In intracellular bacterial pathogens, phospholipases have been shown to play a key role in host cell exit [1,2], whereas the role of this class of proteins in intracellular protozoans, including Plasmodium parasites, was unknown. In this study we have identified PbPL, a Plasmodium protein with phospholipase activity, as having a key role in disruption of the liver stage PVM. This is the first time that a Plasmodium phospholipase has been implicated in egress from a host cell.
Bhanot et al. have previously found that PbPL has phospholipase activity, is expressed on the surface of sporozoites and that it has a role in damaging host cell membranes, thereby assisting in the migration of sporozoites to the hepatocyte [25]. Together with our observations, demonstrating that PbPL-KO parasites are impaired in their egress from oocysts and from hepatocytes, this protein appears to have important functions in three life cycle stages prior to the blood stage of infection:
(i) in oocysts, where it plays a role in egress of sporozoites, (ii) in sporozoites, where it plays a role in migration through host tissue mediating hepatocyte invasion and (iii) in liver stages, where it is involved in PVM rupture mediating efficient merozoite release.
When Bhanot et al. injected a large number of PbPL-deficient sporozoites intravenously, they did not observe a prolonged prepatent period by blood smear. However, they observed a delayed prepatency when a much lower number of sporozoites were transferred by mosquito bite [25]. Our own observations principally confirm the finding of Bhanot et al., as the first blood stage parasites were detected in all mice 3 days after sporozoite injection. However, determination of parasitemia by the more sensitive method of FACS analysis revealed a significantly lower parasitemia for KO parasites, suggesting that PbPL also contributes to parasite egress in vivo.
In support of the function PbPL has in PVM disruption, the protein is located at the PVM, already being detectable there at 30 hpi, some time before the actual rupture of the PVM occurs. This may suggest that in the developing liver stage parasite, PbPL remains in an inactive state and its activation results in PVM rupture. A coordinated cascade of events involving kinases and resulting in the activation of proteases has been defined as being important for parasite release (reviewed in [31]) and this signaling and protease activation cascade may also include PbPL activation, for example by phosphorylation or proteolytic cleavage. As kinases and proteases are not likely membranolytic, an attractive scenario is that their activation converges on lipases like PbPL. In line with this hypothesis, proteolytic activation of phospholipases was already shown for a Listeria PLC [32] and for several secreted phospholipases of Staphylococcus (reviewed in [33]). Candidates for a potential proteolytic activation of PbPL could be the proteases of the SERA family that, like PbPL, also localize to the PVM in infected hepatocytes [7,8]. However, so far proteolytic cleavage of substrates has not been directly shown for any of the P. falciparum or P. berghei SERAs. Direct and indirect evidence that SERAs with a cysteine residue in the active center are indeed real proteases came from two recent studies: One study showed that the putative active site cysteine of P. falciparum SERA6 is essential and that the P. berghei orthologue of SERA6 could be converted by SUB1-mediated cleavage to an active cysteine protease showing autoprocessing activity [34]. The other study revealed that the exchange of the serine residue in the active center of P. falciparum SERA5 to a cysteine allows peptide binding and cleavage [35]. These observations, the localization of different SERAs in the PV or even in the PVM and the fact that expression of the majority of P. berghei SERAs is restricted to the last few hours before merozoite egress from infected hepatocytes [7,8], suggest a possible role in initiation of PVM rupture by, for example, processing and activation of PbPL or other membranolytic enzymes.
While our study demonstrates that PbPL is involved in disruption of the PVM, our observations also show that a proportion of PbPL-KO parasites are able to disrupt the PVM inside the infected hepatocyte in the absence of PbPL. This indicates that PVM disruption can be brought about by other parasite molecules, albeit less efficiently, or can occur by non-specific (possibly mechanical) processes. The ability of the parasite to utilize multiple exit strategies is further supported by the absence of any obvious blood stage phenotype in PbPL-deficient parasites. Interestingly, the absence of other Plasmodium proteins shown to be involved in egress, such as LISP1 [18] and the perforin-like molecule PPLP2 [14,15], also did not result in an absolute ‘non-egress’ phenotype. Host cell egress is such a critical process for intracellular parasites that presumably several different effector proteins and mechanisms exist, which may have partially overlapping, redundant or even synergistic functions. Therefore, the absence of just one effector molecule, such as PbPL, may only lead to a partial defect in egress. An example of this functional redundancy has been described in Listeria [19], where separate deletion of two phospholipases had only moderate effects on the infectivity to mice (2–20 fold reduction), whereas the deletion of both phospholipases together severely impaired infectivity (500 fold reduction). In line with this, the apparently moderate phenotype resulting from the deletion of PbPL enables the generation of double mutants in the future, providing an opportunity to identify other proteins working in conjunction with PbPL, and thereby may help us to more completely understand the process and the hierarchy of events that facilitate parasite exit from the host cell.
Our current working hypothesis is that PbPL works together with other phospholipases or pore-forming proteins (PFPs), like the already mentioned Plasmodium perforin-like proteins [12]. In general, the combination of membranolytic enzymes might be specific for each egress event in the life cycle and for different Plasmodium species, depending on the composition of the PVM that the parasite has to rupture and the exit strategy employed. It makes sense, for example, that the lysis of the RBC PVM differs from that of infected hepatocytes, as in RBCs the same enzymes probably also rupture the host plasma membrane within seconds, whereas the plasma membrane of hepatocytes needs to be preserved for several hours until merosomes are formed and reach the blood stream [5].
We consider PFPs as promising candidates to be involved in PbPL-mediated PVM disruption, as they were shown to play a key role in host cell egress of intracellular bacteria, for which it has been suggested that PFPs and phospholipases may act in concert [36]. Although the detailed molecular basis of this is unknown, PFPs might make certain membrane leaflets accessible for the action of lipases. In Plasmodium the PFPs of the perforin-like protein family consist of five conserved proteins all containing a membrane-attack complex/perforin domain [12]. Members of this family have been shown to be important for ookinete and sporozoite host cell traversal [37,38,39] and have also been implicated in host cell egress by asexual and sexual blood stage parasites [13,14,15]. However, studies on the role of perforin-like proteins at the end of liver stage development are so far missing and it can only be speculated that these proteins contribute to parasite egress from infected hepatocytes. Interestingly, key roles of PFPs in egress have also been established for several other protozoan parasites, including Trypanosoma cruzii, Leishmania and Toxoplasma gondii [40,41,42], but a putative function of phospholipases in exit of these parasites from their host cells has not been investigated so far.
Taken together, our study identifies PbPL as the first Plasmodium phospholipase that is important for PVM disruption and in turn for parasite exit from the infected hepatocyte and therefore establishes a key role of a parasite phospholipase in egress. We strongly believe that PbPL, because its function is significant but not lethal, offers the unique opportunity to learn more about parasite egress strategies. We now aim to identify the specific combination of membranolytic enzymes needed for membrane rupture and to understand the mechanisms by which these enzymes act together, as they might represent a new class of parasite-specific targets for intervention.
All experiments performed at the University of Bern were conducted in strict accordance with the guidelines of the Swiss Tierschutzgesetz (TSchG; Animal Rights Laws) and approved by the ethical committee of the University of Bern (Permit Number: BE109/13). All experiments performed at the LUMC were approved by the Animal Experiments Committee of the Leiden University Medical Center (Permit Number: DEC 12042). The Dutch Experiments on Animals Act was established under European guidelines (EU directive no. 86/609/EEC regarding the Protection of Animals used for Experimental and Other Scientific Purposes).
Mice used in the experiments were between 6 and 10 weeks of age and were from Harlan Laboratories, Charles River or bred in the central animal facility of the University of Bern. Mosquito feeds were performed on mice anaesthetized with Ketavet/Domitor and all efforts were made to minimize suffering.
For the generation of mCherryhsp70 parasites, Swiss mice were used. The in vivo phenotype of PbPL-KO parasites was analyzed in C57BL/6 mice, while for all other experiments Balb/c mice were used.
Infections of mice were initiated by intraperitoneal injection of P. berghei blood stabilates. When these mice had a parasitemia of 4%, 150 μl or 40 μl of infected blood were injected intraperitoneally or intravenously, respectively, into mice that had received an intraperitoneal injection of 200 μl phenylhydrazine (6 mg/ml in PBS) 2–3 days before. At day 3 or 4 after infection, mice with a parasitemia of at least 7% were anaesthetized for 1 hour to allow feeding of 150 female Anopheles stephensi mosquitoes. The next day, unfed mosquitoes were removed. Mosquitoes were kept at 20.5°C with 80% humidity and for infection experiments, sporozoites were isolated from infected salivary glands 16–27 days after the infective blood meal.
HepG2 cells (obtained from the European cell culture collection) were cultured as described before [30]. For infection, either 3 or 5 x 104 cells were seeded in 24-well plates with or without coverslips. The next day, P. berghei sporozoites were isolated from the salivary glands of infected A. stephensi mosquitoes and added to HepG2 cells in culture medium additionally containing 2.5 μg/ml amphotericin B (PAA Laboratories). After an incubation period of 2 hours, the sporozoite-containing medium was removed and fresh infection medium was added. Subsequently, medium was changed once per day.
Total RNA was isolated from 0.05% saponin-treated P. berghei infected red blood cells and infected HepG2 cells 24, 48, 54 and 60 hpi using the NucleoSpin RNA II kit (Macherey-Nagel). Random-primed cDNA synthesis was performed using GoScript reverse transcriptase (Promega) and the resulting cDNA was then used as a template in PCR reactions using GoTaq Flexi DNA polymerase (Promega) with the primer pairs PbPL-expression-fw/PbPL-expression-rev and GAPDH-expression-fw/GAPDH-expression-rev. All primer sequences are listed in S1 Table.
All PCR reactions were performed using Phusion DNA polymerase (NEB). PCR products were routinely cloned into pJET1.2 (Fermentas) and confirmed by sequencing.
For generation of the PbPL bacterial expression vector parallel-1-His-PbPL195–312, the coding sequence corresponding to a hydrophilic fragment of PbPL ranging from amino acid 195 to 312 was amplified from blood stage cDNA using primer pair PbPL-antiserum-fw/PbPL-antiserum-rev, which was then cloned into the parallel-1-His vector [43] using BamHI and XhoI restriction sites.
The PbPL-GFP expression vector pL0043LSPbPL-GFPCmCherry was generated by first amplifying the PbPL coding sequence from blood stage cDNA using primer pair PbPL-GFP-fw/PbPL-GFP-rev, which was subsequently digested with BglII and ligated into the BamHI digested liver stage-specific expression vector pGFP103464 [26] in frame with GFP. From there the LSPbPL-GFP expression cassette was cloned via EcoRV and KpnI into the pL0043 vector [44], which targets the P. berghei 230p locus by double crossover homologous recombination. Finally, a constitutive mCherry expression cassette was integrated via KpnI, which had been amplified before from the pCmCherry plasmid [45] using primers mCherry-fw and mCherry-rev.
For generation of the PbPL-complementation vector pL0017.1.2-5’FR-PbPL-V5, the vector pL0017.1.2 was generated at first. For this, the GFP coding sequence was excised from pGFP103464 [26] using BamHI and XbaI digestion and then replaced by a double-stranded DNA oligonucleotide (obtained by annealing the single-stranded DNA oligonucleotides 5′-GATCCGCGGCCGCCCTAGGAGGTAAGCCTATCCCTAACCCTCTCCTCGGTCTCGATTCTACGTAGT-3′ and 5′-CTAGACTACGTAGAATCGAGACCGAGGAGAGGGTTAGGGATAGGCTTACCTCCTAGGGCGGCCGCG-3′), bearing the coding sequence for the V5-tag (underlined) as well as a NotI restriction site. In a second step, the PbPL coding sequence and 1067 bp of the upstream region (endogenous promoter) were amplified in two parts from P. berghei genomic DNA. The upstream region and the N-terminus of PbPL were amplified with primer PbPL-5FR-fw (containing a SacII restriction site) and primer PbPL-nterm-rev. The C-terminal part of PbPL was amplified with primer PbPL-cterm-fw and primer PbPL-cterm-rev (containing a NotI restriction site). Next, a SnaBI restriction site present within the PbPL coding sequence was used to join both parts together in a 3-way-ligation, for which a SacII/NotI digested intermediate vector was used. The resulting vector was linearized with SacII and subsequently blunted, followed by NotI digestion. This resulted in release of the joined endogenous promoter region and the PbPL coding sequence, which could be then ligated in frame with the V5 tag into the EcoRV/NotI-digested pL0017.1.2 vector, thereby replacing the lisp2 promoter and generating pL0017.1.2-5’FR-PbPL-V5. All primer sequences are listed in S1 Table.
His-PbPL195–312 was expressed from plasmid parallel-1-His-PbPL195–312 in the BL21 (DE3) [pAPlacIQ] E. coli strain and purified using Ni-NTA agarose (Qiagen). For generation of the His-PbPL antiserum, 50 μg of the purified protein were mixed with one volume of Freund’s adjuvant complete (Sigma) and subcutaneously injected into a Balb/c mouse. After two weeks, the mouse was boosted with the same amount of protein mixed with Freund’s adjuvant incomplete (Sigma), followed by a second boost two weeks later. The immunized mouse was sacrificed, blood was collected and antiserum was obtained after centrifugation of the coagulated blood.
The linearized plasmid pL0043LSPbPL-GFPCmCherry was transfected into blood stage schizonts of the GIMOANKA parasite line [44] as described previously [46] and selection of transfected parasites was carried out by adding 5-FC (Abcam) to the drinking water of infected mice [47].
3 x 104 HepG2 cells were seeded on coverslips in 24-well plates and infected the following day with P. berghei sporozoites. At different time points post-infection, cells were fixed with 4% PFA in PBS for 20 minutes at room temperature, followed by permeabilization with ice-cold methanol. Unspecific binding sites were blocked by incubation in 10% FCS in PBS, followed by incubation with primary antibodies (rabbit anti-GFP (Invitrogen), rat anti-V5 (Invitrogen), mouse anti-PbPL, chicken anti-Exp1 and rat anti-MSP1 (both generated at the Bernhard-Nocht Institute, Hamburg, Germany)) and subsequently with fluorescently labeled secondary antibodies (anti-rabbit AlexaFluor488 (Invitrogen), anti-rat AlexaFluor488 (Invitrogen), anti-chicken Cy5 (Dianova), anti-mouse AlexaFluor488 (Invitrogen)) diluted in 10% FCS in PBS. DNA was visualized by staining with 1 μg/ml DAPI (Sigma). Labeled cells were mounted on microscope slides with Dako Fluorescent Mounting Medium (Dako) and analyzed by confocal point scanning microscopy using a Zeiss LSM5 Duo microscope and a Zeiss Plan-Apochromat 63×/1.4 oil objective. Image processing was performed using ImageJ.
The construct pL1694 was used to generate the reporter line mCherryhsp70 that expresses mCherry under control of the hsp70 regulatory sequences. This construct was used to target the GIMOANKA mother line using the ‘gene insertion/marker out’ (GIMO transfection) procedure [44]. To create pL1694, we modified the existing construct pL1628 [44]; this is a Pb230p GIMO targeting construct that contains a gene encoding mCherry where expression is driven by the P. berghei eef1α promoter and transcription is terminated by a P. berghei dhfr 3’ UTR region. We removed both the eef1α promoter and 3’ dhfr UTR regions from pL1628 and replaced them with the promoter region and 3’ UTR (transcription terminator) sequences of P. berghei hsp70 (PBANKA_071190). The hsp70 regulatory regions were amplified from P. berghei ANKA genomic DNA using primers Hsp70-Promoter-fw/Hsp70-Promoter-rev for the promoter and primers Hsp70-3'UTR-fw/ Hsp70-3'UTR-rev for the 3’UTR. These promoter and 3’UTR elements were cloned into pL1628 vector using the Asp718/BamHI and SpeI/Asp718 restrictions sites respectively. This construct was linearized by digestion with KspI before transfection. The linearized DNA construct was introduced into GIMOANKA parasites using standard methods of GIMO-transfection [44]. Transfected parasites were selected in mice by applying negative selection by providing 5-FC in the drinking water of mice [47]. Negative selection results in selection of parasites where the hdhfr::yfcu selectable marker in the 230p locus is replaced by the mCherry reporter-cassette. Selected transgenic parasites (mCherryhsp70) were cloned by limiting dilution. Correct integration of the constructs into the genome of mCherryhsp70 parasites was analyzed by diagnostic PCR on parasite gDNA. All primer sequences are listed in S1 Table.
PbPL-KO parasites were generated by transfection of the plasmoGEM vector PbGEM-099883 [27,28] into mCherryhsp70 parasites as described before [46], targeting the PbPL coding sequence by double crossover homologous recombination, and were selected by pyrimethamine (Sigma). For generation of marker-free KO parasites, the selectable marker was removed by negative selection with 5-FC in the drinking water of infected mice as described previously [47]. Subsequently, complemented parasites were generated by transfection of the linearized pL0017.1.2-5’FR-PbPL-V5 vector into marker-free KO parasites, leading to expression of PbPL-V5 under its endogenous promoter from the c- or d-ssu-rRNA locus. Parasite genomic DNA (gDNA) was isolated from 0.05% saponin-treated infected red blood cells using the Nucleospin Blood QuickPure kit (Macherey-Nagel) and all genetic modifications of parasites were confirmed by diagnostic PCR using GoTaq Flexi DNA polymerase. All primer sequences are listed in S1 Table. Clonal parasite lines were either generated by detached cell injection [48] or by limiting dilution.
Exflagellation of male gametocytes was analyzed using standard in vitro gametocyte activation assays basically as described previously [49]. In brief, 2 μl of tail blood were added to 8 μl of ookinete medium (RPMI1640 containing 25 mM HEPES, 20% FCS, 100 μM xanthurenic acid [pH 7.4]) and the mixture was placed under a Vaseline-coated coverslip. By light microscopy (100x objective) we counted the number of unemerged but activated male gametocytes (defined as activated male gametocytes with moving flagella inside the erythrocyte) and of emerged activated male gametocytes (with extracellular male gametes) in four 2 minute intervals starting at 8 minutes post-activation. 20 minutes post-activation, the characteristic exflagellation centers were counted in 20 fields of view using a light microscope (40x objective).
9 days after the infective blood meal, midguts of 15–23 mosquitoes were dissected into PBS and the pooled midguts were fixed in 4% paraformaldehyde (PFA) in PBS for 20 minutes at room temperature. The midguts were then washed with PBS and stored in PBS at 4°C in the dark. The next day, fixed midguts were mounted on glass slides containing a small amount of Dako Fluorescent Mounting Medium and imaged using a fluorescence microscope with a 5x objective. The average number of oocysts per midgut was then determined using an ImageJ-based counting macro [50].
On day 18 and 26 after the infective blood meal, mosquito midguts and salivary glands were harvested for determination of sporozoite numbers. 10 mosquitoes were dissected, organs pooled and homogenized, and released sporozoites were counted using a hemocytometer. For determination of hemolymph sporozoite numbers, hemolymph from 10 mosquitoes was collected on day 18 after the infective blood meal by perfusion of the thorax and abdomen with 50 μl of PBS and sporozoites were counted using a hemocytometer.
1,000 WT, KO or complemented sporozoites were injected intravenously into 6–7 C57BL/6 mice per group. Subsequently, blood stage parasitemia was determined between day 3 and 6 post-infection by FACS analysis using a FACSCalibur flow cytometer (BD Biosciences) and the mCherry fluorescence of parasites.
For determination of blood stage growth in mice, 1,000 infected red blood cells (containing mixed blood stages) of WT, KO or complemented parasites were injected intravenously into 4–5 C57BL/6 mice per group and subsequent parasitemia was determined between day 3 and 6 post-infection by FACS analysis.
For determination of liver loads, 4–5 C57BL/6 mice per group were injected intravenously with 10,000 WT, KO or complemented sporozoites. After 38 hours, whole livers were removed and homogenized on ice in 5 ml of denaturing solution (4 M guanidine thiocyanate, 25 mM sodium citrate [pH 7.0], 0.5% N-lauroylsarcosine, 0.7% β-mercaptoethanol) using a Polytron homogenizer (Kinematica). Total RNA was isolated from 100 μl of liver homogenate using 900 μl of TRIzol (Ambion). Random-primed cDNA synthesis was performed using 2 μg of total RNA and GoScript reverse transcriptase. Levels of parasite 18S ribosomal RNA and mouse hypoxanthine guanine phosphoribosyltransferase (HPRT) cDNAs obtained from the reaction were quantified by real-time PCR using previously described primers [10]: Pb18S-fw and Pb18S-rev for P. berghei 18S ribosomal RNA and MmHPRT-fw and MmHPRT-rev for the Mus musculus housekeeping gene hprt. To quantify gene expression, MESA GREEN qPCR MasterMix Plus (Eurogentec) was used according to the manufacturer's instructions. Reactions were performed in triplicate in an ABI Prism 7000 sequence Detection System (Applied Biosystems) with 1 μl of cDNA in a total volume of 25 μl and the following reaction conditions: 1 step of 2 min at 50°C, 1 step of 5 min at 95°C, 40 cycles of 15 sec at 95°C and 1 min at 60°C. Relative expression levels were calculated using the ΔΔCt method [51]. All primer sequences are listed in S1 Table.
5 x 104 HepG2 cells per well were seeded in 24-well plates and infected the next day with 10,000 WT, KO and complemented sporozoites. After 48 hpi, the average number of infected host cells per well was quantified in triplicate.
5 x 104 HepG2 cells per well were seeded in 24-well plates and infected the next day with WT, KO and complemented sporozoites. 48 hpi, parasite size (area) was determined by density slicing using ImageJ and infected cells were counted. At 65 hpi, the number of detached cells (DCs) in the supernatant was counted in triplicate. The percentage of DC formation was then calculated by dividing the number of DCs in the supernatant by the number of infected cells at 48 hpi. For quantification of DC morphology, DCs were harvested at 65 hpi and DCs with normal (merozoites freely distributed in host cell cytoplasm) and abnormal morphology (merozoites still being clustered) were counted.
3 x 104 HepG2 cells per well were seeded in 24-well plates on coverslips and infected the next day with WT, KO and complemented sporozoites. They were fixed at 54 and 65 hpi and stained for IFA with an anti-MSP1 antiserum as already described. Subsequently, attached hepatocytes containing schizont, cytomere and merozoite stage parasites were counted based on parasite plasma membrane morphology.
2 × 106 HepG2 cells were transfected with 4 μg pEGFP-N3 plasmid (Clontech) using an Amaxa nucleofector (Lonza) as described previously [30] and subsequently seeded into 4-chamber glass bottom dishes (In Vitro Scientific). The following day, two wells of GFP-expressing cells were infected with mCherry-expressing WT and the other two either with KO or complemented sporozoites. The percentage of merozoite-forming parasites that rupture the PVM, as defined by influx of GFP into the PV, as well as the time between successful formation of merozoites and PVM rupture was subsequently analyzed by live-cell time-lapse imaging. For this, a Zeiss LSM5 Duo microscope with a Zeiss Plan-Apochromat 63×/1.4 oil objective was used in the LIVE mode (confocal line scanning). Development of parasites shortly before or in the cytomere stage was followed for 12 hours starting between 54 and 56 hpi using the Zeiss LSM Multitime-Macro and an image was acquired every 10 minutes. During imaging, cells were kept in a CO2 incubator at 37°C. Only parasites that developed to the merozoite stage within the first 6 hours of imaging and displayed completely normal development (e.g. absence of merofusosomes [52]) were used for further analysis. Image processing was performed using ImageJ.
Statistical analyses were performed using GraphPad Prism (GraphPad Software). For comparisons between groups, a one-way analysis of variance (ANOVA) followed by a Holm-Sidak multiple comparison test was perfomed. P values of < 0.05 were considered significant.
P. berghei phospholipase (PBANKA_112810), P. berghei glyceraldehyde-3-phosphate dehydrogenase (PBANKA_132640), P. berghei LISP2 (PBANKA_100300), P. berghei exported protein 1 (PBANKA_092670), P. berghei heat shock protein 70 (PBANKA_071190), P. berghei 18S ribosomal RNA (berg07_18S), P. berghei merozoite surface protein 1 (PBANKA_083100), M. musculus hypoxanthine guanine phosphoribosyltransferase (NM_013556.2).
|
10.1371/journal.pgen.1000148 | Evolutionary Convergence on Highly-Conserved 3′ Intron Structures in Intron-Poor Eukaryotes and Insights into the Ancestral Eukaryotic Genome | The presence of spliceosomal introns in eukaryotes raises a range of questions about genomic evolution. Along with the fundamental mysteries of introns' initial proliferation and persistence, the evolutionary forces acting on intron sequences remain largely mysterious. Intron number varies across species from a few introns per genome to several introns per gene, and the elements of intron sequences directly implicated in splicing vary from degenerate to strict consensus motifs. We report a 50-species comparative genomic study of intron sequences across most eukaryotic groups. We find two broad and striking patterns. First, we find that some highly intron-poor lineages have undergone evolutionary convergence to strong 3′ consensus intron structures. This finding holds for both branch point sequence and distance between the branch point and the 3′ splice site. Interestingly, this difference appears to exist within the genomes of green alga of the genus Ostreococcus, which exhibit highly constrained intron sequences through most of the intron-poor genome, but not in one much more intron-dense genomic region. Second, we find evidence that ancestral genomes contained highly variable branch point sequences, similar to more complex modern intron-rich eukaryotic lineages. In addition, ancestral structures are likely to have included polyT tails similar to those in metazoans and plants, which we found in a variety of protist lineages. Intriguingly, intron structure evolution appears to be quite different across lineages experiencing different types of genome reduction: whereas lineages with very few introns tend towards highly regular intronic sequences, lineages with very short introns tend towards highly degenerate sequences. Together, these results attest to the complex nature of ancestral eukaryotic splicing, the qualitatively different evolutionary forces acting on intron structures across modern lineages, and the impressive evolutionary malleability of eukaryotic gene structures.
| The spliceosomal introns that interrupt eukaryotic genes show great number and sequence variation across species, from the rare, highly uniform yeast introns to the ubiquitous and highly variable vertebrate intron sequences. The causes of these differences remain mysterious. We studied sequences of intron branch points and 3′ termini in 50 eukaryotic species. All intron-rich species exhibit variable 3′ sequences. However, intron-poor species range from variable sequences, to uniform branch point motifs, to uniform branch point motifs in uniform positions along the intronic sequence. This is a more complex pattern than the clear relationship between intron number and 5′ intron sequence uniformity found previously. The correspondence of sequence uniformity and intron number extends to species of the green algal genus Ostreococcus, in which the single intron-rich genomic region shows far more variable intron sequences than in the otherwise intron-poor genome. We suggest that different concentrations of spliceosomal complexes may explain these differences. In addition, we report the existence of 3′ polyT tails in diverse eukaryotic protists, suggesting that this structure is ancestral. Together, these results underscore the complexity of ancestral eukaryotic splicing, the qualitatively different evolutionary forces acting on intron sequences in modern eukaryotes, and the impressive evolutionary malleability of eukaryotic genes.
| Spliceosomal introns are genomically-encoded sequences that are removed from RNA transcripts by the spliceosome, a massive RNA-protein complex. The spliceosome and spliceosomal introns are common and ancestral to eukaryotes [1]–[4], however spliceosomal organization shows striking divergence across species. Intron number per genome differs by orders of magnitude, from fewer than ten known introns in the genomes of some protist species [4],[5] to nearly ten per gene in some metazoans [6],[7]. Introns also vary dramatically in length, from the ‘bonzaied’ 19 nt introns of the Bigelowiella natans nucleomorph [8] to the giant kilobases-long introns of humans and other mammals.
Intron sequence elements, which are important for intron recognition by the RNA and protein components of the spliceosome, also vary significantly across species. In some species, the consensus of an intron sequence is largely restricted to an initial 5′ GT dinucleotide (the “donor” site), a terminal 3′ AG (the “acceptor”), and a degenerate few nt “branch point” site, lying somewhere within the intron. In other species, sequences are more conserved. For instance, in the apicomplexan parasite Cryptosporidium, 84% of introns begin with the most common sixmer GTAAGT, enabling close complementary base pairing with the U1 RNA of the spliceosome (compared with only 14% for humans; see examples in Figure 1). We previously studied the phylogenetic pattern of conservation of 5′ intronic sequences and found a strong correspondence between species with very few introns and those with such strong 5′ consensus sequences [9].
Clear differences are also observed in other structures. For instance, branch points vary across species from a highly conserved seven-mer (TACTAAC) in hemiascomycetous yeasts (e.g., Saccharomyces cerevisiae) to several lineages in which previous studies have failed to find branch points [10],[11]. Here we report studies of the evolution of 3′ intron structures including the branch point, poly-pyrimidine tract and 3′ splice site across 50 species spanning all major eukaryotic kingdoms (opisthokonts, amoebozoans, red and green plants, chromalveolates, excavates and rhizarians, Figure 2).
The branch point is an internal intronic sequence that initiates the splicing event through a hydrophilic attack by an adenosine 2′ hydroxyl group at the 5′ splice site [12],[13]. As with donor sites, branch point consensus strength varies across species: 84% of Saccharomyces cerevisiae introns utilize the sequence TACTAAC and 94% have ACTAAC, compared to fewer than 20% of human introns with the exact same sixmer at the branch point site (Figure 1), and with Caenorhabditis nematodes, where no branch points have been identified.
We first studied intron branch points in available genomes from intron-poor species (Table 1; defined, as before, as those species with ∼0.1 or fewer introns per gene on average [9]). Many of these species exhibited clear branch point consensus sequences. First, we found that the S. cerevisiae branch point consensus ACTAAC (throughout, the putative branch point A is underlined) is found in all fully-sequenced hemiascomycetous yeasts, with 66–100% of introns in each species containing this motif. Extended branch points of up to eight bases were found in some species (Table 1). We also found an extended single ACTAACC branch point motif in all 26 known introns of the red alga Cyanidioschyzon merolae. Other intron-poor species containing strong putative branch point motifs were the two excavate species: 32/34 (94.12%) probable introns (see Materials and Methods) in the flagellated protist Trichomonas vaginalis use WCTAAC, and all five known introns in Giardia lamblia (four published plus one unpublished instance) contain CTRACA. Finally, excluding two questionable predicted introns (see Discussion), all 13 introns in the microsporidium parasite Encephalitozoon cuniculi contain a TAAYTT hexamer (9/13 have CTAAYTT). Thus all lineages with strong branch points conformed to a general WCTRAYN consensus.
However, in at least one intron-poor lineage we failed to find such clear branch point sequences. Visual inspection and computational analysis (see below) of the intron-poor apicomplexan parasite Cryptosporidium parvum failed to reveal a potential branch point site. The case for the few introns of the Guillardia theta nucleomorph is less clear. 14/16 predicted G. theta introns contain a YAAY branch point-like sequence between 2 and 6 nucleotides from the acceptor AG (compared to only 2/16 that contain such a motif within the next 10 positions (7 to 17)). Intriguingly, a second YAAY signal exists further upstream – 8/16 introns have a YAAY 24–28 nt from the 3′ terminus. The single known intron in each of the sequenced Trypanosomatid genomes does not show a clear branch point structure; however it is of note that despite having few cis-spliced introns, these species do have large numbers of 3′ splicing boundaries due to the ubiquity of spliceosome-mediated spliced leader trans-splicing [14],[15].
We next studied conservation of branch points in 33 more intron-rich species. We studied the occurrence of motifs conforming to the consensus branch point WCTRAY or CTRAYN with varying levels of two-fold degeneracy (i.e. allowing two possible nucleotides at each site). For instance, the ACTAAC hexamer common to all C. merolae introns has no degeneracy, whereas the CTRACA motif of G. lamblia has one degenerate site, and the general consensus WCTRAY contains three two-fold degenerate sites. As with the other sites, either one or two nucleotides at a time were allowed at the final “N” site (CTRAYA, CTRAYC, CTRAYR, CTRAYY…).
For each species, and for each level of degeneracy (zero to three degenerate sites), we calculated the fraction of introns that contain the same motif within the last 200 nt of the intron (see Methods). Then, for each level of degeneracy, we identified the motif that was present in the largest fraction of introns for that species. These values are given in Table 2. As expected, the intron-poor species discussed above give high values at all levels of degeneracy. All intron-poor species with strong consensus discussed above have values of at least 72% for one-site degenerate branch point motifs.
By contrast, every studied intron-rich species shows much lower scores, with lower than 22% of introns with the same putative branch point motif (for example, 8.73% (ACTAAT) in Drosophila melanogaster or 10.97% (ACTGAC) in Aspergillus fumigatus), and less than 36% allowing one degenerate site. This relation between intron numbers and branch point consensus strength is underscored in Figure 3. The species are clearly distributed in two main groups (with two exceptions, G. thetha NM and C. parvum (red asterisks)): intron-rich/weak branch point consensus and intron-poor/strong branch consensus (intron-poor, fewer than 0.15 introns per gene; strong branch points, same BP-like hexamer in more than 50% of introns (red lines in Figure 3A). Overall, there is a negative correlation (r = −0.75) between intron numbers and branch point consensus conservation (by a linear regression analysis, Figure 3B).
Only the species Ostreococcus lucimarinus appeared to represent an intermediate between strong and weak branch point species, with 41.15% of predicted introns containing the same branch point-like ACTGAC sequence (Table 2). However, given the high evolutionary distance of O. lucimarinus from other species with fully-sequenced genomes and the relatively small number of available transcript sequences, annotation in O. lucimarinus and distant congenitor O. tauri is difficult and some intron predictions may thus be unreliable. To identify a confident set of introns, we performed BLASTN searches of the predicted intron-containing coding sequences against the 17,592 available O. lucimarinus EST sequences. Further computational and manual filtering for ambiguities and potential problems associated with reverse transcriptase [16],[17] yielded a total of 560 confirmed intron sequences.
The confirmed introns show a stronger signal, with 50.7% containing the branch point sequence ACTGAC (and 37.9% showing an extended branch point GACTGACG). The pattern becomes even more striking when intron sequences are divided along the lines of the previously reported genomic heterogeneity of O. lucimarinus, with roughly half of chromosome 2 differing from the rest of the genome in a variety of ways including much higher intron density [18]. Confirmed chromosome 2 introns show very weak branch point signal, with only 4.6% sharing the same sixmer (CTGACG), while 87.2% of introns outside of chromosome 2 contain ACTGAC (Table 3), and 66.5% contain an extended GACTGACG motif. Confirmed introns outside of chromosome 2 also show very strong 5′ splice sites consistent with intron-poor structures (4.8 bits), whereas confirmed chromosome 2 introns show much weaker boundaries (1.9 bits) (Table 3). Notably, 5′ splice boundaries outside chromosome 2 exhibit the atypical consensus GTGCGTG, whereas chromosome 2 introns prefer a more typical GTRNGT.
In contrast to O. lucimarinus, predicted introns in O. tauri show far lower conservation of branch points (Table 2). The lack of EST sequences for this species makes confirmation difficult, however a few putative intron sequences identified by TBLASTN searches of intron-containing non-chromosome 2 O. lucimarinus genes against the O. tauri genome exhibited sequences similar to O. lucimarinus introns, suggesting that O. tauri may exhibit a similar pattern. Given this seeming discrepancy between all predicted and confirmed O. tauri introns, we excluded O. tauri from the analysis.
Next, we studied acceptor sequence conservation within genomes. The 3′ sequences of spliceosomal introns generally show more similarity across species, with most species showing a terminal YAG, sometimes preceded by a poly-pyrimidine tract (Figure 1), although most fungal species and some others lack the poly-pyrimidine tract [19]. A survey of 50 widely diverged eukaryotic species showed 5 clear exceptions to this pattern. Three species, the protists T. vaginalis and G. lamblia, and Yarrowia lipolytica, a hemiascomycetes fungus (relatives of the baker's yeast S. cerevisiae), show strikingly similar patterns (Figure 4). The three species lack the 3′ polypyrimidine tract, and show clear anchoring of the branch point site at a specific number of basepairs away from the 3′ terminal AG (branchpoint-AG (BP-AG)) distance; 2 nt (AC) in Y. lipolytica, 4 nt (ACAC) in T. vaginalis and G. lamblia; Figure 4).
Non-chromosome 2 O. lucimarinus introns show a preference for CGCAG, though notably weaker than that found in the in the other three species (Figure 4). Interestingly, O. lucimarinus introns' conserved 3′ boundaries are associated with conserved BP-AG distance, as branch points for confirmed non-chromosome 2 O. lucimarinus introns show a broad peak ranging from around 20–35 bp (data not shown). It is interesting then that both O. lucimarinus and the constrained BP-AG distance species prefer a C at position –5 and a R at position –4. We were unable to find an explanation involving snRNA sequences for this preference. Finally, C. elegans introns also show stronger 3′ consensus, matching TTTCAG with nucleotides -6 and -5 significantly more conserved (Figure 4), as had been previously shown [20].
To better understand this pattern, we next studied available relatives of Y. lipolytica. We studied all six hemiascomycetes species with full genome annotations, as well as three additional Candida species for which some intron sequences were available. The species show pronounced differences, with three (including S. cerevisiae) showing large variations in BP-AG length and five species showing clear BP-AG constraint (differences in BP-AG length across hemiascomycetes was previously reported in [10]). In Debaryomyces hansenii, 65% of introns show a BP-AG distance between 6 and 8, and 88% of introns have a BP-AG distance between 5 and 10 nt. 75% of introns in Eremothecium gossypii with well defined branch points have BP-AG between 6 and 9 nt, with 66% between 6 and 11. The small number of available introns in C. lusitaniae and C. guilliermondii suggested preferred BP-AG distances of 4–5 and 3 nt, respectively. This BP-AG constraint could partially reflect differences in intron lengths, as mean/median lengths are lower for some of these species across the clade (Figure 5). However, the species with the clearest pattern of constraint, Y. lipolytica, has rather long introns relative to the other species.
Intriguingly, in E. gossypii introns, sequences between the BP and intron terminus varied considerably based on the BP-AG distance. The 30 introns with a BP-AG distance of 6 nt (the shortest distance with more than a few introns) showed a strong sequence consensus at the 3′ end of the sequence, with 80.0% having a G at position -4 (compared to only 39.3% for other introns; p = 0.00004 by a Fisher's Exact test), and 50.0% having an A at position -5 (compared to only 22.0% of other introns; p = 0.002 by a Fisher's Exact test). However no clear general trend towards stronger boundaries for short BP-AG introns was observed: introns with a BP-AG distance of 7 nt did now show a stronger 3′ consensus than introns with larger BP-AG distances. Y. lipolytica did not show differences in strength of 3′ sequence consensus for different BP-AG distances.
Mapping of BP-AG distances across hemiascomyces shows a complex phylogenetic pattern (Figure 5). The five species with strong BP-AG constraint are intermingled on the tree with the three less constrained lineages, suggesting convergent evolution of BP-AG constraint. Importantly, the “preferred” BP-AG distance varies across species – for Y. lipolytica, the most common BP-AG distance is 2 nt, compared to 3 in C. guilliermondii, 4–5 for C. lusitaniae, 6–8 for E. gossypii and 7–8 for D. hanseii.
It seems unlikely that a species with a strong preference for a certain BP-AG distance would convert to a different distance, since this would require indels of a very specific length occurring in dozens of already constrained introns. It seems more likely that this condition reflects ancestrally relatively unconstrained BP-AG distance, and convergent evolution of constrained BP-AG distances in different lineages.
Convergent evolution of retained intron sequences in intron-poor species is likely due to a combination of two factors: preferential retention of introns with consensus-like (i.e. strong) sequences and change of retained intron sequences to consensus boundaries. However, the relative impact of these two factors is unknown. We attempted to address this issue by identifying introns in intron-rich species that were present at the exact homologous position to introns in any available intron-poor species, and thus are likely to be ancestral to both species. If strong consensus sequences in intron-poor species are due to differential retention of introns with conserved ancestral sequences, it is possible that orthologous introns in intron-rich species could retain some of this signal. For each intron-rich species (the apicomplexan Toxoplasma gondii, H. sapiens, S. pombe, A. fumigatus, and A. thaliana), we compared 5′ splice site strength and branch point conservation between introns putatively orthologous to introns retained in at least one intron-poor species and the total set of introns in these species. Significant differences for both intron structures were found for A. fumigatus and T. gondii introns (Table 4).
Despite this analysis being perhaps the most direct way to test the hypothesis of preferential intron retention available, it is deeply undermined by the great phylogenetic distances between intron-poor species and even their closest relatives (T. gondii-C. parvum and A. fumigatus-hemiascomycetes diverged both many hundred million years ago) and associated large amounts of sequence change. Therefore, the finding of a positive signal for any of these comparisons is surprising and intriguing. To further test whether the observed stronger intron consensus sequence signal in intron-poor organisms could truly reflect retained greater boundary strengths from ancestor, we further divided the orthologous set into those introns shared with a more closely related intron-poor species (Y. lipolytica for A. fumigatus; C. parvum for T. gondii) and those shared only with distantly related species. Whereas the former could conceivably retain some specific ancestral signal due to lack of change, the second set represent divergences dating back upwards of a billion years ago to early eukaryotic evolution, seemingly precluding similarities in trends across individual intron boundary strengths representing lack of sequence change since that time. Unexpectedly, we observed that this second subset of introns, shared only with older relatives, show stronger signal than those introns shared with the closest intron-poor ancestor in T. gondii (Table 4). This suggests that boundaries with greater strength in intron-poor species does not reflect retained ancestral signal. (There was an insufficient number of introns in the A. fumigatus distantly-related group for comparison).
Another argument also argues that these intron subsets' stronger boundaries reflects not retention of ancestral boundary strength, but something else: A. fumigatus and Y. lipolytica exhibit different 5′ splice site consensus sequences, thus while A. fumigatus introns with homology to retained Y. lipolytica introns do show greater homogeneity in 5′ splice site boundaries (matching the consensus GTAAGT), they do not more closely resemble Y. lipolytica boundaries (consensus GTGAGT). Indeed, we observe the opposite trend: only 15.2% (26/171) of the A. fumigatus introns shared with Y. lipolytica have a G in position +3, whereas in the whole set of A. fumigatus introns, 21.4% have a G in that position.
Finally, we studied the distribution of characteristic intronic polyT motifs along intron length. For each species, we calculated frequencies of intronic minimal “polyT motifs” (following previous studies, we define these as six consecutive nucleotides containing at least 3 T's and no A's [19],[21],[22]) as a function of distance from the acceptor site.
Almost all species conformed to one of 3 broad patterns, which tend to be conserved within large phylogenetic groups (Figure 6). For the most common distribution (found in metazoans, plants, most apicomplexa and the heterokont Phytophtera species) polyT motifs concentrate near the intron terminus (Figure 6A). The 5′ limit of the distribution is likely determined by branch point position in some species (∼30 nt, similar to mean BP-AG distance 31.5 nt in mammals [23] and 27.6 nt in Arabidopsis thaliana [24]). In other species (Caenorhabditis elegans, Ciona intestinalis, Drosophila melanogaster, and the apicomplexan Theileria parva) polyTs are concentrated in the last ∼10 nt, and are underrepresented 10–15 nt from the terminus. This pattern could suggest a more 3′ branch point position, although branch points in these species are difficult to determine [25]. The rhizarian Plasmodiophora brassicae seems consistent with this broad pattern; however, the small number of available introns renders confident conclusions difficult.
Second, in most fungi, polyTs are roughly equally common across the intron (with the exception of the position of the branch point site, which typically falls ∼13–25 nt from the terminus [19]) (Figure 6B). This pattern resembles that found in some intron-poor species, including the T-rich introns of C. parvum as well as the more moderate T-rich introns of hemiascomycete yeasts (Figure 6C). S. cerevisiae shows a partial exception to the pattern, with a pronounced peak ∼10 nt from the terminus.
The third pattern is found in the two fully-sequenced amoebozoans (Dictyostelium discoideum and Entamoeba histolytica) and in the intron-poor fungus Ustilago maydis (Figure 6D). This pattern shows a single peak in polyT occurrence, centered between 15 and 40 nt from the acceptor site.
Although the numerous exceptions make firm conclusions difficult, the broad phylogenetic distribution of the first pattern (in animals, plants, apicomplexans and heterokonts, and perhaps rhizarians), suggests that the ancestral intronic structure had polyT motifs concentrated between the BP and the terminal AG, and that broader polyT distributions evolved early in fungal evolution.
We report convergent evolution of strong branch point consensus sequences and constrained branch points positions in eukaryotic lineages ranging from fungi to plants to protists. These observations join our previous findings of convergent evolution to strong 5′ splice site boundaries [9], as well as the pattern of recurrent nearly-complete intron loss, as examples of convergent intron-exon structure evolution across eukaryotes [26]. Interestingly, these five patterns appear to be closely related. Those lineages that are highly derived in intron number, having lost most of their introns, are the same ones that exhibit constraint of their few remaining introns' sequences.
However, different intron sequence characteristics show different degrees of co-evolution with intron number. Whereas strong splice sites show a one-to-one correspondence with low intron number across species [9], only a subset of intron-poor lineages have strong branch point sequences, of which only a subset have highly constrained branch point positions. Thus while intron paucity may be necessary for the emergence of branch point sequence and position constraint, it is not sufficient.
Difference in levels of intron structure constraint likely is associated with (even perhaps driven by; see below) changes in the spliceosomal machinery that have led to increased requirements for adherence to consensus sequences [9],[27]. Indeed, the intron-poor species whose splicing apparatus has been most extensively studied, S. cerevisiae, shows considerable alterations in the mechanisms and protein components of its spliceosome [2],[19]. Future work should explore spliceosomal changes in other intron-poor lineages, in particular the possibility in evolutionary convergence in spliceosomal machinery across lineages.
Notably, we failed to find intermediate stages. 5′ splice site strength shows a clear gap between intron-poor lineages (at least 5 bits of information content), and intron-rich (1–4 bits) [9]. Almost all species have either clear and strong branch point consensus (66–100% introns with same branch point hexamer) or much weaker conservation (<30%). Branch point position also seems bimodal, as clearly seen among the hemiascomycetous yeasts, where either >80% of branch points fall within a few base pairs, or fewer than 40%.
For 5′ splice sites, this lack of intermediates is consistent with some qualitative difference in the selective regimes acting within intron-poor and intron-rich species, leading to a lack of intermediate strengths. For branch point sequences and positions the case is more subtle. Do weak branch points in some intron-poor lineages reflect an ongoing process, or are these lineages somehow refractory 3′ intron convergence? Repeated evolution of constraint in hemiascomycetes may suggest an ongoing process, however in this case we might expect to observe intermediate stages. Possibly, once put in motion, intron structure constraint proceeds rapidly, which could explain the lack of observed intermediates.
Widespread sequencing has underscored the complexity of eukaryotic genome structure. While some genomes seem generally complex (with large numbers of genes containing numerous long introns and ubiquitous transposable elements) or simple (with short intergenic regions flanking a modest complement of nearly intronless genes), many genomes defy such straight-forward characterization. Intron-exon structures provide a clear example: the three genomes with the shortest known intron structures each have relatively high intron densities (Paramecium tetraurelia, the nucleomorph of B. natans, and Dicyemids, so-called mesozoans), whereas introns in very intron-poor species are not particularly short (Table 2).
Interestingly, these two classes of reduced lineages appear to show opposed patterns of intron evolution. Whereas intron-poor lineages tend towards highly-constrained intron sequence elements, short-intron lineages seem to show very weak sequence constraint. Available Dicyemid introns give the weakest known score for 5′ intron boundaries (0.5 bits) and 5′ splice sites of P. tetraurelia and B. natans are largely restricted to GT(A). These three lineages also show no signature of branch points (Table 2). This does not simply reflect an inability of short introns to accommodate branch points or reduced splicing constraints associated with short introns per se: E. cuniculi introns (35.8 nt on average) and many T. vaginalis introns (∼25 nt) are short, yet both show conserved 5′ splice site and branch point sequences (note that this also suggests that species with both types of genome reduction exhibit strong consensus, reflecting their intron paucity). This pattern underscores the importance of intron number, and not simply genome reduction, in driving the emergence of strong consensus sequences.
The finding of a general inverse correspondence between intron number and splicing signals' strength is unexpected and remains unexplained. Previously, we suggested that in intron-poor species, selection against aberrant splicing of cryptic splice sites would drive changes in the spliceosome towards stricter splicing requirements, which would in turn drive sequence change in (or loss of) non-conforming introns [9]. In intron-rich species, this evolutionary pathway would not be available since increased spliceosomal strictness would imply deleterious inefficient splicing of a much larger number of non-consensus introns [9].
The genome of the ultra-small green alga Ostreococcus lucimarinus provides a rare natural experiment to test this hypothesis. While genes in the majority of the genome exhibit very low intron densities, the genes spanning roughly half of chromosome 2 show a much higher density, well within that of “intron-rich” species [18]. That the two sets diverge so clearly in level of intron sequence constraint is clearly not predicted by general alterations of splicing strictness due to changes in a (assumed) single spliceosome.
One possibility is that the two intron sets are serviced by different spliceosomes (as is the case of U2 and U12 in different eukaryotic lineages). However, a computational search turned up only single copies of spliceosomal RNA components in congenitor O. tauri [28]. Conceivably snRNA changes to complement the divergent Ostreococcus 5′ splice sites and branch points (GTGCGTG and GACTGACG in O. lucimarinus) could have thwarted their identification in the previous study, and this possibility is worth exploring. Alternatively (though more difficult to test), a single core RNA splicing machinery could associate with different sets of protein components in distinct spliceosomes with different splicing activities.
More likely, however, O. lucimarinus contains a single spliceosome, strongly suggesting that the constrained introns through most of the genome, as well as those of intron-poor species that they so closely resemble, do not reflect inherent changes in the spliceosomal machinery. A simpler possibility is that differences in local (in O. lucimarinus) or cellular (in other species) concentrations of spliceosomal complexes is the driving factor. It seems possible or even likely that spliceosomal complexes in intron-poor species are downregulated. Such downregulation could reflect either selection to reduce incorrect splicing of truly exonic sequence (i.e. fewer spliceosomes, less chance of false splice boundaries being spliced), or could be favored by reducing energetic costs associated with transcription, processing, and translation of spliceosomal components. If so, the lowered concentration of spliceosomal components would require stronger binding affinity of individual splice sites to corresponding snRNAs for efficient splicing, which would in turn drive the evolution of stronger boundaries (or intron loss). Differential local concentrations across genomic regions in Ostreococcus could be maintained if spliceosomes were preferentially recruited to the intron-rich genomic region. This scenario is similar to our previous hypothesis in invoking a tradeoff between the costs of efficient splicing of weak boundaries (maintenance of high spliceosome concentration) and the costs of mis-splicing, which we argue would likely be different in intron-rich and -poor species. This hypothesis makes the testable prediction that spliceosomal components should show reduced expression in intron-poor species relative to intron-rich species.
Another hypothesis concerning the concentration of spliceosomal complexes, suggested to us by Tony Russell, sees a very different role for selection. A predicted consequence of increasing the length of snRNA-intron element base-pairing interactions is a reduction in the overall rate of splicing, simply due to a tighter association between intron and spliceosome. Such a decrease in splicing rate will be tolerated in intron-poor species if spliceosomal components are in excess relative to the number of introns. However, in intron-rich species spliceosomal components may not be in excess, in which case stronger base-pairing between intron and snRNAs could be disfavored. Notably, these two hypotheses make qualitatively different predictions. Whereas the latter hypothesis predicts that strong boundaries would be disfavored in intron-rich species, the former predicts that they would as or more fit than weaker boundaries. Comparative analysis of closely related species to test these predictions is underway.
Two factors could drive evolutionary convergence to strong intron boundaries: sequence changes in existing intron sequences to consensus sequences, and preferential loss of non-consensus introns. The relative contributions of these factors may depend on the precise evolutionary pathway from (ancestral) genomes with many introns with weak boundaries and relatively lax splicing requirements to fewer introns with strong boundaries and stricter requirements.
First, widespread (mostly random) intron loss could lead to selective conditions favoring the evolution of a spliceosome with stricter sequence requirements for splicing (as argued above and in reference [9]). Introns with non-consensus sequences would then impose a burden, which could be resolved by sequence change or intron loss. If intron loss rates in these lineages are at least comparable to substitution mutation rates (for instance, 90% intron loss over 500 million years is consistent with a constant loss rate of 5×10−9 per year, comparable to some estimated mutation rates [29],[30]), preferential loss could play an (or even the) important role in convergence. In this case, intron loss would be a self-catalyzing process, with widespread intron loss leading to increased splicing requirements driving yet faster intron loss.
Alternatively, increased splicing requirements could come first, driving intron loss (consistent with [31]). However evolution of stricter splicing requirements in intron-rich organisms, where there are large numbers of non-consensus introns, would lead to widespread deleterious mis-splicing. Thus it is hard to imagine such strict requirements arising prior to widespread intron loss.
Finally, even under lax splicing requirements, loss of non-consensus introns could be more highly selected due to the effects of the less efficient splicing of these introns (e.g. [32]), with stricter splicing requirements gradually enabled by loss of non-consensus introns. The viability of this scenario depends on significantly higher selective costs of suboptimal boundaries in intron-rich species respect to the optimal introns. Attempts to estimate these selective costs by comparative analysis are underway.
In any case, it seems likely that intron loss, increased splicing constraints, and intron sequence change will all be reinforcing of one another, such that (perhaps past some critical threshold), the three will proceed in tandem. Differences in the relative contributions of the three phenomena will depend on mutation rates and selective coefficients for different kinds of changes (basepair substitutions, intron loss, spliceosomal changes), which may vary considerably across times and lineages.
The convergent U2 intron structures observed here and elsewhere in some intron poor species – strong 5′ and branch point consensus sequences, constrained BP-AG length – are strikingly reminiscent of the U12-type intron structures found across a wide variety of lineages [33],[34]. One possibility is that the U12-type intron structures represent a derived state (possibly evolved in the ancestor of eukaryotes), and that these convergent cases have similar causes – that U12-type introns' low genomic number has subjected them to the same pressures as those experienced by U2-type introns in intron-poor lineages. Notably, if the conservation across lineages of specific conserved U12 intron sequence elements, as opposed to the differences in consensus structures observed in U2-type introns across diverse intron poor lineages reflects the emergence of these structures early on in eukaryotic history, this interpretation would imply that U12-type introns have been rare since very early in eukaryotic history. Alternatively, the similar structures of U12 introns in general and U2 introns in intron-poor species could have different explanations. One possibility is that the U12 system represents an intermediate between the type II introns that initially proliferated in early eukaryotic ancestors, with their highly similar sequences and structures, and typical highly degenerate U2 introns. In this case the persistence of strong consensus sequences in U12-type but not U2-type introns remains somewhat mysterious, as does the preponderance of U2 introns relative to U12.
We find different patterns of polyT motif distribution along the introns in different lineages, which likely reflect differences in polyT functionality and in how polyT-binding factors regulate splicing in these lineages. Indeed, the differential conservation and evolution of spliceosomal proteins binding polyT tracts (PTB, SXL, TIA1, Nam8, etc.) in each species is likely to determine the position of polyT motifs along introns and what role these motifs play in splicing regulation.
Among these locations, we find that the so-called “polyT tail”, spanning from a position likely corresponding to the branch point to the 3′ intron terminus, is common to a wide variety of groups ranging from plants to animals to various, widely diverged, protists, strongly suggesting that the existence of a polyT tail may be ancestral. Furthermore, we show that, as in the case of 5′ splice sites, strong branch point site consensus have evolved independently in only intron-poor species, whereas all intron-rich species have weak branch point sites. Since a wide variety of studies have shown that eukaryotic ancestors have harbored relatively high intron numbers [3], [35]–[38], our results suggest that the eukaryotic ancestors also had weak branch point site consensus, as with most modern eukaryotic groups.
These conclusions extend an excellent recent study from Schwartz et al. [27], who studied BP and polyY motifs in 19 opisthokonts and 3 other eukaryotic species, and reached similar conclusions about ancestral intron structures. Our inclusion of a more diverse set of species spanning all of known eukaryotic species allows us to reach deeper into eukaryotic evolution, potentially getting much closer to the initial origin of spliceosomal introns. In particular, we find that sequences from the first characterized rhizarian, as well as for various heterokonts, follow the patterns found in other kingdoms. This striking similarity over very broad evolutionary distances significantly strengthens our conclusions about ancestral eukaryotic splicing, rendering them independent of the placement of the root of the eukaryotic phylogeny.
The presence of these intronic features, polyT tail and weak branch point sites, in the eukaryotic ancestors adds to the developing picture of the spliceosomal system in early eukaryotes – with highly developed spliceosome, weak 5′ and branch point sequences, a polyT tail and complex splicing patterns [2],[9],[39],[40].
One limitation of the data deserves comment. We analyze annotated intron sequences, making our analysis subject to the quality of available annotations. Problems with the annotations may very directly influence the characteristics studied here, since for instance more consensus-like sequences are more likely to be identified as introns. Such concerns may also affect comparative analysis if the annotation efforts for different species are differentially sensitive to different kinds of introns.
However, it is very unlikely that such limitations are likely to drive the qualitative differences we see here. For a “weak” species with for instance 30% of its true introns exhibiting the same branch point to be incorrectly identified as a “strong” species (with, say, 80% of predicted introns with the same branch point) would require that the vast majority of its non-consensus introns have gone unannotated. For the reverse to occur, it would be required that there were so many falsely predicted introns that it had drowned out the signal almost entirely. Thus, while it is important to point out that our results are likely not accurate to the second decimal place due to problems with annotations, such problems are very unlikely to be driving the large qualitative differences observed.
Possibly, the most important impact of annotation errors could be to reduce the signal in very intron-poor species. For instance, further scrutiny suggested that 2 of the 15 predicted introns in E. cuniculi may not in fact be introns at all: both are a multiple of 3 basepairs and lack inframe stop codons; these two introns have the weakest 5′ boundaries (matching the consensus (GT)AAGT at 1 and 2 out of 4 positions, compared to at least 3 matches for the other 13 introns); and one intron has similar sequences at the two boundaries, suggesting that this intron prediction could reflect a reverse transcriptase artifact in EST preparation [17]. Notably, in our previous work on donor splice sites [9], E. cuniculi represented the only intron-poor species lacking very clear strong boundaries. Excluding these two questionable introns, E. cuniculi has a donor site information content of 6.2 bits, comparable to the other intron-poor lineages.
These results attest to plasticity of spliceosomal intron structures through the history of eukaryotes. The availability of large numbers of eukaryotic genomes now allows comparative analysis of an increasing diversity of genomic structures. Present and previous works have provided an increasingly detailed picture of the patterns and determinants of intron-exon structures, one of the hallmarks of eukaryotic genome organization. Definitive identification of the causes of highly regular 3′ intron structures awaits the identification of additional lineages exhibiting this pattern.
GenBank fully-sequence genome annotations were downloaded from NCBI webpage (http://www.ncbi.nlm.nih.gov) or Ensembl database (http://www.ensembl.org) for 7 metazoa: human (Homo sapiens (NCBI 36 Ensembl 38.36)), zebra fish (Danio rerio (Zv5 Ensembl 38.35e)), Strongylocentrotus purpuratus (AAGJ00000000.2), Drosophila melanogaster (release 4.1), mosquito (Anopheles gambiae (AgamP3 Vectorbase 38.3a)), Caenorhabditis elegans (WS150 Wormbase 38.150a), Ciona intestinalis (CINT1.95); 16 fungi: Aspergillus fumigatus Af293 (AAHF00000000.1), Aspergillus nidulans FGSC A4 (AACD00000000.1), Neurospora crassa OR74 A (AABX00000000.1), Gibberella zeae PH-1 (AACM00000000.1), Yarrowia lipolytica CLIB122 (CR382127-31.1), Saccharomyces cerevisiae YJM789 (AAFW00000000.1), Candida glabrata CBS138 (provided by J. E. Stajich), Candida albicans SC5314 (provided by J. E. Stajich), Kluyveromyces lactis NCYC 2644 (AADM00000000.1), Eremothecium gossypii ATCC 10895 (AE016814-20.1), Debaryomyces hansenii CBS767 (CR382133-9.1), Schizosaccharomyces pombe 972h (AL672256-8.1), Ustilago maydis 521 (AACP00000000.1), Cryptococcus neoformans B3501-A (NC_006670, NC_006679-87, NC006691-4), Encephalitozoon cuniculi GB-M1 (AL391737.1, AL590442-50.1), Rhizopus oryzae RA 99-880 (AACW00000000.2); 2 amoebae: Dictyostelium discoideum AX4 (AAFI00000000.1), Entamoeba histolytica HM-1:IMSS (AAFB00000000.1); 2 plants: Arabidopsis thaliana (NC_003070.5, NC_003071.3, NC_003074.4, NC_003075.3, NC_003076.4), Oryza sativa (Build 2.1); 3 green algae: Chlamydomonas reinhardtii (JGI Chlamy 3.0), Ostreococcus lucimarinus CCE9901 (GenBank version 1, CP000581- CP000601), Ostreococcus tauri OTH95 (GenBank version 1, CR954201- CR954220); 6 aplicomplexans: Plasmodium falciparum HB3 (AANS00000000.1), Plasmodium yoelii yoelii (AABL00000000.1), Plasmodium chabaudi (CAAJ00000000.1), Theileria parva Muguga (AAGK00000000.1), Cryptosporidium parvum Iowa (AAEE00000000.1) and Toxoplasma gondii RH (AAQM00000000.1); and 2 heterokonts: Phytophthora ramorum strain Pr102 (AAQX00000000.1) and Phytophthora sojae strain P6497 (AAQY00000000.1).
All reported intron sequences were obtained from published supplementary material for: Guillardia theta NM [41], Cyanidioschyzon merolae 10D [42], Trichomonas vaginalis [5],[16] and Bigellowiella natans NM [8] and Giardia lamblia ATCC 50803 strain WB C6 [43] (plus 2 introns we have identified), the mesozoan Dicyema acuticephalum [44], and the rhizarian Plasmodiophora brassicae [45].
Available gene sequences for unpublished or incomplete genomes were downloaded one by one as CDS annotations from NCBI web page (http://www.ncbi.nlm.nih.gov/): Brassica oleracea (1096 introns in 314 genes). Introns for Paramecium tetraurelia were extracted from nucleotide links of NCBI taxonomy (1082 introns in 401 genes). Introns for Candida guilliermondii (13 introns), Candida lusitaniae (10 introns) and Candida tropicalis (34 introns) were provided by J. E. Stajich.
Studying the clear branch point consensus from a wide variety of intron-poor species, we first define an extended branch point consensus, WCTRAYN, consistent with the minimal consensus NYYNAN described for a wide variety of eukaryotic groups [19],[23],[24],[46].
For each 50 species, we next studied the percentage of introns showing the most common hexamer matching this extended consensus, allowing two-fold degeneracy at zero, one, two and three sites of the putative branch point hexamer. The four measures are complementary.
We did not aim to identify and study branch points in all introns. Instead, the percentage of use of the most common motifs gives a straight-forward measure of the strength of the signal for a given species. The use of a similar approach to measure the strength of the 5′ splice site (whose definition is trivial) shows a clear correspondence between the measure of strength as the percentage of introns with the most common sequence motif (i.e. GTAAGT, GTATGT, etc.) and as information content, used broadly in the literature, with a coefficient of correlation between both variables for the species included in this study is r = 0.96.
We aligned the final 20 nt of each intron for each species using WebLogo (http://weblogo.berkeley.edu/logo.cgi). To better characterize the evolution of BP-AG constraint in Y. lipolytica, we further studied BP-AG distance in the 9 hemiascomycetes species. In all hemiascomycetes species, the vast majority of introns contained a single TACTAAC sequence, used as the branch point. The BP-AG distance was defined as the number of base pairs (Nn) for 5′TACTAAC|Nn|AG-exon3′.
For 5′ splice sites we used a similar methodology as described in [9]. The first 6 bases of each intron were extracted, and information content in bits for positions +3 to +6 was calculated using PICTOGRAM software online (http://genes.mit.edu/pictogram.html).
We downloaded available EST sequences for O. lucimarinus from NCBI on April 4th, and performed standalone BLASTN searches for each predicted intron-containing O. lucimarinus gene against the ESTs. Preliminary confirmed introns were identified as those in which an EST hit with >60 bits and with >90% sequence identity spanned the intron position (reaching at least 3 nt on each side of the intron position). Each of these introns was then analyzed by eye to exclude non-canonical intron positions as well as those for which sequence similarity between the regions spanning the 5′ and 3′ predicted splice boundaries were consistent with a template-switching artifact during the reverse transcription step of the EST library preparation [16],[17].
For each species considered, databases of intron/exon structures of predicted gene transcripts were prepared from the genome annotations. Homologs were identified by one-way BLASTP searches of intron-containing genes from intron-poor species against predicted proteomes from intron-rich species. Putatively orthologous introns were identified as those present at identical alignment positions (position and phase) in both species. Related species were defined as: C. parvum for T. gondii; Y. lipolytica (chosen as the representative hemiascomycete due to its greater intron density (3 times higher than S. cerevisiae), in order to increase sample size) for S. pombe and A. fumigatus; and the C. merolae and the G. theta nuclemorph for A. thaliana. Due to the lack of related intron-poor species, H. sapiens introns were divided into only two groups.
We used the minimal definition for polyT motif, defined as six consecutive nucleotides containing at least 3 T's and no A's [19],[21],[22]. The study of introns for each species was performed using custom Perl scripts. The last 2 and the first 10 base pairs of each intron were excluded.
|
10.1371/journal.pcbi.1002938 | A Simple Iterative Model Accurately Captures Complex Trapline Formation by Bumblebees Across Spatial Scales and Flower Arrangements | Pollinating bees develop foraging circuits (traplines) to visit multiple flowers in a manner that minimizes overall travel distance, a task analogous to the travelling salesman problem. We report on an in-depth exploration of an iterative improvement heuristic model of bumblebee traplining previously found to accurately replicate the establishment of stable routes by bees between flowers distributed over several hectares. The critical test for a model is its predictive power for empirical data for which the model has not been specifically developed, and here the model is shown to be consistent with observations from different research groups made at several spatial scales and using multiple configurations of flowers. We refine the model to account for the spatial search strategy of bees exploring their environment, and test several previously unexplored predictions. We find that the model predicts accurately 1) the increasing propensity of bees to optimize their foraging routes with increasing spatial scale; 2) that bees cannot establish stable optimal traplines for all spatial configurations of rewarding flowers; 3) the observed trade-off between travel distance and prioritization of high-reward sites (with a slight modification of the model); 4) the temporal pattern with which bees acquire approximate solutions to travelling salesman-like problems over several dozen foraging bouts; 5) the instability of visitation schedules in some spatial configurations of flowers; 6) the observation that in some flower arrays, bees' visitation schedules are highly individually different; 7) the searching behaviour that leads to efficient location of flowers and routes between them. Our model constitutes a robust theoretical platform to generate novel hypotheses and refine our understanding about how small-brained insects develop a representation of space and use it to navigate in complex and dynamic environments.
| Pollinating bees, along with bats, hummingbirds, rodents and primates, typically develop circuits (traplines) to visit multiple foraging sites in an efficient stable sequence. The question of how animals encode and process spatial information to develop these impressive foraging patterns remains poorly understood. Previously we showed that an iterative improvement heuristic model of bumblebee traplining can replicate the establishment of stable routes by bees between flowers distributed over several hectares. Here we tested the model against a variety of datasets with different configurations of flowers and found it to give good agreements with all these observations. We have thus shown how these complex dynamic routing problems can be solved by small-brained bees using simple learning heuristics and without acquiring a ‘map-like’ memory. The proposed heuristic shows how bees develop optimal routes simply by following multi-segment journeys composed of learnt flight routines (local vectors), each pointing towards target locations (flowers) and coupled to a visual context (landmarks or panoramas). Such a decentralized representation of space relying on learnt sensorimotor routines is akin to ‘route-based’ navigation as described in desert ants, where spatial information is thought to be processed by separate, potentially modular, guidance systems.
| Bees, bats, hummingbirds, rodents and primates which exploit patchily distributed foods that replenish over time often visit resource locations in predictable sequences [1]–[12]. In pollinating insects, such as bumblebees, these traplines are often the shortest circuits to visit all the known flower locations exactly once before returning to the nest and so are solutions of the well-known travelling salesman problem (TSP) [13]. Just how these animals solve this problem with relatively low computational power has long been a mystery [14]–[16]. The TSP is, after all, one of the most intensively studied problems in combinatorial optimization [13]. There are no efficient algorithms for even solving the problem approximately (within a guaranteed constant factor from the optimum) because the problem is NP-complete (nondeterministic polynomial time complete) and it is believed that there is no algorithm that can find a solution where the processing time increases as a finite order polynomial in N [17]. The most direct approach would be to try all of the permutations and then select the shortest one, but this becomes impractical even for only 20 locations as the number of permutations is 20!. Nonetheless, approximate solutions can be found using linear programming methods, neural networks, simulated annealing and genetic algorithms [17]. The best approximate algorithms can typically find solutions within 1–2% of the optimum but these are unlikely to be implemented by biological organisms because they are computationally demanding [13].
Several algorithms have been proposed to explain how animals might optimise multi-location routes [18]. Perhaps the simplest candidate model of bumblebee trapline development is the ‘nearest neighbour’ or ‘greedy’ heuristic, in which a model bee chooses the nearest unvisited flower as its next move until all flowers have been visited. It has been suggested that this simple heuristic explains the routing behaviour of some animals [14], [19]–[21] but it is incompatible with observations of bumblebees foraging at various spatial scales [15], [16]. No better is a simple random ‘k-opt’ iterative improvement heuristic [22] in which a model bee (1) tries to improve the route between known flowers by randomly shuffling the order in which a number (k) of randomly selected flowers are visited, and (2) the route change is kept if the new route is shorter than the previous one (otherwise it is rejected). This heuristic significantly over-predicts the number of foraging bouts executed before the first appearance of an optimal (shortest-path length) foraging route and unlike bumblebees does not create stable traplines [16]. Recently we proposed that bumblebees use a simple learning heuristic (‘The Basic Traplining Heuristic Model’) to develop optimal traplines between distant feeding locations in the field. This heuristic is based on our general knowledge of bee navigational strategies [23], including bees' tendency to discover flowers in relation to their distance to the nest [16], the fact that they learn sequences of vector flights between familiar locations using the visual context (landmarks and/or panoramas) [24]–[26], and their ability to measure travel distances through the image movement over their retina (optic flow) experienced during flight [27], [28]. In this heuristic, model bees try a limited number of possible route iterations, so that route segments (between pairs of flowers) that shorten the overall route are reinforced in memory, while others are abandoned, allowing bees to develop an adaptive (and occasionally optimal) ‘trapline’ whilst retaining some ability to adjust their route in response to changes in the spatial configuration of flowers [16]. This model predicts that bees: (1) occasionally visit fewer than all flowers especially during early bouts; (2) regularly revisit empty flowers during the same bout; (3) decrease their frequency of returns to just-visited, empty flowers with experience; (4) establish stable optimal routes in some spatial configurations but not others; (5) can sequentially adjust their routes to incorporate newly discovered flowers in an optimal way when the number of locations is relatively small. Quantitative evaluation of the simulated data with bees' optimisation performances at an array of five artificial flowers arranged in a regular pentagon (50 m side length) set up in the field showed full agreement (as quantified by p-values for the probability of the data given the model) for the number of bouts: (1) to the first appearance of an optimal sequence; (2) the number of bouts to the stabilization of an optimal sequence into a trapline; (3) the number of different routes experienced; (4) the net travel length per bout; (5) the number of revisits per bout: and (6) the similarity indices between successive bouts [16]. In accordance with empirical data [16] the model also predicts correctly that bee flight paths are constrained by previous experience and that bees cannot compute entirely novel solutions quickly. Our simple model relies on reactive navigation rules rather than a “cognitive map” and might require relatively low cognitive demands. It may therefore provide an important indication of how bumblebees encode and use spatial information when developing traplines. Nonetheless, detailed analyses of the model are necessary to refine our understanding of this strategy and clarify whether similar learning heuristics apply to bumblebees foraging at different spatial scales and configurations.
In this paper we show that our model is consistent with all published observations [14]–[16], [29]–[32] made at small spatial scales that have established how bumblebees optimize their routes between fixed resources [15], [29], re-optimize their routes after identifying a new resource [31] and how and when bees prioritize high-reward resources [32] in various floral arrays (Fig. 1). These studies have been performed in flight rooms where bees could potentially see all artificial flowers from any vantage point. The dimension of these flight rooms varied. The study of Saleh and Chittka [29] was carried out in an indoor flight arena measuring 105 (L)×75 (W)×30 (H) cm. The studies of Ohashi et al. [14] were carried out in an indoor flight cage (788×330×200 cm). The studies of Lihoreau et al. [15], [16], [31], [32] were carried out in a greenhouse (870×730×200 cm). We begin by showing that the model can account for the observed increasing propensity of bees to find optimal routes with increasing spatial scale and show that is predicts correctly the formation of stable optimal traplines for some arrangements of flowers but not others [16], [32]. We then show that our model predicts that bumblebee flight patterns made during the course of a day (between 65 and 80 foraging bouts) between 10 or fewer irregularly distributed flowers will often converge onto the shortest possible path or find good approximate solutions of it. This is an impressive feat because there are different ways of travelling between 10 flowers. We also show that the model predicts that after locating more flowers than necessary to fill their crop capacity (nectar stomach size), bumblebees can develop highly effective traplines, by visiting only a set of flowers with an appropriate spatial configuration. Finally, we show that the bee searching behaviour is consistent with their adopting an optimal searching strategy.
The basic model – an iterative improvement heuristic - is described in Lihoreau et al. [16]. The heuristic mimics the behaviour of a bumblebee collecting nectar in a stable array of flowers and returns to its nest over multiple consecutive bouts. At the end of each foraging bout, flowers replenish with a new load of nectar. At each stage, a model bee chooses to move between flowers according to six assumptions: (1) the bee can uniquely identify each flower using information from path integration and/or the visual context (landmarks, panoramas) [24], [25]; (2) the bee has a finite probability of using transition vectors joining each pair of flowers; (3) the initial probability of using a vector depends on the distance between the two flowers (in our simulations these probabilities are inversely proportional to the squared distance between flowers and are normalized with respect to all flowers); (4) the bee computes the net length of the route travelled using optic flow (odometer) [27], [28], by summing the distances of all vectors comprising the flower visit sequence; (5) having completed a route passing through all the flowers at least once (and thus filled their crop capacity), the bee compares the net length of the current route to the net length of the shortest route experienced so far that passes through all the flowers; (6) if the new route is no longer, the probabilities of using the vectors forming this new route in the next foraging bout are multiplied by a common factor and then all probabilities are rescaled with respect to all flowers so that they sum to unity. Repeating the shortest route therefore reinforces it.
The model was used to predict the distributions of the number of bouts before the first appearance of an optimal (shortest) route (if found) and the number of bouts before the optimal routes became established as the only foraging route stabilised. These distributions were based on 1000 runs of the model. These distributions were then used to calculate the probability of a real bee doing at least as well given the model (i.e. the null hypothesis) is correct, i.e., the numbers of bouts/routes were ordered and then the ranking of the real bee observation was determined. This probability is a p-value. A p-value of 0.3 means that the numbers of bouts/routes for the real bee was equal to the 70th % quickest result in the numerical simulations. The model can be rejected if the p-value is lower than 0.05. Typically p-values are much larger than this threshold.
Aside from comparisons with our own data [15], [16], [31], [32] we will also compare our model with the empirical study of Ohashi et al. [18]. For each pair of flowers (i, j), Ohashi et al. [18] recorded the numbers of transitions from flower i to flower j, and from flower j to flower i made during 9–10 successive foraging trips. These transition matrices were characterised by asymmetry indices , where P is the binomial probability of the observed departure from a 1∶1 expectation of the observed number of transitions, i.e. if there were N transitions between flowers i and j with n transitions being from flower i to flower j then the associated binomial probability. P values with fewer than 6 observations were omitted. The asymmetry indices were then standardized by dividing by the number of pairs tested, which varied with foraging stage and among the bees. This standardization was not mentioned in the paper by Ohashi et al. [14] (K. Ohashi, private communication).
Lihoreau et al. [32] reported on bee optimisation performances in an array of five artificial flowers arranged in a regular pentagon (5 m side length) in a flight room (Fig. 1a). All flowers had the same reward value and their spatial arrangement was similar to the one used in the field study of Lihoreau et al. [16]. However, unlike at the field scale, at these scales the bees could potentially detect all of the flowers visually from any location. Nonetheless, stable optimal routes (visiting each flower once and returning to the nest using the shortest possible path) only became established after bees had made 34 or more foraging bouts. This is significantly more than in the field experiment using a scaled-up arrangement of flowers with side length 50 m where around 26 bouts were required for the establishment of the optimal route [16]. This suggests that a bee's “motivation” to optimise its route increases with spatial scale because the costs of travelling suboptimal routes are lower when flying a few metres than when flying several hundred metres [16]. We tested whether this difference in tendency to optimise can be captured by the model by adjusting the common factor by which vectors are reinforced each time a short route is found (see Methods).
Good model agreement with the data collected at the field-scale was only obtained when the probability enhancement factor fell between about 1.5 and 4 [16]. Smaller probability enhancement factors, less than about 1.1, brought the model into good agreement with the data collected in the flight room by Lihoreau et al. [32] (Table 1). This suggests that the probability enhancement factor in our model is scale-dependent and can be associated with motivation to optimize a route. Similarly, good model agreement with the data collected in the same flight room with 4 flowers [31] and with 6 flowers that were between 1 and 10 m apart [32] and with data collected in a smaller flight arena with 6 flowers less than 1.0 m apart [29] was only obtained with probability enhancement factors less than 1.5 (flower arrangements shown in Fig. 1b–d).
We then examined how a bee's tendency to repeat visitation sequences increases with experience using a similarity index (SI), described in Saleh and Chittka [29], which quantifies the similarity between pairs of flower visitation sequences. SI takes into account the length of sequences and the order of visits to flowers. SI ranges between 0 (completely different sequences, e.g. 123 vs 456) and 1 (identical sequences, e.g. 12345 vs 12345). The model was in closest agreement with the observational data [29], [32] (i.e. the p-values were largest) when the probability enhancement factor was about 1.1 (Fig. 2). For example, for the case of Lihoreau et al. (Expt. 2, Bouts 1–40) [32], p-values ranged between 0.16 and 0.43 when the probability enhancement factor was 2.0 and ranged between 0.43 and 0.62 when the probability enhancement factor was 1.1
Ohashi et al. [14] reported on the ontogeny of foraging paths in 3 different spatial configurations of 10 flowers that were less than 1.0 m apart (Fig. 1e–g). In their ‘independent’ array, 10 flowers were arranged in a triangular pattern so that bees can choose distance and turning independently; in the ‘positive’ flower array, proximity and directionality were positively linked, so that the nearest neighbouring flower could be reached by straight-ahead movement; and in the ‘negative’ flower array, proximity and directionality were negatively linked, so that choosing the nearest neighbour flower as the next flower to visit required bees to make turns. Ohashi et al. [14] reported that bumblebees preferred to choose short distances over straight flights and showed little plasticity in this regard, and as a consequence are less able to approximate the TSP solution in a ‘negative’ flower array compared to other arrays. Ohashi et al. [14] reported that 3 out of the 6 bees tested in the positive array established optimal traplines, just 1 out of 5 bees tested in the independent array established an optimal trapline and none of 5 bees tested in the negative array established optimal traplines. Our model is consistent with these observations. After 65 bouts, the model predicts that about 80% of the bees will have established stable optimal traplines in the positive array; 10% of the bees will have established stable optimal traplines in the independent array; and no (0 out of 100) bees will have established stable optimal traplines in the negative array. The latter prediction arises because the initial probability of using a vector depends on the distance between the two flowers (in our simulations these probabilities are inversely proportional to the squared distance between flowers). In accordance with the observations of Ohashi et al. [14] our model predicts that stable optimal traplines cannot be established for all spatial configurations of rewarding flowers. This is true at the scale of the experiments (probability enhancement factor 1.1) and at the field scale (probability enhancement factor 1.5).
Ohashi et al. [14] reported that the asymmetry index increased with foraging experienced and for this reason they reported on median rather than mean values of the standardized asymmetry index. The median standardized asymmetry indices for the positive, independent and negative arrays were 5.47±1.10 (mean±s.e.), 4.37±0.37 and 4.56±0.56. Comparable model predictions (with overlapping ranges) are obtained when the probability enhancement factor is less than about 1.5. Model predictions for a probability enhancement factor of 1.5 are 5.65±2.69, 4.86±1.78 and 5.58±2.22.
Having demonstrated good agreement between our model and various datasets of the literature, we then used the model to predict the optimization performance of bees when foraging on randomly rather than regularly distributed flowers. The simulation data were obtained for 100 different random arrangements of N flowers, and for 100 bees per arrangement. The probability enhancement factor is two, as this brought the model into good agreement with the observations. The model predicts that the numbers of bees that find the optimal path between N randomly distributed flowers during the course of a day (65 foraging bouts) decreases as the number of flowers increases but remains sufficient even for 10 flowers (Fig. 3a,b). Nonetheless, some of the random arrangements of flowers form ‘negative’ arrays as proximity and directionality were negatively linked and in these cases no model bees found an optimal route. For other random arrangements of the flowers, almost all of the models bees found the optimal route. The average path length as a proportion of the minimum path length increases as the number of flowers increases but is less than 1.25, the value obtained using the nearest neighbour algorithm, when there are 10 or fewer flowers (Fig. 3c).
It is computationally prohibitive to test the optimality of routes made between 20 or more flowers. Nonetheless, most (about 98.5%) of the shortest routes made between 20 randomly distributed flowers that were found by the model bees during the course of a day (65 foraging bouts) could be shortened by simply switching the order in which two of the flowers were visited, i.e. by changing the flower visitation sequence 5875431… to 5835471… by switching the order in which flowers 3 and 7 are visited. The routes that could not be shortened in this way could be the shortest of possible routes between the flowers. After 2 days of foraging (130 foraging bouts) without overnight memory loss, about 96.5% of the shortest found routes could be shortened by such pairwise switching of the visitation sequences. Model bees that do not find optimal traplines gradually reduce the number of distinct routes taken between the flowers, but generally do not form stable non-optimal traplines during the course of a day (Fig. 3d).
Bees foraging on several distant patches are predicted to eventually minimize overall travel distances between patches but not necessarily travel distances within patches. Most model bees (with constant probability enhancement factors) foraging on 4 patches located at the corners of a square (Fig. 3e) did, for example, follow optimal clockwise or anticlockwise routes when flying between patches during the course of the day. Optimal clockwise or anticlockwise routes were flown about 70% less frequently within patches.
The aforementioned results together with the observations of Ohashi et al. [14] (3 different arrangements of 10 flowers in a small flight cage) suggest that a bee's ability to optimise its foraging route may depend largely on how it selects a set of flowers or patches of flowers to visit. If it has sufficient options, a bee might select a set of flowers or patches for which the route between them can be optimized. This tendency would be limited by the number of located resources and possibly because bumblebees avoid intensive overlap of their foraging areas with competitors [33], [34]. We used the model to predict the optimization performance of bees when the crop capacity is filled after visiting some but not all known flowers. When the crop capacity is filled, a model bee returns directly to the nest. The simulation data were obtained for 100 different random arrangements of the 8 flowers, and for 100 bees per arrangement. The probability enhancement factor was two. The model predicts that a significant proportion (≥20%) of bees can find the optimal route between a few known randomly distributed flowers during the course of a day (65 foraging bouts) (Fig. 4a). Irrespective of whether or not the optimal route is found, the model bees do tend to form stable traplines so that some flowers are repeatedly revisited during the day whilst others are largely neglected (Fig. 4b,c). In accordance with the observations of [29] [6 flowers in a small flight cage], the non-optimal routes are predicted to be dependent upon an individual's foraging history.
Lihoreau et al. [32] demonstrated that traplining bees trade-off between minimizing travel distance and prioritization of the most rewarding locations. After the introduction of a highly rewarding flower to the pentagon array, the bees re-adjusted their routes visiting the most rewarding flower first provided that the departure distance from the shortest route was sufficiently small (18%). However, when routes optimizing the initial rate of reward were much longer (42%), bees prioritized short travel distances. This behaviour can be captured qualitatively by the model by enhancing the initial value of the probability for flying between the nest and the highly rewarding flower. If there are more flowers than necessary to fill a bee's crop capacity and if flowers vary in their reward value, then the model bees tend to establish stable optimal traplines at the field scale and the most rewarding flowers are visited more frequently than are the least rewarding flowers. The tendency to prioritise the most rewarding flowers decreases as the number of flowers necessary to fill a bee's crop capacity increases, i.e. as the typical reward value decreases (Fig. 5).
Naïve bees need to search for the flowers, and experienced bees were found to search after removal of a found artificial flower [16]. Lihoreau et al. [16] were the first to record these searching flights and this allows for the development of a model of bee searching behaviour during trapline development. These searches comprise loops centred on the location of a found flower or the location of a missing flower [16]. The size of a typical loop tends to decrease with experience (bout number) eventually becoming comparable with the ‘learnt’ typical distance between flowers. The typical size of loops made by experienced bees searching after removal of a flower also appears to be comparable with the learnt distance between flowers [16]. Here, using a simple mathematical model, we show that a looping searching strategy is near optimal for the location of flowers when the expected distance between flowers is known (has been learnt from experience), and when the typical loop size is comparable with that distance. Our finding suggests that the naïve bees gradually optimize their loop searching strategy by utilizing information they gain about the distance between flowers, and then this use optimal strategy when searching after removal of a flower, i.e., when searching after a known food source becomes depleted.
In the model of searching developed here, a bee travels out from the origin of its search (the nest initially or the location of a previously found or missing flower) along a randomly orientated straight-line (the outward leg of a loop) whose length, , is drawn from an exponential distribution where is the average length of the outward leg of a loop. The bee then flies continuously in that direction whilst constantly searching for the flower. The search ends if the flower lies within a ‘direct perception’ distance, r, of the bee. If the flower is not sighted, the bee stops after traversing the distance, , and returns to the origin of its search by retracing its outward flight. It then randomly chooses a new direction and a new distance before travelling out again. The search is centred on the origin because, initially at least, that is the most likely location of the flower.
The number of loops, , in a searching flight can be estimated by simply noting that a search will end when the length of the longest loops, , become comparable with the distance from the centre of the search to the flower, , i.e. by noting that which dictates that a loop with length longer than occurs at most once in the search pattern [35]. This condition gives . It follows from this that the average length of an entire search path, , is given by and so is minimal when , i.e. when the average length of the outward leg of a loop equals the expected distance to the flower. This optimization is not specific to exponential loop-length distributions and has been validated in numerical simulations (data not shown). It can be understood intuitively. If loops tend to be shorter than the distance between flowers then the search will be very long because most loops will fall short of the nearest flower so prolonging the search (sufficiently long loops will be rare). If loops tend to be longer than the distance between flowers then the search is unnecessarily long as the bee will frequently fly beyond where a new or missing flower is expected to be. We could not actually test this model in detail because of the small amount of empirical data available.
Previously we showed that a simple iterative improvement heuristic model of bumblebee traplining can accurately replicate the establishment of stable foraging routes by bees between five flowers distributed over several hectares [16]. In this paper, we have confronted this model to five other datasets from the literature and demonstrated that it also captures the development of traplines at smaller spatial scales in different arrangements of flowers. We showed that the model predicts correctly the formation of stable optimal traplines for some arrangements of flowers but not others, and accounts for the observed increasing propensity of bees to find optimal routes with increasing spatial scale [16], [32]. Bees foraging on several distant patches are therefore expected to eventually minimize overall travel distances between patches but not necessarily travel distances within patches. The model can also be modified to account for the observed trade-off between travel distance and prioritization of high-reward sites [32].
The model predicts that bees can, during the course of a day (ca. 65 foraging bouts), find solutions or good approximate solutions to the TSP. These approximate solutions tend to have a certain level of instability because bees never quite abandon interfacing exploration with the exploitation of known resources in a known order, so that an optimal route can be followed by a sub-optimal route. The bumblebee algorithm as encoded by our model also becomes impractical for 20 or more locations. However, it is effective for up to about 10 locations, which in practice could facilitate the linking up of flower patches or large plants (trees or bushes) with an optimal or near optimal routes rather than individual flowers as bumblebees typically visit 100's or even 1000's of individual flowers before returning to their nests [36]. The algorithm is less effective at linking up individual flowers within a patch. The model also predicts that after locating multiple flowers whose total nectar volume is in excess of their crop capacity, bumblebees can develop highly effective traplines, by visiting only a set of flowers with an appropriate spatial configuration. This selection arises naturally within our model without the need for additional modelling.
Despite a long history of research on bee learning and navigation, most knowledge has been deduced from the behaviour of foragers travelling between their nest and a single feeding location [23]. Only recently, studies of bumblebees foraging in arrays of artificial flowers fitted with automated tracking systems have started to describe the learning mechanisms underpinning complex route formation between multiple locations [14]–[16], [29]–[32]. The demonstration that all these observations can be accurately replicated by a single learning heuristic model holds considerable promises to further investigate these questions and fill a major gap in cognitive ecology [37]. We also provided theoretical evidence that the searching strategies employed by bumblebees and reminiscent of those seen in desert ants and in desert isopods [38], [39] become optimized over time as the bees gain knowledge about the spacing between flowers. They can be contrasted with the ‘scale-free’ strategies adopted by honeybees when searching for their hive or after the only known food becomes depleted; situations lacking a characteristic spatial scale [40], [41]. Future developments of our modelling platform will allow us to generate specific empirically testable predictions about how different organisations of spatial memory might produce different movement patterns and optimisation dynamics by bees in more ecologically relevant situations, for instance in the presence of competitors or in environments containing resources of different nutritional values. In the future, by incorporating searching behaviours and flight trajectories into the model, we will be able to make even more robust and precise predictions about trapline development.
|
10.1371/journal.ppat.1003419 | DNA Methylation Impacts Gene Expression and Ensures Hypoxic Survival of Mycobacterium tuberculosis | DNA methylation regulates gene expression in many organisms. In eukaryotes, DNA methylation is associated with gene repression, while it exerts both activating and repressive effects in the Proteobacteria through largely locus-specific mechanisms. Here, we identify a critical DNA methyltransferase in M. tuberculosis, which we term MamA. MamA creates N6-methyladenine in a six base pair recognition sequence present in approximately 2,000 copies on each strand of the genome. Loss of MamA reduces the expression of a number of genes. Each has a MamA site located at a conserved position relative to the sigma factor −10 binding site and transcriptional start site, suggesting that MamA modulates their expression through a shared, not locus-specific, mechanism. While strains lacking MamA grow normally in vitro, they are attenuated in hypoxic conditions, suggesting that methylation promotes survival in discrete host microenvironments. Interestingly, we demonstrate strikingly different patterns of DNA methyltransferase activity in different lineages of M. tuberculosis, which have been associated with preferences for distinct host environments and different disease courses in humans. Thus, MamA is the major functional adenine methyltransferase in M. tuberculosis strains of the Euro-American lineage while strains of the Beijing lineage harbor a point mutation that largely inactivates MamA but possess a second functional DNA methyltransferase. Our results indicate that MamA influences gene expression in M. tuberculosis and plays an important but strain-specific role in fitness during hypoxia.
| Tuberculosis is a disease with a devastating impact on public health, killing over 1.5 million people each year around the globe. Tuberculosis is caused by the bacterium Mycobacterium tuberculosis, which over millennia has evolved the ability to survive and persist for decades in the harsh environment inside its human host. Regulation of gene expression is critical for adaptation to stressful conditions. To successfully tackle M. tuberculosis, we therefore need to understand how it regulates its genes and responds to environmental stressors. In this work, we report the first investigation of the role of DNA methylation in gene regulation and stress response in M. tuberculosis. We have found that DNA methylation is important for survival of hypoxia, a stress condition present in human infections, and furthermore that DNA methylation affects the expression of several genes. In contrast to methylation-regulation systems reported in other bacteria, in which the effects of methylation vary from one gene to the next, M. tuberculosis appears to use a concerted mechanism to influence multiple genes. Our findings identify a novel mechanism by which M. tuberculosis modulates gene expression in response to stress.
| Mycobacterium tuberculosis is a pathogen of tremendous global significance, causing 9 million cases of tuberculosis annually and latently infecting up to a third of the world's population [1]. Untreated, M. tuberculosis can persist for decades in the infected host. Over such timescales, the bacterium must tune gene expression patterns to match conditions in the host environment, including hypoxia, nutrient deprivation, and low pH, and maintain these adaptations over long periods of time.
How might M. tuberculosis durably maintain gene expression patterns? While eukaryotes use a variety of mechanisms to heritably ensure expression states, DNA methylation is the only known mechanism by which prokaryotes might achieve epigenetic inheritance. Both adenine and cytosine can be methylated in DNA, resulting in N6-methyladenine, N4-methylcytosine, and 5-methylcytosine (accurately termed N6-methyl-2′deoxyadenosine, N4-methyl-2′deoxycytidine, and 5-methyl-2′deoxycytidine, and abbreviated here as N6-MdA, N4-MdC, and 5-MdC, respectively). Cytosine methylation is an important mechanism of repressing gene expression in higher eukaryotes and recent reports suggest that 5-MdC has regulatory roles in prokaryotes [2], [3]. However, in prokaryotes N6-MdA is the best-characterized epigenetic regulator of gene expression [4]–[9].
Regulation of gene expression by adenine methylation has been described mainly in the Proteobacteria where it is primarily mediated by the Dam methyltransferase in the Gammaproteobacteria and CcrM in the Alphaproteobacteria, although other methyltransferases of unknown function have been identified [5], [10]. Dam-mediated methylation has pleiotropic roles that include directing DNA mismatch repair, suppressing transposition, and regulating genes involved in cell cycle timing and antigenic variation [5]–[9], [11]–[17]. In Escherichia coli, genetic disruption of dam causes a modest growth defect [18], an increased mutation rate [19], [20], and numerous gene expression changes [21]–[23]. Some of these expression changes result directly from the methylation state of a given promoter, but most seem to reflect the downstream consequences of cell cycle changes and perturbed DNA repair [7]–[9], [11], [15]–[17], [24]–[28]. Even where Dam methylation has been shown to regulate gene expression directly, the mechanistic details are highly locus-specific [7], [17], [29], [30]. There are several known transcriptional repressors that bind DNA in a methylation state dependent manner. Methylation may permit or prevent repressor binding, depending on the repressor and the spatial relationship between the Dam site and other promoter elements. However, the pleiotropic roles of Dam methylation in cell cycle regulation and DNA repair make it difficult to distinguish between direct and indirect effects on gene expression. Furthermore, over half of the ORFs in the E. coli genome have two or more Dam sites in the 500 base pair region upstream [21], making the presence of Dam sites a poor indicator of Dam-mediated regulation.
Virulent M. tuberculosis has been reported to contain both N6-MdA and 5-MdC [31]. However, there are no predicted dam or dcm homologues in the genome and canonical Dam and Dcm sites are not methylated [31], [32]. Van Soolingen and colleagues identified a site in the lppC gene that was protected from restriction digest in clinical M. tuberculosis strains [33] and predicted this to be due to DNA methylation. However, nothing further was known about the mechanism or functional consequences of DNA methylation in M. tuberculosis.
Interestingly, the extent of lppC protection differed among strains from the different phylogeographic lineages of M. tuberculosis, with strains of the Beijing lineage showing reduced lppC protection compared to strains from other lineages [33]. The various lineages of M. tuberculosis are associated with different epidemiological characteristics. Most notably, strains of the Beijing lineage appear to be increasing in prevalence globally, suggesting that this lineage has a competitive advantage in the modern world [34]–[36]. While the success of the Beijing lineage is likely multifactorial, some of its unique characteristics have been hypothesized to arise from differences in regulatory circuitry that may alter adaptation to specific host environments [35], [37]–[40].
Based on these findings, we hypothesized that DNA methylation might regulate gene expression in M. tuberculosis, with functional significance in specific host environments or genetic contexts. We identify a methyltransferase, MamA (M.MtuHIII according to systematic DNA methyltransferase nomenclature [41]), and show that it methylates a six base pair sequence in the M. tuberculosis genome in a strain specific manner. We demonstrate that MamA methylation affects expression of several genes. Using a novel approach to map the transcriptional start sites of these genes we demonstrate that in each case, a methylation site overlaps with the sigma factor binding site in an identical configuration. Importantly, we show that loss of MamA reduces the ability of M. tuberculosis to survive in hypoxia, a stressor thought to mimic the environment that the bacterium encounters in the human host.
In order to investigate the determinants of DNA methylation in the M. tuberculosis genome, we began by examining a site in the lppC gene that had been previously reported to be protected from restriction enzyme cleavage [33]. Consistent with the published data, we confirmed that this site was largely protected from cleavage by PvuII in M. tuberculosis strains from the Euro-American lineage and the vaccine strain M. bovis BCG (Figure 1, A and B), but was fully susceptible to PvuII in strain HN878, a member of the Beijing lineage of M. tuberculosis (Figure 1, B and C). As the PvuII recognition sequence was present in all strains, it had been postulated that differential methylation was the most likely explanation for the variable PvuII cleavage [33]. A 10 base pair sequence containing the PvuII recognition site was shown to be protected from PvuII cleavage [33]; methylation of the adenine residues within this sequence is expected to block PvuII cleavage [42] and the effects of cytosine methylation are unknown (Figure 1E).
There are two predicted DNA methyltransferases encoded in the M. tuberculosis genome, neither of which is associated with a cognate restriction endonuclease. To determine if either of these methyltransferases was responsible for the DNA modification at the lppC site, we constructed unmarked deletion mutants of these genes in H37Rv, a commonly used lab strain of M. tuberculosis that belongs to the EuroAmerican lineage. Deletion of Rv3263 abolished protection of the lppC site from PvuII cleavage (Figure 1C). In contrast, deletion of hsdM did not affect protection of this site. Complementation of the Rv3263 deletion strain with an ectopic copy of the gene restored protection (Figure 1D and Figure S1). The Rv3263 gene product from H37Rv will be called M.MtuHIII according to standard DNA methyltransferase nomenclature [41]. As systematic methyltransferase names are strain-specific, we have also chosen a generic name that can be applied to all M. tuberculosis strains. We therefore refer to Rv3263 and its gene product as mamA and MamA, respectively (Mycobacterial adenine methyltransferase). MamA is conserved in relatives of M. tuberculosis including M. bovis BCG (Figure 1), the pathogens M. leprae and M. avium, and the saprophyte M. smegmatis (TB Database, [43]).
To identify the base that MamA methylates, we constructed an episomal plasmid containing the 10 base pair sequence sufficient to enable protection from PvuII cleavage and propagated the plasmid in both wildtype M. tuberculosis and the mamA deletion mutant. We then assessed the methylation status of the 10 base pair sequence using sequence trace comparison. This method is based on differing incorporation of dye terminator nucleotides complementary to methylated adenine or cytosine residues in conventional Sanger sequencing, allowing methylation status to be inferred by comparing sequencing traces from identical sequences of DNA propagated in the presence and absence of the methyltransferase [44], [45]. The change in nucleotide incorporation depends on the methylated base in the template: N6-MdA results in increased incorporation of dideoxythymidine nucleotides yielding higher thymine peaks while 5-MdC and N4-MdC result in less and more dideoxyguanosine incorporation, respectively, and thus lower and higher guanine peaks [10], [44], [45]. We propagated the plasmid in methylation-proficient and methylation-deficient M. tuberculosis and E. coli, then purified and sequenced it. Representative sequence traces are shown in Figure 2A. The thymine peak in position 5 of the top strand sequence showed increased intensity in plasmid isolated from the methylation-proficient M. tuberculosis strain H37Rv, relative to the equivalent peak in sequences of plasmid isolated from E. coli, H37Rv ΔMamA, and M. tuberculosis strain HN878. Similarly, the thymine peak in position 3 of the opposite strand was relatively higher in plasmid from H37Rv. Quantification of differences in peak area is shown in Figure S2. These alterations in relative peak height reflect increases in dideoxythymidine incorporation, suggesting presence of N6-MdA in the complementary templates isolated from H37Rv (Figure 2B).
We also noted a reduction in the height of the guanine peak following the elevated thymine peak in H37Rv-derived DNA (Figure 2A, “top strand” and Figure S2A). This reflects decreased dideoxyguanosine incorporation and would be consistent with the presence of 5-MdC in the template, but bisulfite sequencing of H37Rv-derived plasmid indicated that no methylcytosine was present (data not shown). The peak height difference is therefore likely a result of the preceding N6-MdA causing an effective change in sequence context. Similar alterations in the incorporation of nucleotides neighboring the base complementary to the site of methylation have been observed previously [10], [46].
To determine the minimal recognition sequence required for methylation by MamA, we systematically mutated the 10 base pair sequence shown in Figure 2B and performed sequence trace comparison on the resulting plasmids. A central core of six base pairs (bold in Figure 2B) was sufficient to direct methylation in H37Rv (Figure 2C). Any further changes to this six base pair sequence abrogated methylation (Table S1). The MamA recognition site “CTGGAG” is predicted to be present in 1947 locations in the H37Rv genome. The sites are distributed across the genome, without any obvious skew with respect to the origin of replication (Figure 2D). Interestingly, there is a strong bias regarding the orientations of MamA sites relative to the coding strand within open reading frames. Of the 1816 times that MamA sites occur within annotated coding regions, the sequence reading “CTGGAG” is located on the coding strand in 1511 cases, while it is located on the non-coding strand in only 305 cases (p<0.0001, Chi square test with Yates correction). This may be at least partially a result of codon bias, as the codons “CTG” and “GAG” are both favored in M. tuberculosis while “CTC,” “TCC,” and “CCA” are all relatively disfavored [47]. Two other bacterial DNA methyltransferases, M.GsuI and M.BpmI, are known to recognize an identical sequence to MamA; however, the roles of these enzymes are not known so they did not provide clues regarding the function of MamA (Rebase, [42]).
To investigate the role of MamA within the broader DNA methylation landscape of M. tuberculosis, we defined the spectrum of methylated nucleobases in M. tuberculosis DNA using liquid chromatography-coupled tandem mass spectrometry (LC-MS/MS). Genomic DNA was enzymatically digested to individual nucleosides and subject to LC- MS/MS to quantify N6-MdA, 5-MdC, N4-MdC, and 5-hydroxymethyl-2′deoxycytidine (5-HMdC). In DNA from wildtype H37Rv, N6-MdA occurred at a rate of 4.9+/−2.2 per 104 nucleotides (nts) (0.27% of adenosines) (Figure 3A), which would yield 2048+/−910 N6-MdA per strand in the 4.4 million bp genome, correlating well with the expected 1947 MamA sites (+/− denotes SD). Deletion of mamA reduced N6-MdA to less than five N6-MdA per genome strand. Taken together, these data indicate that MamA is the major adenine methytransferase active in H37Rv. Complementation of H37Rv ΔmamA with mamA restored N6-MdA to wildtype levels. 5-MdC, N4-MdC, and 5-HMdC were not detected in any of the M. tuberculosis strains (limits of detection were approximately five per 108 nts for 5-MdC and N4-MdC, and one per 105 nts for 5-HMdC). The absence of 5-MdC was somewhat surprising given that this modification has been previously reported in H37Rv [31], [32]. However, it is consistent with the absence of a predicted cytosine methyltransferase in the genome.
To confirm that MamA is itself a DNA methyltransferase (rather than an activator of other methyltransferases, for example), we made point mutations predicted to disrupt the MamA active site. MamA is predicted to be a Type II DNA methyltransferase with architecture of the gamma subtype [48], and is therefore homologous to the well-characterized methyltransferases M.EcoKI and M.TaqI. Biochemical studies of these and other adenine methyltransferases demonstrate that mutations of key residues in the S-adenosylmethionine binding site inhibit methyltransferase activity without disrupting overall protein structure [49]–[54]. We therefore mutated these key residues and expressed constructs encoding MamAN127D and MamAY130A in H37Rv ΔmamA. We then assessed the effects of these mutations on DNA methylation using LC-MS/MS analysis. Both mutants displayed little or no adenine methylation (Figure 3A), indicating that the predicted active site of MamA is critical to its ability to confer methylation.
We sought to understand the loss of adenine methylation at the PvuII site in lppC in Beijing lineage strains of M. tuberculosis. Sequencing of mamA revealed that the Beijing lineage strain HN878, and all modern Beijing lineage strains for which genome sequences are publicly available, have a point mutation of nucleotide A809C causing the amino acid substitution Glu270Ala. While Glu270 is not located in the predicted active site, it is part of a conserved motif, suggesting that a non-conservative mutation to Ala could disrupt enzyme activity. Consistent with this hypothesis, mamAE270A failed to restore protection from PvuII cleavage in an H37Rv ΔmamA strain (Figure 1D). LC-MS/MS analysis revealed that H37Rv harboring mamAE270A had N6-MdA levels that were 50–100 fold lower than the wildtype parent (Figure 3A).
Interestingly, strain HN878 had only a 3-fold reduction in N6-MdA compared to H37Rv, despite harboring the mamAE270A allele (Figure 3A). This suggested that HN878 has a substantial amount of MamA-independent adenine methylation, in contrast to H37Rv. We therefore further examined the contributions of the two predicted DNA methyltransferases, MamA and HsdM, to total adenine methylation levels in the two strain backgrounds. In H37Rv and most members of the Euro-American lineage of M. tuberculosis, hsdM contains a mutation resulting in the amino acid change Pro306Leu in the active site, which is predicted to abolish HsdM activity [48], [55]. Indeed, LC-MS/MS analysis of H37Rv ΔhsdM demonstrated that deletion of hsdM did not reduce levels of N6-MdA suggesting that in H37Rv, hsdM does not appreciably contribute to the N6-MdA content of the genome. Consistent with the idea that a Pro306Leu mutation is responsible for the lack of detectable HsdM activity in H37Rv, reintroduction of a wildtype Pro306 allele of hsdM to H37Rv ΔmamA significantly increased N6-MdA levels (Figure 3B). Since HN878 naturally encodes a wildtype Pro306 allele of hsdM, the excess N6-MdA in HN878 relative to H37Rv mamAE270A is likely to reflect greater HsdM activity in HN878 as compared to H37Rv.
We also predicted that complementing HN878 with a wildtype Glu270 allele of mamA would increase total N6-MdA levels. Interestingly, restoration of wildtype MamA to HN878 resulted in a quantitatively greater increase in N6-MdA than expected based on the effect of complementing H37Rv ΔmamA with the same construct expressing mamA (Figure 3). These data suggest that strain genetic background affects expression and/or activity of individual methyltransferases.
As DNA methylation regulates gene expression in other organisms, we sought to determine if MamA serves a similar function in M. tuberculosis. We used an Affymetrix microarray platform to perform global transcriptional profiling of triplicate log-phase cultures of wildtype H37Rv, ΔmamA, and complemented strains (Table S2 for complete dataset; GEO accession number GSE46432). Table 1 lists genes with expression differences of 1.5-fold or greater between wildtype H37Rv and either of the other two strains. Because we saw only a modest number of expression differences of limited magnitude, we felt that the microarray experiment was best used as a hypothesis-generating tool. Recognizing that small changes in a bulk expression assay may reflect larger changes in heterogeneous subpopulations of bacteria, we hypothesize that such apparently subtle changes might be functionally important. Several genes showed lower expression in ΔmamA compared to wildtype and complemented strains and had MamA sites in the region upstream of their annotated start codons (Table 1). These genes were considered to be candidates whose expression might be directly regulated by DNA methylation. Other genes showed altered expression only in the complemented strain relative to the wildtype and ΔmamA strains. These genes were located in the vicinity of the integrating complementation vector and their expression changes were thus likely be a result the strain construction strategy and not related to methylation status. Rv0102, Rv0142, corA, whiB7, and the Rv3083 operon were the strongest candidate methylation-affected genes. We therefore re-tested their expression levels by quantitative PCR (qPCR), using RNA derived from independent cultures, and confirmed that the ΔmamA strain had significantly reduced expression of Rv0102, Rv0142, corA, and whiB7 (Figure 4).
To understand how MamA affects gene expression, we mapped the transcriptional start sites (TSSs) of the qPCR-confirmed genes. We employed a novel strategy based on mRNA circularization in order to map TSSs quickly and accurately (Figure 5A). Total RNA preparations were subject to rRNA depletion and treated with a 5′polyphosphatase to convert 5′ triphosphate “caps” to 5′ monophosphates. The resulting mRNA-enriched samples were circularized by T4 RNA ligase to create molecules containing junctions between 5′ and 3′ ends. Linear cDNA molecules were synthesized from the circular templates by random priming. The 5′-3′ junctions were amplified by gene-specific primers and sequenced by a reverse primer annealing shortly downstream of the start codon. Because the 3′ ends of mRNAs are variable, the 5′-3′ junctions appeared as transitions from monomorphic to polymorphic sequence (Figure 5B). We validated this method by mapping the TSS of whiB1, and found that our method predicted a TSS identical to that identified by 5′ Rapid Amplification of cDNA Ends (Figure S3) [56].
The TSSs of Rv0102, Rv0142, corA, and whiB7 were each located four or five base pairs downstream of a MamA methylation site (Figure 5B). In each case, the predicted sigma factor −10 binding site overlaps with the MamA site such that the last nucleotide of the −10 site is predicted to be methylated on the template strand while the nucleotide three base pairs downstream of the −10 site is predicted to be methylated on the non-template strand (Figure 5B). The conserved spatial relationship between methylation sites and sigma factor binding sites is striking and potentially suggestive of a shared regulatory paradigm among these genes.
Given a role for MamA in influencing gene expression, we then sought to determine the functional consequences of losing MamA function. Because DNA methylation plays roles in cell cycle regulation, genome stability and pathogenicity in Proteobacteria [57], [58], we investigated the effects of mamA deletion under a number of different conditions. There were no distinguishable differences in growth between wildtype H37Rv, ΔmamA, and complemented strains in vitro under standard growth conditions (Figure 6A and data not shown). Sensitivity to reactive nitrogen and oxygen species was assessed and no significant differences in survival were observed among the strains (Figures S4A and S4B). ΔmamA and complemented strains did not differ in their abilities to compete with wildtype H37Rv in murine infections (Figure 6B and Figure S5). Mutation rates were likewise unaffected by deletion of mamA; this was unsurprising given that Mycobacteria lack homologs of the key proteins involved in methyl-directed DNA repair in Proteobacteria (Table S3) [59].
Changes in gene expression are often associated with adaptation to new conditions, and we reasoned that MamA might be necessary for adapting to an environment that is not modeled accurately in the mouse. Humans develop hypoxic granulomas that are thought to slow or even arrest the growth of M. tuberculosis, but the lesions observed in mice are less organized and remain oxygenated [60]–[62]. Therefore, we therefore tested the ability of wildtype H37Rv, ΔmamA and complemented strains survive in hypoxic culture. Bacteria were seeded into vials with a defined headspace/liquid ratio, sealed and incubated with slow stirring, allowing the bacteria to gradually deplete the enclosed oxygen supply and enter a non-replicating state [63]. While all strains displayed reduced viability over time as measured by colony forming units (CFU), the ΔmamA strain died at a significantly faster rate than the wildtype parent and the complemented strain (Figure 7A). To assess the viability of the bacteria exposed to hypoxia using a complementary method, a portion of each culture was removed and treated with fluorescein diacetate at day 28. Only viable cells have active intracellular esterases that convert fluorescein diacetate to fluorescein, inducing fluorescence [64]. Consistent with the colony counts, hypoxic cultures of the wildtype and complemented strains had a high proportion of viable cells (∼37–62%), while hypoxic cultures of the ΔmamA strain had significantly fewer viable cells (∼23%) (Figure 7B–C).
In this work we demonstrate that Rv3263 encodes an adenine methyltransferase, MamA, which is responsible for all detectable DNA methylation in the Euro-American lineage strain of M. tuberculosis, H37Rv. By deleting mamA, we show that in M. tuberculosis, adenine methylation alters gene expression. Furthermore, mamA is required for optimal bacterial survival in a hypoxic environment. The expression changes mediated by MamA appear subtle in the bulk assays we used. One possible explanation is that methylation may direct greater expression differences in a subpopulation of cells. Single-cell methods will be required to explore this possibility and to determine whether DNA methylation allows heritable (epigenetic) regulation of gene expression in M. tuberculosis.
How does MamA alter gene expression? In each case that we have identified, the MamA site overlaps the sigma factor −10 binding site, a promoter region that is directly bound by the RNA polymerase holoenzyme during the initiation of transcription. Strikingly, the MamA sites are located at exactly the same position relative to the −10 sites in the four MamA-affected genes that we examined. This shared spatial configuration contrasts with the locus-specific relationships between Dam methylation sites and promoters at the known methylation-regulated genes in Proteobacteria [5], [65]. The overlap between the promoter MamA sites and the sigma factor −10 binding sites is highly suggestive of a direct effect of methylation on expression of these genes. The lack of broad transcriptional changes or growth rate changes in the ΔmamA strain also suggests that the global physiology of the mutant is unperturbed under normal conditions, making indirect effects on transcription less likely. Indeed, the apparently restricted role of MamA under normal growth conditions may have allowed us to detect a category of subtle but direct effects on transcription that might also exist in E. coli but which are difficult to detect given the high frequency of Dam sites and dramatic effects of dam deletion on cellular physiology.
In the Proteobacteria, methylation has been shown to affect transcription by two broad mechanisms: (1), modulation of repressor binding, and (2), direct modulation of RNA polymerase's interactions with the promoter. Either of these mechanisms could underlie the MamA-dependent expression changes we observe and we propose several potential models for this effect. Methylation could potentially prevent binding of a transcriptional repressor, enhance promoter recognition by the RNA polymerase holoenzyme, increase the melting efficiency of the promoter, or enhance the stability of the open complex.
Three E. coli genes are known to be regulated by methylation sites that overlap their sigma factor −10 binding sites, as we observe here, but they do not share a common regulatory paradigm. One of these, dnaA, is regulated by repressor binding [6], [8]; another, IS10 transposase, is regulated by direct effects of methylation on RNA polymerase interaction with the promoter [16]; and the third, glnS, is Dam-regulated by unknown mechanisms [66]. The dnaAp2 promoter harbors several Dam sites, one of which overlaps the −10 region. When the Dam sites are hemimethylated following DNA replication a repressor, SeqA, binds and inhibits transcription [6], [8], [11]. Later in the cell cycle, when the promoter becomes fully methylated, expression resumes [12]. This regulatory paradigm clearly falls into the category of methylation-state-dependent repressor binding, which includes other genes with Dam sites in different configurations, such as pap and agn43 [7], [9], [15], [17].
In the case of the IS10 transposase, methylation is thought to alter RNA polymerase interaction with the promoter. Here, methylation directly inhibits expression from the transposase promoter in vitro. Hemimethylated promoters have activity that is intermediate between fully methylated and unmethylated promoters [16]. These findings suggest that methylation directly affects either the binding of the RNA polymerase holoenzyme to the promoter, open complex formation, or open complex stability. We note that the effect of methylation on IS10 expression is the opposite of what we observe for MamA-affected genes, and that the methylated bases lie at different positions within the −10 site in the two systems [16].
There are several mechanisms by which DNA methylation may affect open complex formation and stability. DNA is initially melted over a region extending from the second bp of the −10 site to just past the TSS in order to form open complexes [67]. This process involves physical interaction between the −10 site and the 2.4 region of sigma as well as a more recently identified interaction between the DNA shortly downstream of the −10 site and the 1.2 region of sigma [68]–[71]. N6-MdA can reduce the melting temperature of DNA heteroduplexes in vitro [72], which may make open complex formation more thermodynamically favorable. DNA melting efficiency is thought to be important in open complex formation in part because of the AT-rich nature of −10 sites in most bacteria, and because a GC-rich region between the −10 site and TSS is important for stringent control of some promoters [70], [73], [74]. In MamA-affected genes, N6-MdA in the −10 site and between the −10 site and TSS could therefore potentially enhance open complex formation and consequently increase expression.
It is also possible that region 1.2 of sigma could make direct contact with the N6-MdA located between the −10 site and TSS on the non-template strand, and that such an interaction could increase open complex stability directly. Region 1.2 of sigma was shown in E. coli to make direct contact with bases on the non-template strand between the −10 site and TSS of rrnB P1 and λPR [70], [75]. In Bacillus subtilis, changing the base at position −5 of rrnB P1 from T to A resulted in a decrease in sensitivity of the promoter to GTP levels, indicating that the stability of its open complex was increased [76]. Together, these data suggest that the sequence of the region between the −10 site and the TSS matters for reasons beyond GC content, and the non-template strand in particular plays an important role. In vitro studies will be required to elucidate the effects of MamA-mediated methylation on interactions between M. tuberculosis promoters and RNA polymerase holoenzyme.
The MamA-affected genes are not obviously related with respect to pathway or function, although several are involved in stress responses. Rv0102 is an essential gene predicted to encode an integral membrane protein of unknown function [77]. It is not reported to undergo major transcriptional changes [78], although it may be modestly induced by oxidative stress [79]. Rv0142 is predicted to encode a DNA glycosylase and is strongly induced in response to oxidative stress in a σH-dependent fashion [79]–[81]. It may also be induced by nitrosative stress [79]. CorA encodes a predicted magnesium and cobalt transporter that may be modestly induced by thioridazine, proton gradient disrupters, and oxidative stress [78], [79], [82]. We find that induction of Rv0142 and corA in response to oxidative stress appears to occur normally in a ΔmamA strain (data not shown), suggesting that MamA methylation affects the basal transcription of these genes but not the higher-level transcription that occurs during oxidative stress. WhiB7 is a transcriptional regulator that is induced by stationary phase, multiple antibiotics, heat shock, and iron starvation; disruption of whiB7 leads to increased antibiotic susceptibility [83], [84]. Further work is needed to understand whether the hypoxia-survival defect in ΔmamA is related to the gene expression differences we detected or is a result of other effects of mamA deletion on hypoxia-specific gene expression. Gene expression profiling of 2-week-old hypoxic cultures suggested that a small number of genes are differentially expressed in the absence of mamA, similar to the extent of gene expression changes under standard in vitro growth conditions. These include some genes influenced by MamA in aerobic growth and some novel potential targets (data not shown). Further studies are required to validate these findings and understand their implications for M. tuberculosis survival under hypoxic conditions.
Interestingly, MamA is partially inactivated by a point mutation in most strains of the Beijing lineage of M. tuberculosis, while a second methyltransferase, HsdM, is active in the Beijing lineage and inactivated by a point mutation in most Euro-American strains. Non-synonymous methyltransferase mutations are found in other M. tuberculosis lineages as well (TB Database, [43]), although the effects of these mutations on methyltransferase activity are unknown. A few of the oldest strains of the Beijing and Euro-American lineages, as well as strains from the rim of the Indian Ocean, appear to encode intact copies of both MamA and HsdM. Further work will be needed to understand how the roles of DNA methylation may differ according to genetic background. It is possible that MamA is important for fitness of Euro-American strains during infection-associated hypoxia, but is unnecessary or even detrimental in modern Beijing lineage strains due to their altered hypoxia-response gene regulatory networks [39], [40]. Loss of MamA function may affect other aspects of the Beijing lineage strains. The insertion sequence IS6110 has a MamA site that overlaps both the inverted repeat and the presumed promoter for the transposase gene. The number of IS6110 elements is higher in Beijing lineage strains than in other lineages, suggesting that transposition may be more frequent in Beijing lineage strains [85]. IS6110 activity may be beneficial because it introduces genetic variability into a clonal species that lacks opportunities for horizontal gene transfer [86]. In E. coli IS10 transposition is altered by Dam through both expression-dependent and –independent mechanisms [16]. Although we did not detect changes in IS6110 transposase expression in strains lacking mamA (data not shown), future studies may indicate an effect of MamA on IS6110 transposition rates.
In this work we report the first investigation of the functional effects of DNA methylation in M. tuberculosis, as well as basic characterization of where and how DNA methylation occurs in this globally important bacterium. Methylation enhances expression of several genes that have methylation sites located in identical positions within their promoters, consistent with a shared regulatory paradigm. The activity of the methyltransferase MamA is required for normal survival of hypoxia, indicating that it is likely an important mediator of adaptation to this physiologically relevant stressor. Different methyltransferases predominate in different lineages of M. tuberculosis, suggesting that methylation-mediated regulatory pathways may contribute to lineage-specific characteristics.
Animal experiments were performed in strict accordance with the National Institutes of Health guidelines for housing and care of laboratory animals, with institutional regulations after protocol review, and with approval by the Harvard Medical Area Standing Committee on Animals. The animal protocol was approved by the Harvard University IACUC (protocol number 03000).
M. tuberculosis strains were grown in Middlebrook 7H9 or 7H10 media supplemented with 10% OADC (Oleic Albumin Dextrose Catalase, Becton Dickinson), glycerol, and 0.05% Tween 80 unless otherwise specified. H37Rv strains were derived from the ATCC lineage. Unmarked mamA (Rv3263) and hsdM deletion strains were constructed by a two-step process. Plasmids pSS002 and pSS004 were derived from pJM1 [87] and contained 1 kb of the sequence upstream and downstream of hsdM and mamA, respectively, with 24–27 base pairs of coding sequence and a stop codon in between the two flanks. Plasmids were linearized with restriction enzymes cutting within one of the flanks before transformation into H37Rv. Integrants were selected with 50 µg/ml hygromycin. Counterselection with 7% sucrose was followed by PCR screening to identify isolates that subsequently underwent second crossovers resulting in loss of the plasmid and hsdM or mamA coding sequences. Complementation vectors were derived from pJEB402, which integrates as a single copy into the L5 attB site [88]. The mamA coding sequence and 33 upstream bases (assumed to contain the RBS) were cloned behind the MOP promoter present in pJEB402, creating plasmid pSS030. We performed PCR with primers containing a central mutation in order to change nt 809 of the coding sequence from “A” to “C” creating pSS040. An equivalent strategy was used to insert the active site mutations N127D and Y130A into plasmids pSS075 and pSS077, respectively. For complementation with wildtype hsdM, the hsdM coding sequence and 23 upstream nt assumed to contain the RBS were cloned behind the MOP promoter in pJEB402 to produce plasmid pSS079.
Cultures were grown to an optical density of between 0.7 and 1.1 unless otherwise specified. Cell pellets were inactivated by chloroform-methanol (ratio 2∶1), pelleted, resuspended in 0.1M Tris and 1 mM EDTA, pH 8–9, and lysed with lysozyme overnight at a final concentration of 100 µg/mL. Lysates were treated with 1% SDS and 100 µg/mL proteinase K (IBI Scientific) (final concentrations) for 3 hours at 50°C followed by phenol-chloroform extraction according to standard procedures, RNase (MO BIO) treatment (25 µg/mL for 1 hour at 37°C), and a second phenol-chloroform extraction. Two µg DNA was digested with PvuII (NEB) for Southern blotting. Blotting was performed according to standard protocols. DIG-labeled probe was made and detected with Roche DIG DNA Labeling and Detection Kit and Roche DIG Wash and Block Buffer Set according to the manufacturer's instructions.
Plasmids for sequence trace comparison were constructed by digesting pMV762 [89] with BamHI and HindIII (NEB) and ligating in an annealed oligonucleotide duplex containing the sequence of interest (see Table S1 for oligonucleotide sequences). pSS012 contains the full 10 base pair sequence and other variants are listed in Table S1. pMV762 contains multiple complete and partial MamA sites, and these were sequenced as well. Plasmids were isolated from M. tuberculosis using a variation of a published protocol [90]. Briefly, 30 ml of culture was pelleted and inactivated by overnight incubation with a 4∶1 ratio of chloroform∶methanol at 4°C. After centrifugation and removal of the liquid phases, pellets were resuspended in 200 µl of lysozyme solution [90] and incubated 4–18 h at 4°C. 400 µl of alkaline SDS solution [90] was added and samples incubated 30 min at 4°C with agitation. Buffer N3 (700 µl) from a Qiagen miniprep kit was added and samples centrifuged at maximum speed for 10 min. The supernatant was then applied to a Qiagen miniprep spin column and sample was processed according to the manufacturer's instructions. E. coli derived plasmids were propagated in both DH5-alpha and a dam dcm deletion strain (NEB). No differences were observed between sequence traces from the two E. coli strains.
Five µg of DNA was digested to nucleosides enzymatically with the addition of deaminases to reduce artifactual deamination due to contaminating deaminases in commercial enzyme preparations [91]. Isotopically labeled internal standards for 5-MdC and N6-MdA were synthesized and spiked into the digestion reactions (EGP manuscript in preparation and [92]). An HPLC method that separates all four methylated products and the canonical nucleosides was developed utilizing a Cogent Diamond Hydride aqueous normal phase column (2.1×250 mm, 4 µm particle, 100 Å pore size; Microsolv Technology Corporation, Eatontown, NJ) with an isocratic step gradient of 0.1% acetic acid in acetonitrile/water (EGP manuscript in preparation). The LC-MS/MS analysis was performed on an Agilent 1100 HPLC coupled to an AB Sciex API 3000 triple quadrupole mass spectrometer in positive ion multiple reaction monitoring mode utilizing only 50–200 ng of DNA. Transitions monitored were m/z 266-150 (N6MdA), m/z 271-154 ([15N5]6-MdA), m/z 242-126 (5-MdC/N4MdC), m/z 254-133 ([13C915N3]-5-MdC), m/z 258-142 (5-hydroxymethylcytidine), m/z 252-136 (dA), m/z 243-127 (dT), m/z 228-112 (dC), and m/z 268-152 (dG). For each case, the monitored transition represents the loss of the 2′-deoxyribose. The areas under the curve of each nucleoside transition were quantitated and compared to calibration curves (r = 0.99). There were three biological replicates and at least two technical replicates per sample.
Female C57BL/6 mice were purchased from Jackson Laboratory (Bar Harbor, ME). Freshly grown cultures of wildtype H37Rv (kanamycin sensitive), ΔmamA::pJEB402, ΔmamA::pSS030, and ΔmamA::pSS040 (kanamycin resistant) were washed and cell densities were estimated by optical density. Equal quantities of wildtype bacteria were mixed with each of the three marked strains in order to perform three separate competition experiments. Mice were infected by the aerosol route with approximately 104 CFU. Four mice per group were sacrificed at the indicated time points and bacterial burden in the lung and spleen were determined by plating homogenized organs on plates both with and without 25 µg/ml kanamycin. Animal experiments were performed in accordance with the National Institutes of Health guidelines for housing and care of laboratory animals, with institutional regulations after protocol review, and with approval by the Harvard Medical Area Standing Committee on Animals.
RNA was isolated from cultures grown to OD 0.8–0.9 in the absence of antibiotics. Twenty ml of culture was added to 20 ml of RNAlater (Ambion) and incubated for 10 min. Ten ml of water was added immediately before centrifugation for 15 min. Pellets were resuspended in one ml Trizol (Invitrogen) and subject to bead-beating for 45 s and 30 s in a FastPrep-24 instrument (MP) before continuing according to the manufacturer's instructions. RNA samples were then treated with 10 U DNase Turbo (Ambion) for 1 h and purified with an RNeasy kit (Qiagen) according to the manufacturer's instructions, with the addition of RNaseOUT (Invitrogen) to the water used for elution. For quantitative PCR and TSS mapping, cDNA was synthesized as follows. One µg of RNA was mixed with 1.3 µl of 3 mg/ml random hexamers (Invitrogen), denatured at 70°C for 10 min and snap-cooled on ice before adding 4 µl 5X Superscript First Strand Buffer, 1 µl of dNTPs at 10 mM each, 0.4 µl of 500 mM DTT, 1 µl RNaseOUT, and 1 µl Superscript III (Invitrogen). Reactions were performed overnight at 42°C. RNA was degraded with the addition of 10 µl each 500 mM EDTA and 1 N NaOH and heating to 65°C for 15 min, followed by neutralization with 25 µl of 1M Tris pH 7.5. cDNA was then purified over Qiagen MinElute columns according to the manufacturer's instructions.
qPCR primers are listed in Table S4. Each 20 µl reaction contained 100–200 pg of cDNA, 2.5 pg of each primer, and 10 µl of iTaq SYBR Green Supermix with ROX (Biorad). Reactions were run in an Applied Biosystems 7300 Real Time PCR System with the following program: 50°C/2 min, 95°C/5 min, and 40 cycles of 95°C/15 s and 61°C/30 s. Expression values normalized to sigA were calculated by the Δct method. Expression differences were compared by ANOVA with Tukey's post-test using GraphPad Prism 5. qPCR was performed on separate biological replicates from those used for microarray analysis.
RNA was extracted from triplicate cultures of indicated strains as described above for expression analysis with the Affymetrix custom-designed GeneChip MTbH37Rva520730F for M. tuberculosis (GEO platform number GPL17082, designed at the Broad Institute). Microarrays were run by the Boston University Microarray Core, who prepared the probes, hybridized and scanned the arrays according to the manufacturer's directions for prokaryotic samples with high GC content. Expression estimates were derived from probe-level hybridization intensities using RMA [93] in Expression Console (Affymetrix). Differential expression of non-intergenic features was assessed using 1-way ANOVA and for each ANOVA p-value we calculated a False Discovery Rate (FDR) using the method of Benjamini and Hochberg [94] to account for the large number of genomic features we interrogated. The ANOVA and FDR calculations were done using version 2.14 of the R Language for Statistical Computing [95]. Data are available on GEO, accession number GSE46432.
Total RNA samples were subject to two consecutive rounds of rRNA depletion with the MICROBExpress kit (Ambion) according to the manufacturer's instructions. To convert the natural 5′ triphosphates of mRNAs to 5′ monophosphates, approximately one µg of enriched mRNA was treated with 5′Polyphosphatase (Epicentre) for 30 minutes at 37°C in a 10 µl reaction containing 1 µl of enzyme and the supplied buffer, followed by RNeasy purification (Qiagen). The resulting sample was then circularized in 50 µl reactions contained 200 ng of RNA, 2 µl of T4 RNA ligase I (Epicentre), and ATP and buffer according to the manufacturer's recommendations in a final volume of 50 µl. Reactions were allowed to proceed for 2 h at 37°C and purified with RNeasy. cDNA was synthesized as described. Primer sets for genes of interest were designed such that the forward primer annealed approximately 100 base pairs upstream of the stop codon and the reverse primer annealed approximately 150–200 base pairs downstream of the start codon (Table S5). PCR reactions were in 25 µl volumes and contained 0.25 µl of Phusion polymerase (Finnymes), 1.25 pg of each primer, 0.2 µl of 25 mM each dNTPs, 1X GC buffer (Finnzymes), and 12–15 ng of cDNA. Reactions were performed with genomic DNA and with cDNA derived from non-circularized RNA for comparison. Cycler conditions were 98°C/2 min, 30 cycles of 98°C/15 s, 60°C/15 s, 72°C/15 s, and a final extension of 72°C/5 min. Entire reactions were then run on a 1% agarose gel. Bands present in reactions templated from circularized samples but absent in reactions templated from non-circularized samples or genomic DNA were excised and purified with Qiagen spin columns. One or two bands were identified for each gene, and some were sharp and distinct while others appeared as smears. Entire gel-extracted products were then concentrated under vacuum and subjected to a second round of PCR with the same primers, scaled up to 50 µl and with the addition of 1 µl of DMSO. Entire reactions were again run on a gel and the purified products were sequenced directly with their respective PCR primers.
The indicated strains of H37Rv were grown in 7H9 media supplemented with ADC and Tween-80 containing selective antibiotics if necessary. Seed cultures were washed twice, normalized and inoculated at a calculated density of 3×106 cfu/ml into 31 ml fresh media without antibiotics in a rubber stopper-sealed serum bottle (62 ml total volume). Cultures were shaken at ∼120 rpm at 37°C with an intermittent manual homogenization in case of cell precipitation. Two bottles per strain were opened at the indicated time points and cfu/ml was determined after serial dilutions and plating onto 7H10 agar supplemented with OADC.
At 28 days, each hypoxic culture used for CFU determination was also was stained with fluorescein diacetate. 10 ml of culture was pelleted and resuspended in 2 ml of PBS with 0.05% Tween-80. Fluorescein diacetate was prepared as a 100X stock in acetonitrile and methanol (1∶1) and added to resuspended cells to a final concentration of 50 µg/ml. After 30 minutes of incubation at 37°C, the dye-treated cells were washed with PBS-tween to remove the residual dye and then fixed with formalin. Fluorescent cells were visualized microscopically (DeltaVison, AppliedPrecision Inc.) using identical exposure settings for all strains. Death rates were calculated by linear regression analysis of log10-transformed data for time points between day 14 and day 35 (inclusive) in GraphPad Prism 5, which compares the significance of differences in slope using the method described in [96].
|
10.1371/journal.pcbi.1000637 | How Synchronization Protects from Noise | The functional role of synchronization has attracted much interest and debate: in particular, synchronization may allow distant sites in the brain to communicate and cooperate with each other, and therefore may play a role in temporal binding, in attention or in sensory-motor integration mechanisms. In this article, we study another role for synchronization: the so-called “collective enhancement of precision”. We argue, in a full nonlinear dynamical context, that synchronization may help protect interconnected neurons from the influence of random perturbations—intrinsic neuronal noise—which affect all neurons in the nervous system. More precisely, our main contribution is a mathematical proof that, under specific, quantified conditions, the impact of noise on individual interconnected systems and on their spatial mean can essentially be cancelled through synchronization. This property then allows reliable computations to be carried out even in the presence of significant noise (as experimentally found e.g., in retinal ganglion cells in primates). This in turn is key to obtaining meaningful downstream signals, whether in terms of precisely-timed interaction (temporal coding), population coding, or frequency coding. Similar concepts may be applicable to questions of noise and variability in systems biology.
| Synchronization phenomena are pervasive in biology, creating collective behavior out of local interactions between neurons, cells, or animals. On the other hand, many of these systems function in the presence of large amounts of noise or disturbances, making one wonder how meaningful behavior can arise in these highly perturbed conditions. In this paper we show mathematically, in a general context, that synchronization is actually a means to protect interconnected systems from effects of noise and disturbances. One possible mechanism for synchronization is that the systems jointly create and then share a common signal, such as a mean electrical field or a global chemical concentration, which in turn makes each system directly connected to all others. Conversely, extracting meaningful information from average measurements over populations of cells (as commonly used for instance in electro-encephalography, or more recently in brain-machine interfaces) may require the presence of synchronization mechanisms similar to those we describe.
| Synchronization phenomena are pervasive in biology. In neuronal networks [1]–[3], a large number of studies have sought to unveil the mechanisms of synchronization, from both physiological [4],[5] and computational viewpoints (see for instance [6] and references therein). In addition, the functional role of synchronization has also attracted considerable interest and debates. In particular, synchronization may allow distant sites in the brain to communicate and cooperate with each other [7]–[9] and therefore may play a role in temporal binding [10],[11] and in attention and sensory-motor integration mechanisms [12]–[14].
In this article, we study another role for synchronization: the so-called collective enhancement of precision (see e.g. [15]–[17]), an intuitive and often quoted phenomenon with comparatively little formal analysis [18]. We explain mathematically why synchronization may help protect interconnected nonlinear dynamic systems from the influence of random perturbations. In the case of neurons, these perturbations would correspond to so-called “intrinsic neuronal noise” [19], which affect all of the neurons in the nervous system. In the presence of significant noise intensities (as experimentally found in e.g. retinal ganglion cells in primates [20]), this property would be required for meaningful and reliable computations to be carried out.
It should be noted that “protection of systems from noise” and “robustness of synchronization to noise” are two different concepts. The latter concept means that the synchronized systems remain so in presence of noise, whereas the former concept means that, thanks to synchronization, the behaviors of the coupled systems are close to the noise-free behaviors. This difference is further addressed in the Discussion.
The influence of noise on the behaviors of nonlinear systems is very diverse. In chaotic systems, a small amount of noise can yield dramatic effects. At the other end of the spectrum, the effect of noise on nonlinear contracting systems is bounded by where is the noise intensity – which can be arbitrarily large – and is the contraction rate of the system [21]. Between these two extremes, it has been shown analytically that some limit-cycle oscillators commonly used as simplified neuron models, such as FitzHugh-Nagumo (FN) oscillators, are basically unperturbed when they are subject to a small amount of white noise [22]. Yet, a larger amount of noise breaks this “resistance”, both in the state space and in the frequency space [Figures 1(A)–(D)]. This suggests that both temporal coding and frequency coding may be unusable in the context of large neuronal noise.
One might argue that it could be possible to recover some information from the noisy FN oscillators by considering the activities of a large number of oscillators simultaneously [19],[23]. Figure 2(A) shows that the spatial mean of the noisy oscillators still carries very little information when the noise intensities are large, making the population coding hypothesis also unlikely in this context. In other words, if the underlying dynamics are fundamentally nonlinear, as in the case of our FN oscillators, the spatial mean of the signals is “clean,” but contains very little information: the nonlinear nature of the systems dynamics prevents the familiar “averaging out” of noise through multiple measurements, and getting rid of the noise also gets rid of the signal.
By contrast, one can observe that when oscillators are synchronized through mutual couplings, then they become “protected” from noise, whether in temporal [Figure 1(E)], frequential [Figure 1(F)] or “populational” aspects [Figure 2(B)]. Thus, in some sense, the linear effect of averaging noise while preserving signal [24] can be achieved for these highly nonlinear dynamic components through the process of synchronization. Our aim in this article is to give mathematical elements of explanation for this phenomenon, in a full nonlinear setting. It is also to suggest elements of response to a more general question, namely: what is the precise meaning of ensemble measurements or population codes, and what information do they convey about the underlying dynamics and signals?
Consider a diffusive network of -dimensional noisy non-linear dynamical systems(1)where is a function. Note that the noise intensity is intrinsic to the dynamical system (i.e. independent of the inputs), which is consistent with experimental findings [20]. For simplicity, we set to be a constant in this article, although the case of time- and state-dependent noise intensities can be easily adapted from [21].
We consider four mathematical assumptions that will enable us to relate the trajectory of any noisy element of the network to the trajectory of the noise-free system driven by equation(A1) is an assumption on the form of the network. (A2) gives a bound on the nonlinearity of the dynamics . (A3) states that the system trajectories are resistant to small perturbations. Finally, (A4) requires that the dynamical systems in the network are synchronized.
We now give conditions to guarantee assumption (A4) for all-to-all networks of FN oscillators with identical couplings. The dynamics of noisy FN oscillators coupled by (gap-junction-like) diffusive connections is given by(2)where . We show in Methods that, after exponential transients of rate ,(3)Thus, (A4) is verified with(4)For large , we have , which converges to 0 when . Figure 3(A) provides a comparison of this theoretical bound with simulations.
Assumption (A1) is also verified because an all-to-all network with identical couplings is symmetric, therefore balanced. Since the are oscillators with stable limit cycles, it can be shown that the trajectories of the are bounded by a common constant . Thus (A2) is verified with . Finally, (A3) may be adapted from [22]. Indeed, we believe that the arguments of [22] can be extended to the case of non-white noise. Making this point precise is the subject of ongoing work.
Using now the “general analytical result”, we obtain that, given any (non necessarily small) noise intensity , in the limits for and and after exponential transients, the behavior of any oscillator will be arbitrary close to that of a noise-free oscillator (Figure 1).
This statement can be further tested by constructing a model-based nonlinear state estimator (observer) [29]. Let be a noisy synchronized oscillator and consider the observer(5)
If has the same trajectory as a noise-free FN oscillator, then it can be shown that tends exponentially to , independently of the observer's initial conditions [29]. Thus the squared distance indicates how close is from a noise-free oscillator [see Figure 3(B) for a comparison this theoretical result with simulations].
We provide in this section simulation results which show that similar observations can be made even for more general network classes that are not yet covered by the theory. We believe that this simulations show the genericity of the concepts presented above.
We have argued that synchronization may represent a fundamental mechanism to protect neuronal assemblies from noise, and have quantified this hypothesis using a simple nonlinear neuron model. This may further strengthen our understanding of synchronization in the brain as playing a key functional role, rather than as being mostly an epiphenomenon.
It should be noted that the causal relationship studied here – effect of synchronization on noise – is converse to one usually investigated formally in the literature – effect of noise on synchronization: under certain conditions, adding noise can de-synchronize already synchronized oscillators (destructive effect) [32]; under other conditions, adding noise can, on the contrary, synchronize oscillators that were not synchronized (constructive effect) [33],[34]; for a review, see [35]. Also, previous papers have studied a similar phenomenon of improvement in precision by synchronization. Enright [28] shows improvement in a model of coupled relaxation oscillators, all interacting through a common accumulator variable (possibly being the pineal gland). This improvement has been experimentally shown in real heart cells [36]. More recently, [37] shows a way to get better than improvement. However, their studies primarily focused on the case of phase oscillators, which are linear dynamical systems. In contrast, we concentrate here on the more general case of nonlinear oscillators, and quantify in particular the effect of the oscillators' nonlinearities. The assumptions we consider are also different: while most existing approaches (including [37]) assume weak couplings and small noise intensities, we consider here strong couplings and arbitrary noise intensities.
The mechanisms highlighted in the paper may also underly other types of “redundant” calculations in the presence of noise and variability. In otoliths for instance, ten of thousands of hair cells jointly compute the three components of acceleration [38],[39]. In muscles, thousands of individual fibers participate in the control of one single degree of freedom. Similar questions may also arise in systems biology, e.g., in cell mechanisms of quorum sensing where individual cells measure global chemical concentrations in their environment in a fashion functionally similar to all-to-all coupling [25]–[27], in mechanical coupling of motor proteins [40], in the context of transcription-regulation networks [41],[42], and in differentiation dynamics [43].
Finally, the results point to the general question: what is the precise meaning of ensemble measurements or population codes, what information do they convey about the underlying signals, and is the presence of synchronization mechanisms (gap-junction mediated or other) implicit in this interpretation? As such, they may also shed light on a somewhat “dual” and highly controversial current issue. Ensemble measurements from the brain can correlate to behavior, and they have been suggested e.g. as inputs to brain-machine interfaces. Are these ensemble signals actually available to the brain [44], perhaps through some process akin to quorum sensing, and therefore functionally similar to (local) all-to-all coupling? Are local field potentials [45] plausible candidates for a role in this picture?
In the noise-free case (), it can be shown that, for strong enough coupling strengths, the elements of the network synchronize completely, that is, after exponential transients, we have in (A4) [6]. Thus, all the tend to a common trajectory, which is in fact a nominal trajectory of the noise-free system , because all the couplings vanish on the synchronization subspace.
In the presence of noise, it is not clear how to relate the trajectory of each to a nominal trajectory of the noise-free system. Nevertheless, we still know that the live “in a small neighborhood” of each other, as quantified by (A4). Thus, if the center of this small neighborhood follows a trajectory similar to a nominal trajectory of the noise-free system, then one may gain some information on the trajectories of the .
To be more precise, let be the center of mass of the , that is(6)Observe that, after expansion and rearrangement, the sum can be rewritten in terms of the distances of the from Using (A4) then leads to(7)Summing over the equations followed by the and using assumption (A1), we have(8)We now make the dynamics explicit with respect to by letting(9)so that Equation (8) can be rewritten as(10)Using the Taylor formula with integral remainder, we have(11)where is the gradient of or, equivalently, the vector of the Jacobian matrix of . Summing Equation (11) over and using assumption (A2), we get(12)Summing now inequality (12) over and using inequality (7), we get(13)which implies that when .
Turning now to the noise term in Equation (10), we have(14)since the intrinsic noises of the elements of the network are mutually independent.
Thus, for a given (even large) noise intensity , the difference between the dynamics followed by and the noise-free dynamics tends to zero when and . Assumption (A3) then implies that . More precisely, the impact of noise on the mean trajectory (quantified by ) evolves as(15)Finally, Equation (7) and the triangle inequality(16)imply that the trajectory of any synchronized element of the network and that of the noise-free system are also similar [compare Figure 1(A) and Figure 1(E)].
|
10.1371/journal.ppat.1004682 | A Crystal Structure of the Dengue Virus NS5 Protein Reveals a Novel Inter-domain Interface Essential for Protein Flexibility and Virus Replication | Flavivirus RNA replication occurs within a replication complex (RC) that assembles on ER membranes and comprises both non-structural (NS) viral proteins and host cofactors. As the largest protein component within the flavivirus RC, NS5 plays key enzymatic roles through its N-terminal methyltransferase (MTase) and C-terminal RNA-dependent-RNA polymerase (RdRp) domains, and constitutes a major target for antivirals. We determined a crystal structure of the full-length NS5 protein from Dengue virus serotype 3 (DENV3) at a resolution of 2.3 Å in the presence of bound SAH and GTP. Although the overall molecular shape of NS5 from DENV3 resembles that of NS5 from Japanese Encephalitis Virus (JEV), the relative orientation between the MTase and RdRp domains differs between the two structures, providing direct evidence for the existence of a set of discrete stable molecular conformations that may be required for its function. While the inter-domain region is mostly disordered in NS5 from JEV, the NS5 structure from DENV3 reveals a well-ordered linker region comprising a short 310 helix that may act as a swivel. Solution Hydrogen/Deuterium Exchange Mass Spectrometry (HDX-MS) analysis reveals an increased mobility of the thumb subdomain of RdRp in the context of the full length NS5 protein which correlates well with the analysis of the crystallographic temperature factors. Site-directed mutagenesis targeting the mostly polar interface between the MTase and RdRp domains identified several evolutionarily conserved residues that are important for viral replication, suggesting that inter-domain cross-talk in NS5 regulates virus replication. Collectively, a picture for the molecular origin of NS5 flexibility is emerging with profound implications for flavivirus replication and for the development of therapeutics targeting NS5.
| DENV causes widespread mosquito-borne viral infections worldwide and nearly 40% of the world’s population is at risk of being infected. Currently, no licensed vaccines or specific drugs are available to treat severe infections by DENV. NS5 is a large protein of 900 amino acids composed of two domains with several key enzymatic activities for viral RNA replication in the host cell and constitutes a prime target for the design of antiviral inhibitors. We succeeded in trapping a stable conformation of the full-length NS5 protein and report its crystal structure at a resolution of 2.3 Å. This conformation reveals the entire inter-domain region and clarifies the determinants of NS5 flexibility. The inter-domain interface is stabilized by several polar contacts between residues projecting from the MTase and RdRp domains of NS5. Several evolutionarily conserved residues at the interface play a crucial role for virus replication as shown by reverse genetics, although the analogous mutations mostly do not abolish the in vitro enzymatic activities of the recombinant proteins.
| Several flaviviruses such as Dengue virus (DENV), Japanese Encephalitis virus (JEV), West Nile virus (WNV), Yellow Fever virus (YFV) and Tick-Borne Encephalitis virus (TBEV) are major human pathogens. The mosquito-borne DENV serotypes 1–4 cause widespread epidemics and nearly 40% of the world’s population is at risk of being infected [1]. Infection by any of the four serotypes can lead to a broad spectrum of outcomes, ranging from asymptomatic infection, dengue fever, dengue hemorrhagic fever or dengue shock syndrome. A tetravalent vaccine is undergoing phase III of clinical trials, which requires three booster injections and only confers partial cross protection. No antivirals have been approved to treat Dengue so far, although availability of such molecules would be valuable to treat dengue infection.
Flavivirus RNA replication occurs within a multi-protein replication complex (RC), which assembles on ER-derived membranes and comprises both non-structural (NS) viral proteins and host cofactors [2–4]. With 900 amino acid residues, NS5 is the largest enzyme and the most conserved protein component of the flavivirus RC. Its N-terminal domain (residues 1–262 in DENV3) belongs to the S-adenosyl-L-methionine (SAM)-dependent methyltransferase (MTase) superfamily [5]. The MTase domain of NS5 caps the viral RNA genome, a step required for its stability and translation into viral polyproteins by host cell ribosomes [6]. Methylations of the N7 atom of Guanine-0, the 2’-O atoms of the ribose of Adenosine-1 and internal adenosines also contribute to viral escape from the host cell innate immune response [5] [7] [8] [9]. A putative guanylyltransferase activity (GTase) was also proposed for the N-terminal domain of NS5 [10,11] but this notion remains debatable. The C-terminal domain (residues 273–900) of NS5 contains the RNA-dependent RNA polymerase (RdRp) that synthesizes the anti-genome and progeny genome [12,13]. The RdRp domain is composed of the Finger, Thumb and Palm subdomains that are structurally conserved across viral RdRps [13]. Within the RdRp domain, residues 316–415 contain functional nuclear localization sequences that are hotspots for interactions with other viral and host proteins [14–17]. NS5 interacts with the NS3 protease-helicase and with several host proteins, including components of the ubiquitin proteasome pathway, NF90, and eEF1A [17–20]. In addition to its enzymatic functions required for RNA replication, NS5 acts as an antagonist of the host interferon response by interacting with and promoting the degradation of STAT2 [21]. In DENV, NS5 localizes to the nucleus of infected cells in a regulated and serotype-dependent manner that may modulate other host processes [22,23]. The importance of NS5 in viral replication and host immune response modulation makes it an ideal target for developing broad-acting antiviral inhibitors to treat diseases caused by flaviviruses [24–28]. A putative linker (or “inter-domain region” spanning residues 263–272) connects the two catalytic domains of NS5. Its sequence is relatively poorly conserved although its length has been preserved across flaviviruses.
Structural and functional studies of the complete NS5 polypeptide have been initiated by several research groups. Using in silico docking guided by reverse genetics, a first model for NS5 from WNV was proposed [12]. In this model, the MTase active site was placed in an orientation that would facilitate the capping of newly synthesized viral RNA exiting from the template-binding channel of the RdRp domain. The conformation of NS5 from DENV3 was also studied using small angle X-ray scattering (SAXS) [29] which suggested that the isolated NS5 protein from DENV may be flexible and could adopt a range of conformations, from compact (80% of the population) to more extended (20% of the population) structures in solution [29]. It was also reported that the MTase domain of NS5 does not affect its RdRp activity and that the two domains of NS5 responsible for RNA capping and synthesis respectively function essentially independently [30]. Recently, Lu and Gong reported a crystal structure for NS5 from JEV, in which the MTase and RdRp domains adopt a compact conformation stabilized by an interface dominated by hydrophobic interactions [31]. Moreover, a fragment of the RdRp from DENV3 NS5 containing extra residues from the linker has enhanced thermo-stability and polymerase activity, suggesting an impact of residues N-terminal to the polymerase domain on RdRp activity and stability [32]. This observation is supported by a recent study on viral RNA synthesis demonstrating that DENV2 NS5 has superior de novo and elongation activities than the NS5-RdRp domain and proposing that the MTase domain interacts with RdRp domain dynamically to regulate RNA synthesis during virus replication [33]. Clearly, more experimental data are needed to clarify the potential cross-regulatory effects between the two enzymatic domains of NS5, to reveal its distinct conformations, the molecular origin of its flexibility and to understand how its dynamic properties relate to the various steps required for RNA capping and synthesis. Here, we crystallized the full length NS5 protein from DENV3 and determined its crystal structure at 2.3 Å resolution, bound to SAH and also in the presence of GTP. Although the overall molecular shape of NS5 from DENV3 is similar to the NS5 protein from JEV, the relative orientation between their MTase and RdRp domains differs markedly between the two structures. The crystal structure of NS5 from DENV3 reveals a well-ordered linker region and a unique and mostly polar interface between the MTase and RdRp domains. The importance of selected interface residues in the structure and function of NS5 was addressed by site-directed mutagenesis for biochemical and virus replication studies. These studies corroborate the structural insight that inter-domain interactions are critical for viral RNA replication and infection.
BHK-21 cells (Baby Hamster Kidney fibroblast cells, ATCC) were maintained in RPMI 1640 media (Gibco), supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (P/S) at 37°C in 5% CO2.
To facilitate protein crystallization, we removed five flexible amino acids each from the N- and C- terminal ends of NS5, based on inspection of missing electron density in the structures of the individual MTase and RdRp domains from DENV3 (PDB codes 3P97 and 2J7U). The fragment corresponding to amino acid residues 6 to 895 of DENV3 NS5 (GenBank accession number AY662691.1) was amplified and cloned into pNIC28-Bsa4 plasmid using ligation independent cloning methods [34]. The expressed construct comprises the hexa-histidine tag and a Tobacco Etch Virus (TEV) protease cleavage site fused at the N terminus of DENV3 NS5. Transformed E. coli BL21-CodonPlus (Stratagene) was grown in LB medium containing 50 μg/ml kanamycin and 34 ug/ml chloramphenicol to an A600nm of 0.6–0.8 at 37°C. Expression of the recombinant proteins was induced by the addition of 0.5 mM isopropyl β-D-galactoside, and incubation was continued for a further 20 h at 18°C. The cells were rapidly cooled and harvested by centrifugation at 8000 g for 10 min at 4°C and stored at -20°C. Cells were resuspended in buffer A (20 mM Tris-HCl, pH 7.5, 500 mM NaCl, 5mM β-mercaptoethanol, 10% glycerol, 10 mM imidazole) and lysed by sonication. The lysate was clarified by centrifugation at 20,000 g for 60min at 4°C. The supernatant was purified by nickel-nitrilo-triacetic acid affinity chromatography by washing unbound protein with buffer A supplemented with 40 mM imidazole. The DENV3 NS5 protein was eluted using a linear gradient of imidazole ranging from 40 to 500 mM. Fractions containing His6-NS5 were dialyzed against buffer C (20 mM Na-Hepes, pH 7.5, 300 mM NaCl, 5 mM DTT, 10% (v/v) glycerol), and the proteins were cleaved by TEV (substrate-to-enzyme ratio of 40:1, w/w) at 4°C for overnight. NS5 was further purified by size-exclusion chromatography using a HiPrep Superdex-200 gel filtration column (Amersham Bioscience) in buffer C. Fractions containing NS5 were pooled and concentrated to ~10 mg/ml before storage in -80°C. SDS-PAGE analysis of the resulting NS5 protein indicated a purity of 95%.
Crystallization trials were set up at 20°C with a Phoenix crystallization robot using sitting drop vapour diffusion. After extensive robotic crystallization trials, small rhomboidal shaped crystals were obtained using precipitants containing divalent metal salts, including calcium acetate, magnesium acetate and magnesium formate. Crystallization conditions include conditions 18 and 46 (Crystal Screen HR2–110, Hampton Research), condition 20 (PEG/Ion Screen, HR2–126, Hampton Research) and condition 55 (Index screen HR2–144, Hampton Research). Larger crystals of rhombus or trapezoid shape were obtained over 2–5 days by mixing a volume of 1 μl of NS5 (6–895) at 4–6 mg/ml with 1 μl of precipitation solution (0.2 M calcium acetate or magnesium acetate, 0.1 M sodium cacodylate, pH = 6.4, 10–20% (w/v) PEG 8000). Crystals for the GTP complex were obtained by either co-crystallization of NS5 at a concentration of 6 mg/ml (~60 μM), with 0.4 mM GTP using a slightly different precipitating solution (0.1 M sodium cacodylate (pH 6.4), 0.2 M magnesium acetate or 0.2 M calcium acetate and 14% (w/v) polyethylene glycol 8000) at 18°C, or soaking of apo NS5 crystal in the precipitating solution supplemented with 5 mM GTP overnight. Although these crystals grow in slightly different conditions, they are isomorphous. Prior to data collection, crystals were soaked for a few seconds in a cryoprotecting solution containing 20% (v/v) glycerol before being mounted and cooled to 100 K in a nitrogen gas stream (Oxford Cryosystems). Diffraction intensities were collected at the PXIII (X10SA) beamline of the Swiss Light Source, Paul Scherrer Institut, Villigen, Switzerland. Diffraction intensities for cocrystals with GTP were collected at the National Synchrotron Radiation Research Center, beamline 13B1 (Hsinchu, Taiwan). The best diffraction data were obtained for crystals of NS5 grown in the presence of calcium acetate. Integration, scaling, and merging of the intensities were carried out using programs MOSFLM and SCALA from the CCP4 suite [35]. The asymmetric unit contains one NS5 molecule with S-adenosyl-L-homocysteine (SAH) (copurified from E. coli) bound to the MTase domain. Crystal parameters and data collection statistics are summarized in Table 1.
The crystals have an estimated solvent content of 50.2% based on a Matthews coefficient (Vm) value of 2.47 [36]. The structure was solved by molecular replacement with the program PHASER [37] using the DENV3 MTase (Protein Data Bank [PDB] accession code 3P97) [38] and DENV3 RdRp (PDB accession code 2J7U) [13] as search probes. Refinement cycles carried out using REFMAC5 [39] and PHENIX [40] were interspersed with manual model rebuilding sessions using Coot [41]. TLS refinement was introduced in the last refinement steps. The quality of the structure was analyzed using Molprobity [42]. A summary of structure refinement statistics is given in Table 1. Solvent-accessible surfaces areas were calculated using CCP4 program AREAIMOL with a 1.4-Å radius sphere as a probe. Superimpositions of structures were carried out using the program LSQKAB from the CCP4 suite. Figures were prepared using the program Pymol [43]. In the crystal structure of NS5 with GTP bound to the MTase domain (Table 1), the overall conformation of NS5 is very similar to the NS5: SAH structure: The root mean square deviation (r.m.s.d.) is 0.17 Å for a total of 719 Cα atoms superimposed. Buried surface areas were calculated with the PISA server (www.ebi.ac.uk/pisa) using the following domains definition: MTase residues 6–262 in DENV3 and 6–265 in JEV; RdRp residues 273–883 in DENV3 and 276–896 in JEV (S2 Table) [44]. The relative reorientation between domains was assessed using the program DynDom [45]. The refined coordinates were deposited in the PDB under accession codes 4V0Q (NS5: SAH) and 4V0R (NS5: SAH: GTP).
The buffer for the HDX reaction was of the same composition (20 mM Na-Hepes, pH 7.5, 300 mM NaCl, 5 mM DTT, 10% (v/v) glycerol) except that H2O was replaced with D2O (99.99%) and glycerol was omitted. Essentially the deuterium exchange reactions for NS5, or the individual MTase or RdRp domains was carried out by mixing a volume of 4 μl of the respective protein solutions (at a concentration of 18 μM), with 16 μl D2O or H2O buffer and incubated at 4°C for 0s, 30s, 60s, 600s, 1800s. The deuterium exchange (final D2O concentration 80%) for each sample at the various time points was quenched by the addition of 20 μl ice-cold solution consisting of 1 M guanidine hydrochloride and 1.5% (v/v) formic acid followed by rapid freezing in liquid nitrogen and storage at -80°C until HDX MS analysis.
For capillary-flow LC, buffer A was H2O containing 0.3% (v/v) formic acid. Buffer B was acetonitrile containing 0.3% (v/v) formic acid. Deuterium exchanged protein samples from above were then digested online by passing through an immobilized pepsin-coupled column (2.1 mm i.d. × 30 mm) (Invitrogen) at 16°C and were de-salted for 3 min on a home-packed C4 trap (0.75 mm i.d. × 10 mm, C4 beads purchased from Michrom) with buffer A (H2O containing 0.3% (v/v) formic acid) at a flow rate of 150 μl min-1 driven by the LC loading pump (Dionex 3000 RSLC). The peptide samples were separated further on a home-packed C4 column (0.3 mm i.d. × 50 mm, C4 beads purchased from Michrom) using a 20 min gradient (5% to 35%) in buffer B (acetonitrile containing 0.3% (v/v) formic acid), followed by washing with 90% buffer B for 3 min and equilibration with 1% buffer B for 5 min) at a flow rate of 15 μl min-1 driven by LC NC pumps. MS raw data were acquired in the range of m/z 300–2000 for 30 min in positive mode on a LTQ-Orbitrap mass spectrometer (Thermo Fisher Scientific) equipped with an ESI source (capillary temperature 275°C and spray voltage of 5 kV). All the HDX separations were performed at 0°C except for on-line pepsin digestion that was carried out at 16°C. Blank injections were made between runs to remove carry-over peptides. At least three independent deuterium exchange experiments were carried out for each time point. All HDX data were normalized to 100% D2O content, corrected for an estimated average deuterium recovery of 70%, and analyzed by the software HD Desktop [46]. Initial peptic peptide identifications were performed with the same HDX set up as described above. 4 μl of protein sample (20 μM) was injected into the HDX MS system. Product ion (MS/MS) spectra were acquired in linear ion trap LTQ with eight most abundant ions selected in the precursor (MS) scan with a 7.5 sec exclusion time. MS and tandem MS files were extracted and searched by using the Global proteome machine (http://www.thegpm.org) for high-confident peptide identification.
The K95A, Y119A, R263A (R262 in DENV3), E268A (E267 in DENV3), E270A (E269 in DENV3), and R353A (R352 in DENV3) mutants in the DENV4 NS5 sequence (GenBank accession number AF326825) were engineered into the subclone, pACYC-DENV4-F shuttle, using the QuikChange II XL site-directed mutagenesis kit according to the manufacturer’s protocol (Stratagene). This plasmid harbours nucleotides 7564–10653 (from NS3–3’UTR) from the DENV4, MY01–22713 strain, linked at the 3’ end to the Hepatitis D virus ribozyme (HDENVr) sequence [47]. Following sequence verification, the plasmids were digested with NotI and KpnI and inserted with a PCR product comprising the sequence spanning nucleotides 1–7563 under the control of the upstream T7 promoter in which the region from nucleotides 217–2291 in this cDNA had been replaced by renilla luciferase and foot-and-mouth disease virus 2A protease cDNAs [38]. A list of primers used for cloning and mutagenesis is available in S1A Table is S1 Text.
After linearization of the corresponding replicon cDNA plasmids with XhoI, in vitro transcription (IVT) was performed using a T7 mMESSAGE mMACHINE kit according to the manufacturer’s protocol (Ambion; Austin, Texas, USA). IVT RNAs (10 μg) were electroporated into 8×106 BHK-21 cells, after which the cells were re-suspended in a volume of 25 ml of DMEM with 10% FBS. Cell suspensions of 0.5 ml per well were seeded into 12-well plate; and the cells were assayed for luciferase activities at 4, 24, 48, 72, and 96 h post-transfection. Duplicate wells were seeded for each data point. Luciferase assays were performed using a RenLUC assay system following the manufacturer’s protocol (Promega, WI, USA).
NS5 (serotype DENV4) WT and mutant cDNAs were amplified by PCR from the respective WT and mutant pACYC-DENV4-F shuttle plasmids and cloned into the pET28a vector (Stratagene) using the NheI and XhoI sites. Protein expression and heat stability analyses by thermo-fluoresence were performed as described previously [32].
N-7 and 2’-O MTase assays were performed as described [38,48]. In general, the N7-MTase reaction comprised 25 nM protein, 240 nM biotinylated GpppA-DENV nt 1–110 in vitro transcribed RNA, 320 nM [3H-methyl]-SAM in 50 mM Tris-HCl, pH 7.5, 20 mM NaCl, and 0.05% (v/v) CHAPS. The 2’-O MTase reaction comprised 25 nM protein, 40 nM GpppA-7mer RNA (Trilink), 320 nM [3H-methyl]-SAM in 50 mM Tris-HCl, pH 7.5, 10 mM KCl, 2 mM MgCl2, and 0.05% CHAPS. Buffer, RNA substrate, and enzyme were first mixed in a single well in a 96-well half- area, white opaque plates (Corning Costar, Acton, MA), and the reaction was initiated by addition of [3H-methyl]-SAM. The N-7 and 2’-O reactions were incubated at RT for 15 min and 1 hr, respectively. The reaction was stopped with 25 μl of 2× stop solution (100 mM Tris/HCl, pH 7, 100 mM EDTA, 600 mM NaCl, 4 mg/ml streptavidin—SPA beads and 62.5 μM cold AdoMet) and shaken for 20 min at 750 rpm at room temperature followed by centrifugation for 2 mins at 1200 rpm. The plate was read in a Trilux microbeta counter (PerkinElmer, Boston, MA) with a counting time of 1 min per well. All data points were collected in duplicate wells.
The de novo initiation/elongation assay was described [28,49]. Briefly, the reaction comprised 100 nM DENV4 NS5, 100 nM in vitro transcribed DENV4 5´UTR-3´UTR RNA, 20 μM ATP, 20 μM GTP, 20 μM UTP, 5 μM Atto-CTP (Trilink Biotechnologies), in a volume of 15 μl of the assay buffer comprising 50 mM Tris-HCl, pH 7.5, 10 mM KCl, 1 mM MgCl2, 0.3 mM MnCl2, 0.001% Triton X-100 and 10 μM cysteine. The elongation assay reaction comprised 100 nM IVT 244 nt heteropolymeric RNA template, annealed with four primers (C1 primer 3’-AGTCAGTCAGTCAGTGT-biotin-5’, A1 primer 3’-GTCAGTCAGTCAGTCTC-biotin-5’, G1 primer 3’-TCAGTCAGTCAGTCACA-biotin-5’, T1 primer 3’-CAGTCAGTCAGTCAGAG-biotin-5’) [50], 2 μM ATP, 2 μM GTP, 2 μM UTP, 0.5 μM Atto-CTP, and 100 nM of DENV4 NS5 in 15 μl in assay buffer comprising 50 mM Tris-HCl at pH 7.5, 10 mM KCl, 0.5 mM MnCl2, 0.01% Triton X-100, and 10 μM cysteine. RNA was separately pre-annealed to the four different sets of primers at a ratio of 1:2 (w / w) by heating at 95°C for 3 min, and cooled to RT before mixing and used for the assay. All reactions were allowed to proceed for up to 3 hrs at RT. At the indicated time-points, 10 μl of 2.5× STOP buffer (200 mM NaCl, 25 mM MgCl2, 1.5 M DEA, pH 10; Promega) with 25 nM calf intestinal alkaline phosphatase (CIP; New England Biolabs) was added to the wells to terminate the reactions. The plate was shaken and centrifuged briefly at 1200 rpm, followed by incubation at RT for 60 min and the released AttoPhos was monitored by reading on a Tecan Saffire II microplate reader at excitationmax and emissionmax wavelengths 422 nm and 566 nm respectively. All data points were collected in triplicate wells in 384-well black opaque plates (Corning).
Mutations of K95A, Y119A, E268A (E267 in DENV3) and R353A (R352 in DENV3) were constructed into full-length DENV2 cDNA clone (GenBank accession EU081177.1) using the subclone pWSK29 D2 fragment 3 as described [17]. Briefly, mutations were generated using the QuikChange II XL site-directed mutagenesis kit (Stratagene) according to the manufacturer’s protocol. Primers used for mutagenesis are listed in S1D Table in S1 Text. Fragment 3 bearing the mutation was excised from the plasmid by XbaI and SacI, and inserted into subclone pWSK2 D2 fragment 1+2 that was similarly cut with XhaI and SacI.
The genome-length wild-type DENV2 infectious clone plasmid pWSK29 (GenBank accession EU081177.1) and the mutants were linearized with SacI and purified with phenol-chloroform. A quantity of ~2 μg linearized DNA was used for in vitro transcription with T7 mMESSAGE mMACHINE kit (Ambion). BHK-21 cells were trypsinized, washed twice with cold PBS and resuspended in Opti-MEM (Gibco) at a cell density of 1 × 107 cells/ml. 10 μg of in vitro transcribed RNA (wild-type DENV2 and mutants) were mixed with 800 μl cell suspension in a pre-chilled 0.4 cm cuvette, and electroporated at settings of 850 V and 25μ F, 2 pulses at 3 seconds intervals. Electroporated cells were allowed to recover at room temperature for 10 minutes prior to resuspending in complete RPMI 1640 media. 3 × 105 cells were then seeded into a 12-well plate and incubated at 37°C in the presence of 5% CO2. Media was changed to 2% FBS maintenance media after 6 hours post-transfection. Samples were harvested every 24 hours post-transfection until 120 hours. Supernatants were harvested and clarified for determination of virus titer by standard plaque assay on BHK-21 cells and extracellular viral RNA quantification by real time RT-PCR. Cells were then washed once with PBS before lysing with Trizol reagent (Invitrogen) for total intracellular viral RNA quantification.
For extracellular viral RNA quantification, viral RNA was extracted from the supernatant using QIAamp Viral RNA Mini Kit (Qiagen) according to the manufacturer’s instructions. Real time RT-PCR was carried out in Bio-Rad Real time thermal cycler CFX96 by the use of iTaq Universal SYBR Green One-Step kit (Bio-Rad) with primers (forward, 5’- CAGGCTATGGCACTGTCACGAT-3’; and reverse, 5’-CCATTTGCAGCAACACCATCTC-3’) targeting for DENV2 E region (adopted from [51]). A plasmid fragment containing genome sequences of the E region was used to establish a standard curve for quantification of viral genome copy number per ml of supernatant.
For cellular viral RNA quantification, total RNA was extracted using Trizol extraction method and 40 ng RNA was subjected to real time RT-PCR as described above. Absolute copies of the viral RNA genome was normalized to actin expression with β-actin primers (forward, 5’- TCACCACACACTGTGCCCATCTACGA-3’ and reverse, 5’-CAGGGGAACCGCTCATTGCCAATGG-3’); and reported as viral RNA genome copy per μg RNA.
IFA against dsRNA with anti-dsRNA mAb J2 (Scicons) and NS5 protein with anti-NS5 hAb 5R3 were performed as described [52]. Images were captured on an inverted fluorescence microscope (Olympus IX71, Center Valley, USA) at 20× magnification and image analysis was done with ImageJ software.
Overall, the polypeptide chain of NS5 is well defined in the electron density map, including for inter-domain residues 262–273 (Fig. 1 and Table 1). The NS5 protein from DENV3 adopts a compact shape with overall dimensions 87 Å × 72 Å × 55 Å (Fig. 1A and B). The MTase domain is located above the fingers subdomain of the RdRp (Fig. 1A), such that the GTP binding pocket, the K61D146K180E216 catalytic tetrad and the SAH binding pockets of the MTase domain are positioned away from the inter-domain interface. In this conformation, the tunnels of the RdRp domain that permit the entry of NTPs and RNA template remain accessible. The MTase-RdRp inter-domain interface is comprised of two contact areas involving linker residues, the fingers subdomain from the RdRp and three segments (residues 63–69, 95–96 and 252–256) from the MTase domain (Fig. 1). The network of interactions stabilizing the interface mainly comprises charged residues (Fig. 1; S2 Table in S1 Text). In the first cluster, linker residues interact with both the RdRp and MTase domains. The side chain of E267 hydrogen bonds with residues Y119 and R262 from the MTase domain, while E269 forms a salt bridge with the guanidinium group of R361 from the RdRp domain (Fig. 1C). Several polar interactions are also formed between residues K95-K96 from the MTase domain and E296-K300 from the RdRp domain (Fig. 1C; S2 Table in S1 Text). The second cluster of inter-domain interactions is centered at helix α5 (residues F348-K357), which belongs to the evolutionary conserved bNLS motif of the RdRp domain, (Fig. 1B and D). Projecting from helix α5, the guanidinium side chain of R352 plays a central role via the formation of numerous contacts including electrostatic interactions with E67, E252 and Q63. Another salt bridge is formed between K357 and D256 (Fig. 1D). In addition, several water molecules trapped at the interface mediate H-bond interactions between residues projecting from both domains (Fig. 1C and D). In contrast to multiple non-polar contacts established between the MTase and RdRp domains of NS5 from JEV, the only hydrophobic contacts between the two domains of NS5 from DENV3 involve stacking interactions between W64, R68, F348 and P582 (Fig. 1D).
Despite an extensive inter-domain interface (Fig. 2), the MTase and RdRp domains of NS5 display only minimal differences at surface exposed regions compared to the individual enzymatic domains (S3 Fig in S1 Text). Superimposition with the NS5-MTase domain structures (PDB codes 3P97 and 1L9K) returns RMSD values of 0.42 Å and 0.40 Å for 221 Cα and 219 Cα atoms respectively [5,38]. In the MTase domain, the co-purified (SAH) molecule resides in the cofactor-binding pocket and is stabilized by a network of hydrogen bonds and van der Waals contacts (Fig. 1; S1 Fig in S1 Text) essentially identical to those reported earlier [5,53,54]. GTP is stabilized by base stacking with the aromatic ring of F25 and electrostatic interactions with residues L17, N18, and L20 of the MTase domain (S1 Fig in S1 Text) [5,54]. Likewise, superimposition with the DENV3 NS5-RdRp (PDB codes 2J7U and 4A11) yields RMSD values of 0.47 Å and 0.43Å for 492 and 495 Cα atoms respectively [13,32]. Like in the isolated RdRp domain structure reported earlier [13], however, two segments of the polypeptide chain are disordered in the present NS5 structure: residues 406–417 that harbor motif G, which regulate access of the ssRNA substrate to the template tunnel [13] and residues 455–468 which contain motif F. The latter were proposed to bind Stem Loop A, the viral promoter, prior to viral RNA replication [55]. Two Zinc ions bound to the RdRp domain are also observed as reported earlier [12,13].
In order to determine whether the conformation of NS5 captured in the crystal state agrees with that in solution and also to better understand the dynamics of NS5, we performed Hydrogen/Deuterium Exchange Mass Spectrometry (HDX-MS) experiments using NS5, NS5-MTase or NS5-RdRp (Fig. 2A; S2 Fig in S1 Text). HDX-MS results are influenced by protein exposure to solvent and stability in solution, making this technique suitable to study solution-state protein dynamics [17,56–59]. The HDX-MS heat maps were overlaid onto the three respective crystal structures, with a color code indicating the extent of deuterium exchange with the protein backbone amide hydrogen atoms. The missing residues 406–417, 455–468, and 884–891 were modeled based on the homology to JEV NS5 structure. The corresponding HDX-MS profiles of peptic peptides across NS5 are listed in S2A Fig in S1 Text. The interface peptides from the MTase (residues 65–77 and 95–118) and the RdRp (residues 291–301 and 581–587) domains of NS5 display comparable deuterium exchange (Fig. 2A, boxed area; S2E and S2F Fig in S1 Text). However, the extent of deuterium exchange of these peptides (~ 20–30%) is not as low as the domain core region (~ 10% in blue), suggesting that the interface is solvent accessible as indicated by the crystal structure (Fig. 2A; S2 Fig in S1 Text).
Interestingly the HDX-MS also reveals the local protein dynamics of NS5. Peptides from the thumb subdomain (I735-L748, S781-M809, V812-L850, and L853-S895) and the middle finger (K311-M340) of the RdRp domain display higher deuterium-exchange (Fig. 2A; S2 Fig in S1 Text) in NS5 compared with the isolated NS5-RdRp domain. This suggests that in the context of the full length NS5 protein, the thumb subdomain of RdRp is more dynamic in solution than its fingers and palm subdomains. Remarkably the dynamics measurements with HDX-MS correlated very well with the distribution of crystallographic temperature factors observed for the RdRp domain of NS5, where higher temperature factors for the thumb subdomain than the fingers and palm subdomains are observed (Fig. 2B). Likewise, residues containing motifs F and G that are disordered in the crystal structure of NS5, also display a higher degree of amide hydrogen exchanges in HDX-MS (Fig. 2A; S2 Fig in S1 Text) [13,32]. It is worth noting that the conformation adopted by motifs F and G in the crystal structure of the isolated RdRp domain is different in the NS5 protein from JEV (S4 Fig in S1 Text). Overall, the HDX-MS experiment indicates that the conformation adopted by DENV3 NS5 in the crystal structure is also accessible in solution but that other conformations are possible.
Despite a similar molecular shape maintained by intra-molecular interactions between the MTase and fingers subdomain of the RdRp, both the relative orientation and the interfaces between domains differ between the NS5 proteins from JEV and DENV3 (Fig. 3) [31]. Following superimposition of their RdRp domains, an additional rotation of 105° of their MTase domain is needed to bring them to coincidence (Fig. 3; S1 Movie). The inter-domain linker regions also differ markedly between the two proteins: In DENV3 NS5, the linker residues are well ordered (Fig. 3A and C), while in the JEV NS5 structure, residues A266-G270 adopt an extended conformation and residues E271-V272-H273 are disordered (Fig. 3B; S4 Fig in S1 Text) [31]. Since the linker residues are well ordered with clear electron density in the present DENV3 NS5 crystal structure (Fig. 3C), a precise definition of the structural determinants contributing to inter-domain flexibility is now possible. Earlier crystal structures of NS5 individual domains suggested that the evolutionarily conserved R262, whose guanidinium group hydrogen bonds with the main chain carbonyl oxygen of V97, defines the C terminus of the MTase domain (Fig. 1C; S3 Fig in S1 Text) [1,28,32,57,60]. Residues H263-V264-N265-A266, which are located after the C-terminal end of the MTase domain, are the least evolutionarily conserved in the interface region (Fig. 1E and 3D). Interestingly, these residues fold into a short 310-helix giving the polypeptide chain a compact conformation and allowing the two domains of NS5 to form a large interface leading to its observed globular shape (Figs. 1 and 2). Conversely, the segment E267-N272 forms extensive interactions with both the MTase and RdRp domains (Fig. 1C and 3C). The side chain of E267 projects towards the MTase domain forming hydrogen bonds with R262 and Y119, while the side chain of E269 projects towards the RdRp domain and interacts with R361 and K595. The two acidic residues E267 and E269 are highly conserved among all four dengue serotypes, but not in other flaviviruses (Fig. 1E and 3D). Likewise, T270 and N272 establish several polar contacts with the RdRp domain (Fig. 1C). To further delineate the boundaries of the RdRp domain of NS5, we observe that residues 268–271 adopt the same conformation in both the DENV3 NS5-RdRp and in the present NS5 structures (S3 Fig in S1 Text) [32]. Coincidentally, residues 268–272 enhance the thermostability of the truncated NS5-RdRp and also its polymerase activity [32] suggesting that they may be considered as an integral part of the RdRp domain of NS5 (Fig. 3D; S3 Fig in S1 Text). We conclude that DENV NS5 contains a short four-amino-acids long inter-domain linker (residues 263–266). Like a swivel, it is conceivable that structural transitions at this short linker—for instance a conversion from a helix to more extended conformations—are accompanied by important changes in the relative orientation between the two domains of NS5 [61].
The total buried interdomain surface area calculated for DENV3 NS5 is 1502 Å2 which is only slightly larger than that for JEV NS5 (1464 Å2) (Domain definition: MTase residues 6–262 in DENV3 and 6–265 in JEV; RdRp residues 273–883 in DENV3 and 276–896 in JEV) (S2A Table in S1 Text) (http://www.ebi.ac.uk/pdbe/pisa/) [44]. These values suggest that both structures are relatively stable in solution [62]. This is consistent with the rather compact shapes adopted by NS5 with maximum molecular dimensions of approximately 87 Å (DENV3) and 100 Å (JEV) (Fig. 3A and B). Interestingly, major inter-domain interactions in DENV3 NS5 are of polar nature, occurring both directly between charged side-chains and through water-mediated contacts (Fig. 4A; S2B Table in S1 Text). In contrast, the inter-domain interactions in JEV NS5 are predominantly hydrophobic, including residues P113, L115, W121 from MTase and F467 (F464 in DENV3), F351(F348 in DENV3) and P585 (P582 in DENV3) from the RdRp domain (Fig. 4B; S2B Table in S1 Text) [31]. An analysis of interface residues reveals that residues in both interfaces were highly conserved during evolution (Fig. 4C and D) [63]. When superposing MTase domains of the two NS5 structures, the relative positions of the catalytic centers of the MTase domain (KDKE catalytic tetrad) and the RdRp domain (GDD, motif C) are far apart (Fig. 4E). It is thus evident that both DENV and JEV RdRp domains used the finger subdomains to interact with their respective MTase domains from two non-overlapping sides, thus establishing unique interfaces for DENV or JEV (Fig. 4).
To examine the functional importance of individual residues within the two clusters of interactions at the DENV NS5 inter-domain interface (Fig. 1C and D), we introduced single alanine mutations at residues K95, Y119, R263 (R262 in DENV3), and R353 (R352 in DENV3) into the DENV4 NS5 protein and compared their polymerase activities against the WT protein (Table 2). Note also that the residues selected for mutagenesis are conserved across the four DENV serotypes but not across other flaviviruses, except for R263 (R262 in DENV3) and R353 (R352 in DENV3) (Fig. 1E). All purified mutant NS5 proteins retained similar stability as WT as determined by a thermofluor assay, suggesting a native fold (Table 2). The RdRp activity of the mutant proteins compared to WT enzyme was assessed by both the de novo initiation/elongation assay and the elongation assay as reported [32]. The former uses the viral UTR sequence as template to start a new RNA chain while the latter employs an heteropolymeric RNA template annealed with four primers, which is extended during the course of the assay [28,49].
Overall, with the exception of E270A (E269 in DENV3) that exhibits about 30–60% reduction in both polymerase activity, none of the NS5 mutants exhibit significant impairment in their de novo initiation or elongation activities. Mutants Y119A, R263A (R262 in DENV3), and E268A (E267 in DENV3) display a 24–50% increase in de novo initiation/elongation activities compared to WT NS5 (Table 2). The alanine mutation at K95 results in the most active RdRp so far with an almost 2 fold increase in both the de novo initiation/elongation and elongation activities, compared to WT. Residue R353 (R352 in DENV3) is located at the second interface cluster and belongs to a 20-amino-acids sequence conserved in all flavivirus NS5 proteins [15] and, as mentioned earlier, it forms part of the bNLS. The de novo initiation/elongation activity is 2-fold higher (similar to K95A) but the primer elongation activity is similar to WT (Table 2).
Next, we investigated the effect of the mutations on the N7 and 2’-O MTase activities of NS5 (Table 2). Mutation of residues Y119 and R263 (R262 in DENV3) to alanine strongly impacted both N7 and 2’-O MTase activities. Mutant Y119A exhibited less than 50% N7 activity compared to WT protein and minimal 2’-O activity. Both enzymatic activities were almost abolished in R263A (R262 in DENV3), the last residue of the MTase domain. On the other hand, E268A (E267 in DENV3) and E270A (E269 in DENV3)—mutants C-terminal to the inter-domain linker, retained 65–74% of both MTase activities, suggesting minor allosteric effects from the RdRp domain transmitted via the inter-domain interface. In contrast to the elevated polymerase activities, the K95A and R353A (R352 in DENV3) mutants that disrupt the transverse polar contacts appears to have little impact on the MTase activities (Table 2).
We next assessed whether residues involved in MTase-RdRp interactions are important for viral replication by introducing alanine mutations at positions K95, Y119, R263 (R262 in DENV3), E268 (E267 in DENV3), E270 (E269 in DENV3), and R353 (R352 in DENV3) in a DENV4 subgenomic RNA replicon, bearing a renilla luciferase reporter (Figs. 5A and 6) [32]. Both WT and mutant DENV4 replicon cDNAs were transcribed in vitro, electroporated into BHK-21 cells and luciferase activities were monitored at 4, 24, 48, 72 and 96 hr post-electroporation (Fig. 5A). The luciferase activity in cells harbouring WT DENV4 replicon showed a typical pattern, where a peak level is reached at approximately 24 to 48 hr post-electroporation (about 3320-fold above background levels), which steadily declined from 48–96 hr post-electroporation (Fig. 5A). Little or no luciferase activity was detected for DENV4 mutant replicons Y119A, R263A (R262 in DENV3), and E270A (E269 in DENV3), throughout the four days post-electroporation. Mutant replicons E268A (E267 in DENV3) and R353A (R352 in DENV3) exhibited significant luciferase activity that are almost comparable to WT at 48 hr point, although the mutants had some delay at the first 24 hrs (Fig. 5A).
To further validate the biological significance of the interface, four residues including K95, Y119, E268 (E267 in DENV3), and R353 (R352 in DENV3) were selected and mutated into alanine in the DENV2 infectious clone (Fig. 5B, C, and D). Defective mutant K330A, which has previously been reported, was included as a negative control [14,17]. E268A mutant (E267 in DENV3) displayed similar growth kinetics, extracellular viral RNA level and infectious virus recovery compared to WT. Y119A mutation did not yield any viable virus or increased intracellular RNA quantified by qRT-PCR. This is likely due to the defective MTase activity (Fig. 5; Table 2). The K95A mutant virus replicated much slower than WT with a 10-fold lower intracellular and extracellular RNA level at 72 hr post electroporation. No RNA replication or virus recovery was detected for mutant R353A (R352 in DENV3). Immuno-fluorescence analyses at 72 hr using anti-dsRNA and anti-NS5 antibodies confirmed the intracellular levels for WT and mutant viruses (Fig. 5E). Taken together, these data demonstrated: (i) most interestingly, conserved residues that form the major inter-domain polar contacts, K95 and R353 (R352 in DENV3), play critical but non-enzymatic roles in virus RNA replication and infectivity; (ii) alanine mutation of E268 (E267 in DENV3) have no impact on the enzymatic activities of NS5 nor on virus RNA replication and infectivity (iii) Within the MTase domain, Y119, is important for the MTase enzymatic activity of NS5 and is therefore essential for the virus replication.
The present 3D structure of DENV3 NS5 at 2.3 Å provides the first view of a flavivirus NS5 protein showing how its MTase and RdRp domains are physically connected through a fully resolved linker region (Fig. 1 and 3). Based on the present study and recent work [32], we conclude that DENV NS5 contains a four amino-acids long inter-domain linker (residues 263–266), which is able to fold into a short 310-helix as observed in the present structure, but which can probably also adopt extended conformations (Fig. 6). In contrast, the JEV NS5 structure reveals a partially resolved linker region (266–275) which neither forms any secondary structure nor establishes interactions with either enzymatic domain (Fig. 3; S3 and S4 Fig in S1 Text) [31]. The low level of sequence conservation and the diverse conformations observed for the linker region strongly suggest conformational plasticity at the inter-domain interface for the flavivirus NS5 protein (Figs. 1 and 3D; S1 Movie). About 13 Å shorter in its largest dimension, DENV3 NS5 displays a more compact conformation than JEV NS5 (Fig. 3). This is reflected by the differences in the spatial arrangement of the two enzymatic domains (Fig. 4E). These different relative orientations are likely to result from different secondary structures adopted by the linker region that acts like a swivel, leading to the formation of different interfaces and enabling the two enzymatic domains to adopt various relative orientations during the transition from RNA synthesis to RNA capping (Figs. 3 and 6).
Remarkably, the interface in DENV3 NS5 is mostly polar with several salt bridges and fewer hydrophobic stacking interactions than that of JEV NS5, which is mainly of hydrophobic nature (Fig. 4A and B) [60]. The interface peptides from MTase domain and the fingers subdomain of RdRp displayed similar hydrogen deuterium exchange patterns, which suggest that the crystal structure of NS5 represents a predominant conformation in solution (Figs. 1 and 2A). However, given the polar nature of the interface interactions, conformational flexibility and heterogeneity are likely. An earlier SAXS study also suggested that the interaction between the MTase and RdRp domains may be heterogeneous: approximately 80% of DENV3 NS5 were found to adopt a relatively compact structure in which the two domains interact with each other whilst the remaining 20% of the population were in more extended conformations [29]. Therefore the two enzymatic domains of DENV NS5 may be able to associate and dissociate from each other with a relatively low energy barrier, consistent with a limited inter-domain buried surface area and the predominance of polar interactions. The prevalence of a more compact conformation in solution is consistent with our observation of a relatively short but structured four-residue-linker (residues 263–266 in DENV3 NS5) that restricts inter-domain movements. Furthermore, the solution HDX-MS is consistent with the temperature factor analysis of the DENV3 NS5 structure, indicating that the RdRp domain of DENV3 NS5 adopts a more flexible conformation especially in its thumb sub-domain (Fig. 2). These observations rationalize the contributions of a covalently tethered MTase domain to the RNA synthesis activities by RdRp [33]. In addition, the dynamic nature of the DENV NS5 protein may also facilitate the recruitment of other viral or host proteins. In agreement with this proposal, we have shown previously that K330 within the α3-helix of RdRp is critical for binding to the NS3 helicase. Mutation of K330 to alanine did not inhibit the de novo initiation and elongation activities in RdRp nor the 2’O and N-7 MTase activities but completely abolished replication in an infectious clone [14].
Using biochemical and reverse genetic tools—enzymology of MTase and RdRp activities for DENV4 NS5, DENV4 replicon and DENV2 infectious virus, we examined the functional relevance of the interface interactions to viral RNA replication and infectivity by structure-guided mutagenesis. When either the MTase (Y119A and R263A (R262 in DENV3) in DENV4) or the polymerase activities (E270A in DENV4 (E269 in DENV3)) were severely inhibited (over 50% reduction compared to WT), neither viral RNA nor infectious virus could be observed (Fig. 5A; Table 2), Mutation of key polar contacts at the interface (K95A and R353A (R352 in DENV3)), did not impair the enzymatic activities with a slightly increased RdRp activities (Table 2). Indeed regulation of the RdRp de novo initiation activity by the MTase domain was elegantly demonstrated using an array of enzyme assays by Selisko and colleagues [33]. However, this enhancement in enzymatic activity was not beneficial for the virus: we observed a slightly delayed RNA replication as indicated by the reporter activity in the first 24 hrs in the replicon assay (Fig. 5A) and more severely impaired virus replication and infectivity in the infectious clone (Fig. 5B, C and D). For K95A, the intracellular and extracellular RNA levels was about 1 log lower with <50% virus recovery measured by plaque assay compared to WT. In the case of R353A, which had a two-fold higher elongation activity, there was no detectable virus replication. It is possible that the enhanced polymerase activities of mutants K95A or R353A may promote the production of incomplete genomes, truncated viral RNA fragments or catastrophic mutations that abolish reinfection. These mutations may also alter the NS5 conformational dynamics and affect its ability to form competent virus replication complex with other cofactors from the virus and the host. Taken together we conclude that the inter-domain interface of DENV NS5 is essential for successful virus replication and infectivity.
Previous studies on WNV NS5 identified genetic interactions between residues K46A/R47A/E49A (MTase) and L512 (RdRp) from reverse genetic experiments, and a 3D model for WNV NS5 was built based on the hypothesis of direct molecular interactions [12]. However, mutagenesis studies targeting these residues do not affect the interaction between the two domains significantly using in vivo co-expression system [60]. The crystal structure of JEV NS5 revealed a hydrophobic inter-domain interface between the MTase and RdRp domains [31]. The follow-up mutagenesis study addressed the importance for JEV virus replication of residues P113 and L115 from the MTase domain and F351 (F348 in DENV3), F467 (F464 in DENV3) and P585 (P582 in DENV3) from the RdRp domain [64]. In the same study, corresponding mutations (P113D, P115D, F349D, F465D and to a lesser extent P583D) in DENV2 NS5 also yielded severely defective viruses [64]. The residues examined by mutagenesis involved drastic changes of the side-chain chemistry that may impact the folding and overall architecture of the mutants and limit the interpretations of these phenotypes. Subsequently in a follow-up work, the in vitro polymerase activity assays on JEV NS5 showed that proteins with mutations at the interface had only mildly impaired polymerase activities [65]. The authors thus suggested that the defect in virus replication was due to the amplifying effects of the defective enzymatic activity, the impact on the protein-protein interactions, or the combined effect of both [31,64,65].
Careful examination of the two available structures of NS5 revealed two non-overlapping buried surfaces on the MTase domain but one shared buried surface from the RdRp fingers subdomain (Figs. 1 and 3; S4 Fig in S1 Text) [31]. Indeed, F348 and P582 from DENV3 NS5 also form hydrophobic stacking interactions with the MTase domain (Fig. 1D and E). Mutations of these two residues, particularly F348, would disrupt the interface in both structures and the DENV3 NS5 structure explains very well the defective phenotypes of the mutant DENV2 viruses (F349D (F348 in DENV3) and P583D (P582 in DENV3)) (Figs. 1, 3 and 6) [64]. This further argues for the dynamic nature of NS5 and the differential impacts on virus replication within the replication complex machinery.
Based on the present report and comparable studies on NS5 from JEV and other flaviviruses, it is possible that the NS5 protein from different flaviviruses may have evolved to adopt different sets of conformations, stabilized by distinct interfaces and leading to different allosteric mechanisms to regulate their enzymatic activities (Figs. 3 and 6; S5 Fig in S1 Text). This hypothesis is detailed in Fig. 6. However, which molecular conformation(s) participates in the various stage(s) of the virus replication cycle remains elusive (Fig. 6). Thus, during virus replication, NS5 may adopt various conformations upon recruitment of and interaction with NS3, viral RNA and other viral and host cofactors, for instance in order to synthesize a fixed ratio of positive and negative RNA strands, for unwinding dsRNA intermediates and for the effective production of progeny viral genomic RNA. The short linker of DENV NS5 is likely to dictate the degree of freedom of the overall molecular conformation so that the necessary discrete functional conformations can be attained. Inspection of the sequence heterogeneity of the linker residues (Fig. 4E) suggests that the NS5 proteins from other flaviviruses may also be endowed with comparable (or even higher) degrees of freedom (Fig. 6), possibly reflecting the need to accommodate additional host factors that must be recruited to the RC for replication. This hypothesis is also indirectly supported by the disordered linker sequence of JEV NS5 (S3 Fig in S1 Text). Remarkably, the flaviviral NS3 protease-helicase also contains an inter-domain linker that is flexible and variable in sequence and which is able to adopt multiple conformations at different stages of the virus life cycle [66–70]. In this respect, the inter-domain linkers observed in the NS3 and NS5 proteins might have evolved to play a role for the assembly of the RC comparable to the extended and flexible arms that are needed for viral capsid proteins assembly. Further studies exploring possible connections between virus pathogenesis and the dynamics of viral replicative enzymes in various flaviviruses are now needed, which are likely to be of interest to both basic virology and immunology research and also to antiviral therapeutics development.
|
10.1371/journal.pgen.1000681 | Plasticity of the Chemoreceptor Repertoire in Drosophila melanogaster | For most organisms, chemosensation is critical for survival and is mediated by large families of chemoreceptor proteins, whose expression must be tuned appropriately to changes in the chemical environment. We asked whether expression of chemoreceptor genes that are clustered in the genome would be regulated independently; whether expression of certain chemoreceptor genes would be especially sensitive to environmental changes; whether groups of chemoreceptor genes undergo coordinated rexpression; and how plastic the expression of chemoreceptor genes is with regard to sex, development, reproductive state, and social context. To answer these questions we used Drosophila melanogaster, because its chemosensory systems are well characterized and both the genotype and environment can be controlled precisely. Using customized cDNA microarrays, we showed that chemoreceptor genes that are clustered in the genome undergo independent transcriptional regulation at different developmental stages and between sexes. Expression of distinct subgroups of chemoreceptor genes is sensitive to reproductive state and social interactions. Furthermore, exposure of flies only to odor of the opposite sex results in altered transcript abundance of chemoreceptor genes. These genes are distinct from those that show transcriptional plasticity when flies are allowed physical contact with same or opposite sex members. We analyzed covariance in transcript abundance of chemosensory genes across all environmental conditions and found that they segregated into 20 relatively small, biologically relevant modules of highly correlated transcripts. This finely pixilated modular organization of the chemosensory subgenome enables fine tuning of the expression of the chemoreceptor repertoire in response to ecologically relevant environmental and physiological conditions.
| Rapid adaptation and phenotypic plasticity to the chemical environment are essential prerequisites for survival; and, consequently, large families of genes that mediate the recognition of olfactory and gustatory cues have evolved. We asked how flexible the expression of these genes is in the face of rapidly changing conditions encountered during an individual's lifetime. We used the fruit fly, Drosophila melanogaster, to address this question, since both the genetic composition and environmental rearing conditions can be controlled precisely in this experimentally amenable model organism. By measuring expression levels of all chemosensory genes simultaneously, we identified genes that show altered expression at different developmental stages, during aging, in males and females, following mating, and in different social conditions. We asked whether chemosensory genes are regulated independently or whether their regulation is structured. We found that chemosensory genes that are located in close proximity to one another on the chromosome are often regulated independently. However, statistical analysis showed that groups of chemosensory genes are coordinately expressed in response to a range of environmental conditions, revealing an underlying modular organization of the phenotypic plasticity of the chemosensory receptor repertoire.
| Responses to the chemical environment play an important role in animal survival, as chemical cues direct foraging behavior and food selection, predator avoidance, and, in insects, host plant recognition for oviposition and larval feeding. Chemical signals are also essential for the selection of mating partners, maternal behavior, and kin recognition. As a consequence of the profound importance of chemosensation for survival and reproduction, several large families of chemosensory genes have evolved through repeated processes of gene duplication and diversification [1]–[4], including genes that encode odorant receptors (Ors) [4]–[8], gustatory receptors (Grs) [4],[9], and, in insects, odorant binding proteins (Obps) [10]–[12]. In addition, large multigene families aimed at eliminating toxic chemicals have evolved, most prominently the cytochrome P450 superfamily [13]. Detoxification of plant defense chemicals together with development of chemosensors that enable fine tuning to host plants has been instrumental in the establishment of specialized insect-host plant relationships [14]. For example, the black swallowtail butterfly, Papilio polyxenes, has developed cytochrome P450s that can metabolize toxic furanocoumarins, which allows it to feed and oviposit on plants of the Umbelliferae family [15]. Similarly, Drosophila sechellia's host plant, Morinda citrifolia, is toxic to other Drosophila species. A 4 bp insertion in the upstream regulatory region of the D. sechellia Obp57e gene eliminates expression of this odorant binding protein, which elicits avoidance of the Morinda fruit in Drosophila species in which the gene is intact [16].
The rapid evolution of these large chemoreceptor gene families has generated functional redundancy between receptors and their ligands [17],[18], which confers sensitivity and robustness to the chemical recognition process. Animals, however, interact differently with their chemosensory environments under different developmental, physiological and social conditions. Therefore, it stands to reason that expression of the chemosensory repertoire would be dynamically regulated. This raises several fundamental questions: (1) Is the expression of chemoreceptor genes that are organized as clusters in the genome independently regulated or do genes within a cluster act as co-regulated functional ensembles? (2) Are all chemoreceptor genes equally sensitive to environmental fluctuations or is a core group of chemoreceptor genes particularly responsive to environmental or physiological changes? (3) Are certain chemoreceptor genes frequently co-regulated when environmental or physiological conditions change? (4) Is the expression of particular chemoreceptor genes upregulated or downregulated as a function of sex (males versus females), development (e.g. in larval stages, adult stages and aged flies), reproductive state (e.g. virgin or mated) or social context (e.g. solitary or group reared)?
To answer these questions we focused on the chemoreceptor families of Drosophila melanogaster, where both the olfactory and gustatory systems have been well characterized [4], [6]–[12],[19]. D. melanogaster provides an advantageous genetic model as inbred individuals can be readily generated and grown under controlled conditions, enabling control over both the genotype and the environment [20]. We constructed expression microarrays that enable us to survey simultaneously expression of all Obp, Or and Gr genes. We analyzed chemoreceptor expression as a function of sex, development, reproductive state, and social environment, and obtained a systematic description of the plasticity of the chemosensory window through which the fly experiences its chemical environment. We found that genes in clusters are independently regulated in the two sexes, during different developmental stages, and under different physiological and social conditions. Whereas many chemosensory genes showed plasticity in expression, a smaller number of exceptionally plastic genes was evident. Analysis of covariance of transcript levels across all environmental conditions showed that the chemosensory subgenome is structured as a mosaic of 20 small modules of highly correlated transcripts. This finely pixilated modular organization of the chemosensory transcriptome allows finely tuned phenotypic plasticity of expression of the chemoreceptor repertoire under different environmental conditions.
To assess to what extent transcription of chemosensory genes responds to changing conditions, we constructed cDNA expression arrays that represent 50 Odorant binding protein (Obp), 59 Odorant receptor (Or), and 59 Gustatory receptor (Gr) genes, four genes that encode other antenna-specific proteins, and four control genes. To prepare cDNA probes, primer sets were designed to generate unique 400–600 bp amplicons. All amplification products were sequenced and the sequences analyzed using the BLAST algorithm to ensure absence of cross-hybridizing sequences. Cross-hybridization is likely to occur in only two cases. Amplicons for Gr64d and Gr64e do not overlap, but these genes have partially overlapping transcripts and, therefore, could cross-hybridize. In addition, Or19a and Or19d are located 50 kb apart in opposite orientation and share the same sequences, rendering them indistinguishable. The extent of dye effects was assessed by hybridization of a mixture of equal amounts of Cy3 and Cy5 labeled RNA of the same sample. There was generally a close correlation between Cy3 and Cy5 hybridization intensities (Figure S1), indicating overall minor dye effects.
Among the 168 chemosensory genes represented on the microarray, we detected expression of 50 Obp genes, 54 Or genes, and 52 Gr genes, in at least one experimental condition. Expression levels of Obp genes were generally at least one order of magnitude higher than those of Or and Gr genes. Expression of chemoreceptor genes on our customized EST microarrays correlated well with previously obtained transcriptional profiles of chemosensory genes represented on high density oligonucleotide microarrays from Affymetrix, Inc. [21] ((Figure S2; r = 0.818, n = 174), but resolution for detection of chemoreceptor gene expression was substantially improved. We were not able to detect expression of Gr22b, Gr58c, Gr59c, Gr77a, Gr93b, Gr93c, Gr93d, Or10a, Or24a, Or85b, Or85c and Or85d, possibly due to highly localized expression of rare transcripts.
To assess modulation of chemoreceptor gene expression during development we compared expression of Obp, Or and Gr genes in third instar larvae (mixed sexes) and in virgin adult males and females. We also assessed changes in chemoreceptor gene expression in aged males and females. Pairwise comparisons between larvae and adults showed that relative expression of 28 chemoreceptor genes was biased in or specific to larvae at a Bonferroni corrected significance threshold of P<5.68E-5 (corrected for multiple testing at a nominal significance level of P<0.01) with a 2-fold change filter; conversely, 35 chemoreceptor genes showed adult-biased or adult-specific relative expression (Figure 1; Table S1). To validate our microarray observations, we amplified transcripts of the Obp58 and Obp99 gene clusters in larvae and adults. Obp99c was highly expressed in larvae and adults, whereas Obp99b showed strong adult-biased expression (Figure 2). Similarly, Obp58c and Obp58d were virtually undetectable in larvae, but expressed in adults with especially strong adult-specific expression of Obp58c. The results of the microarray analysis showed good concordance with results from RT-PCR experiments (Figure 2).
Since many chemoreceptors occur in clusters in the genome [4], we asked whether individual members of a cluster show coordinated or independent rexpression during development. We examined chemoreceptor gene clusters without intervening genes, including the Gr22a–e cluster, the Obp19a–d, Obp50a–e, Obp56a–f, and Obp57a–c clusters, and the Or43a–b cluster (Figure 3). There were extensive differences between larvae and adults in chemoreceptor gene expression. Gr22d, Gr22e, Obp50d, Obp56a–d, and Or43a showed larva-biased expression. Especially striking was the high larva-specific expression of Gr22d, as well as Gr22e. In contrast, expression of some chemoreceptor genes was observed only in adults, for example the Obp19a–d and Obp57a–c gene clusters and Obp56f.
When we compared relative expression of the same chemoreceptor genes in males and females, we observed extensive sexual dimorphism in transcript abundance levels. Male-biased expression was evident for Obp50c, Obp56d, and Obp56f, whereas female-biased expression was observed for Obp19a, Obp19c, Obp56a, Obp56e, Obp57a, and Or43b (Figure 3; Table S2). These results show that expression of chemoreceptor genes that are located within gene clusters can be regulated independently at different developmental stages and between the sexes.
Next, we asked whether chemosensory gene expression levels are stable throughout adult live or are subject to age-dependent plasticity. We compared transcript abundance levels in 10-day old and 6-week-old virgin males and females maintained under carefully controlled standard laboratory conditions, and found extensive age-dependent changes in transcript abundance in all classes of chemosensory genes (Figure 4). We found 104 chemosensory genes with altered transcript abundance in one or both sexes. Many genes with altered expression in aged flies were shared between males and females. However, sexual dimorphism in age-dependent chemoreceptor gene expression was pervasive. Interestingly, in males 15 Gr genes and 19 Or genes showed alterations in expression levels during aging (Figure 4B), while in females only three Gr genes and four Or genes changed expression levels during ageing (Figure 4A). The ubiquitous odorant receptor Or83b showed decreased expression levels with age in both sexes, whereas expression of Or1a and Gr98a was upregulated in both sexes during aging. Extensive differences among transcript abundance levels of Obp genes in young and old flies were especially prevalent for both sexes. Obp51a, Obp56e, Obp56g, Obp57a, Obp57c, and Obp99b showed altered expression levels during aging in both sexes, but in opposite directions (Figure 4). Again, expression of genes within a cluster appears to be regulated independently from other genes in the same cluster during ageing.
Next, we asked to what extent changes in physiological condition affect expression of the chemoreceptor repertoire. Mating results in physiological changes in females [22] and males [23]–[25]. We compared transcript abundance levels of chemosensory genes in virgin males and females reared separately to those of individuals that were allowed to mate (Figure 5). Following mating, only females showed a reduction in transcript levels of a suite of four Gr and 12 Or genes. In contrast, changes in Obp transcript abundance were seen in both sexes. Here, 16 out of 23 Obp genes with altered transcript abundance showed up-regulation in mated females (Figure 5A). Substantial changes in transcript abundance of Obp genes and Pino (a.k.a. smi21F), a putative odorant binding protein [26], were also evident in mated males (Figure 5B). Twelve Obp genes showed altered expression in both sexes, and among these five showed antagonistic changes in expression levels between the sexes (Figure 5). Thirteen out of 19 Obp genes with altered transcript abundance in mated males showed a reduction in transcript abundance, in contrast to the predominant up-regulation of Obp expression levels seen in mated females. Thus, mating caused profound changes in subsets of chemosensory genes in both sexes. The identities of the chemosensory genes affected or the effect on their transcript levels were distinct between males and females, indicating a profound sexually dimorphic change in the functional composition of the chemoreceptor repertoire after mating.
Our observation that the expression of the chemosensory repertoire is modified dramatically by social contact during reproduction led us to ask whether social context per se can elicit altered expression of distinct chemosensory genes. We compared transcript abundance levels in male and female flies that were reared as single isolated individuals to those of virgin flies reared in corresponding single sex groups. We observed changes in expression levels of few Gr or Or genes under these conditions (Figure 6). However, in females transcript abundance levels of seven Obp genes and Pino were down-regulated when individuals were reared in isolation, whereas two Obp genes were up-regulated (Figure 6A). In males transcription of five Obp genes was down-regulated when individuals were reared in isolation, whereas three Obp genes were upregulated (Figure 6B). Compared to our other experimental conditions, we found less overlap between genes with altered transcript abundance in males and females. Different members of the Obp56 gene cluster featured prominently among transcripts with altered levels in each sex. Only Obp56e, however, showed down-regulation in isolated individuals in both sexes and Obp57b was down-regulated in females and up-regulated in males when flies were reared in isolation (Figure 6).
Chemoreceptors have been implicated in the detection of both volatile [27] and non-volatile [28] social chemical signals. We wanted to assess whether exposure to social odor cues alone could result in altered transcript abundance of chemosensory genes. Therefore, we separated single flies from groups of same-sex or opposite sex flies with a double cheesecloth partition that would allow the transmission of olfactory cues, but would prevent physical interaction (it should be noted that Canton S w− flies used in these experiments are visually impaired). When single flies were maintained for five days under conditions in which they were exposed to same-sex group odors, there were virtually no changes in transcript patterns of chemosensory genes. Only expression of Obp57c was increased in females exposed to female group odor (Figure 7A), whereas expression of Obp84a and Obp83b was increased in males exposed to male group odor (Figure 7B). In contrast, we saw more extensive changes in transcript levels when we exposed single flies to opposite sex group odor for the same time period. Here, nine chemosensory genes in females showed altered transcript levels, including seven Obp genes and the antenna-specific a5 and a10 genes (Figure 7C). With the exception of Obp19c, all of these genes were down-regulated when a single female was exposed to male group odor. In single males exposed to female group odor, six Obp genes and a gustatory receptor gene (Gr2a) showed altered transcript levels (Figure 7D). Remarkably, there was no overlap between the subsets of chemosensory genes that had altered transcript levels when single males or females were exposed to opposite sex group odor. Notably, members of the Obp56 gene cluster (Figure 6) did not show altered expression under these conditions. The lack of concordance between transcript abundance of chemosensory genes when isolated individuals were compared to group reared individuals (Figure 6) and when isolated individuals were limited only to same sex group olfactory exposure (Figure 7A and 7B) shows that physical interactions are instrumental in determining expression of the chemoreceptor repertoire within same sex groups. However, when a solitary female is exposed to a group of males behind a cheese-cloth partition (Figure 7C) or when a solitary male is exposed to a group of females behind a cheese-cloth partition (Figure 7D), the patterns of changes in transcript abundance are distinct from those observed between isolated individuals and individuals maintained within same sex groups (Figure 6). This indicates that odor cues influence chemoreceptor gene expression between individuals of opposite sex (although a possible contribution of courtship song cannot be excluded).
We noticed that environmental plasticity of expression was heterogeneous among chemosensory genes, with certain members of the chemoreceptor ensemble responding more frequently to environmental changes than others. Therefore, we decided to investigate whether groups of chemosensory genes showed correlated transcript levels across all experimental conditions. We analyzed transcript levels using the modulated modularity clustering method. This unbiased, self-organizing paradigm is based on correlations of transcript abundance levels between different conditions, and sorts transcripts into modules such that transcript abundance levels among members within each module are more closely correlated than with members outside that module [29],[30]. The resulting pairwise correlation matrix can be represented graphically such that modules of correlated transcripts are organized in a matrix, with color-coding indicating the strength of each pairwise correlation [29],[30] (Figure 8). This analysis revealed 20 covariant ensembles (Figure 8; Table S3), indicating that transcriptional regulation of the chemoreceptor repertoire is indeed modular. At the same time, however, the large number of modules and their small sizes reflect the overall heterogeneity in transcriptional regulation of chemosensory genes. Whereas genes that are members of the same cluster were by and large independently regulated (e.g. Figure 2), in some instances genes in close proximity to each other within a cluster appeared to co-vary in expression levels. This was the case for Obp58b and Obp58c (located 376 bp apart in different orientations; Module 3), Obp56b and Obp56c (located 855 bp apart; Module 8), Obp83cd and Obp83ef (which have a 56 bp overlap with different orientations; Module 5), Obp 99b and Obp99d (located 1298 bp apart in different orientation with one intervening gene, Dup99B; Module 15), Or42a and Or42b (located 4231 bp apart with one intervening gene, Tsp42A; Module 19) and Or33a and Or33b (located 464 bp apart; Module 14). Strong negative correlations that reflect the antagonistic regulation of chemoreceptor gene expression described above were also observed, e.g. in Module 8. Obp76a (a.k.a. Lush), which binds the courtship pheromone cis-vaccenyl acetate [31] shows a strong positive correlation with Or67d, the transcript that encodes the receptor for cis-vaccenyl acetate [32],[33], and a strong negative correlation with Gr64a and Gr64c (Module 4). However, based on previously published spatial expression patterns of chemoreceptor genes [6],[10],[19],[34],[35] there appears to be no overall obvious correlation between spatial expression patterns and transcriptional covariance. With some exceptions, it appears that by and large Obp genes are segregated in modules that are distinct from modules that contain Or and Gr genes (e.g. Modules 7, 8, 15), and Or and Gr genes are frequently intermixed within covariant ensembles, e.g. Modules 14, 19 and 20 (Figure 8; Table S3). The transcript that encodes the ubiquitous odorant co-receptor Or83b [6] is found in module 17. The CO2 co-receptor Gr63a [36] is in Module 16, while its counterpart Gr21a [36],[37] forms part of Module 13, indicating that Gr63a and Gr21a expression is not closely correlated in the range of environmental conditions investigated in this study. Interestingly, Gr32a and Gr68a, which have both been implicated in pheromone recognition during the Drosophila courtship ritual [38],[39] occur together in Module 16 (Figure 8; Table S3).
Analysis of enrichment for shared transcription factor binding motifs is restricted due to the small size of the modules. Nevertheless, we analyzed in each module 5′ untranslated regions for enrichment of 62 putative transcription factor binding motifs. We found enrichment in module 15 of a transcription factor binding site for mirr shared by Obp83g and Obp99b (P = 0.03), in module 19 enrichment of a transcription factor binding site for pros shared by a5 and Or22b (P = 0.01), and in module 20 enrichment of a transcription factor binding site for Abd-B shared by Or49b, Gr64d, and Gr93a (P = 0.00035). However, even though some promoter regions that control cell-specific expression of odorant receptors have been identified [40],[41], transcription factors that control expression of Or, Gr and Obp genes remain largely unknown and may not be represented among the group of common transcription factors which we analyzed.
The olfactory and gustatory systems in Drosophila melanogaster have been well characterized [4], [6]–[12],[19], but the central problem of how ecologically relevant environmental conditions affect transcriptional variation in expression of the chemoreceptor repertoire has not been addressed previously in a systematic manner. As chemoreceptors are distributed over the entire body of the fly, including the third antennal segment, maxillary palps, proboscis, cibarial taste organs, tarsi, wing margins and the female abdominal reproductive plate, we chose to use a comprehensive analysis whole flies rather than heads. Consequently, some differences in expression between the sexes may be due to expression of chemoreceptors in non-chemosensory tissues. It is of interest to note that expression of odorant receptors in non-chemosensory organs has been observed using similar customized cDNA microarrays in both mice [42] and humans [43].
One should note that, in the absence of corresponding quantitative information about the chemosensory proteome, the relationship between transcript abundance levels and chemosensory function must be interpreted with caution. Although to date there is no evidence for posttranslational modifications of Obps, Ors and Grs might be subject to posttranslational regulatory mechanisms that may affect the amount of active gene product. Similarly, stability of mRNA has been postulated as a contributing factor to phenotypic variation in olfactory response to benzaldehyde associated with polymorphisms in the Obp99 gene cluster in a population of wild-derived inbred lines of Drosophila melanogaster [44].
Here, we have shown that transcriptional profiles of chemosensory genes in D. melanogaster are highly plastic during early development and ageing, as a result of mating, and in social contexts. Expression of chemoreceptor genes is highly sexually dimorphic and frequently sexually antagonistic, and the extent of transcriptional responses to changing conditions is heterogeneous among the chemoreceptor repertoire. Examination of the FlyAtlas expression data base indicates that Obp50c, Obp56d Obp99a and Gr32a are expressed in testes, Obp8a, Obp22a, Obp51a, Obp56e, Obp56f, Obp56g, Obp56i and Or59b in the accessory gland, Obp19c in the ovaries and Pino in both ovaries, testes and accessory glands, which suggests pleiotropic functions of these chemoreceptors and may account in part for the observed sexually dimorphic expression patterns [45]. In this study we have not included an analysis of expression of the recently discovered family of ionotropic odorant receptor (IR) genes, which are expressed in coeloconic sensilla of the antenna and respond, among others, to water and amines [46], and which were not represented on our microarrays. It will be of interest to investigate in future studies whether these genes show similar plasticity in expression as observed for the classical chemosensory genes.
A previous study used in situ hybridization to detect GFP expression of odorant receptors in larvae under the control of odorant receptor-specific promoters [47] through the GAL4-UAS binary expression system [48]. This study showed expression of 25 odorant receptors in the Drosophila larval olfactory system and reported that 14 of these receptors were larval-specific [47]. Although most of the larval expressed Or transcripts reported in this study were also identified on our arrays, the majority of these Or transcripts was also detectable in adults. There was some agreement with specificity of odorant receptor expression in larvae and adults (e.g. Or33a was found to show larval-biased expression and 10 Or genes were found to be expressed in adults as well as larvae both by us and others). However, the concordance between larval specificity detected by GAL4-UAS mediated expression of GFP in olfactory tissues and direct measurements of transcript abundance on our arrays from whole flies was generally poor. This can be due to expression of chemoreceptors in adult tissues not examined by previous in situ hybridization or reporter gene expression, differences in detection thresholds between the techniques used, differences in the strengths of GAL4-linked odorant receptor promoters in larvae and adults, or possibly differences in genetic backgrounds between strains used in the two studies.
A previous study reported sexually dimorphic expression of Obp99a and Obp99b [49]. Here we showed that sexual dimorphism in expression of chemosensory genes is widespread. This is especially evident among Obp genes, but the apparent prevalence of sexual dimorphism among these genes may be caused by their higher expression levels compared to those of Or and Gr genes. These broad sex-dependent differences in levels of expression of chemosensory genes suggest that males and females experience, interact with, and adapt to their chemical environments differently; for example, females have to evaluate the suitability of oviposition sites.
The independent regulation of genes within clusters, which we observed, is perhaps not surprising, as it may be a necessary requirement for subfunctionalization or neofunctionalization during evolution when daughter genes of duplication events either allow refinement and/or expansion in perception of the chemical environment or the acquisition of specialized chemosensory functions. Such functional diversification is reflected in the extensive sexual dimorphism where duplication of an ancestral gene may have resulted in daughter genes with different functions in males and females [49]. Similarly, gene duplication may enable adaptations of daughter genes to specialized chemosensory needs at different developmental stages (Figure 3).
Transcript profiles change drastically after mating, not only in females but also in males. The altered transcript abundance of Obp19d, Obp28a, Obp56a, Obp56g, and Obp99c that we observe in mated females (Figure 5) is consistent with a previous study which compared mating-induced changes of whole genome transcript profiles on high density oligonucleotide microarrays [50]. It is of interest that some odorant binding proteins, including Obp56e, Obp56f, Obp56g and Obp56i are highly expressed in the male accessory glands [50]. Thus, in addition to a function in olfaction, these odorant binding proteins may function also (or primarily) as carriers for physiologically active ligands that are transferred from the male into the female during copulation. Chemically-induced physiological and behavioral changes in females upon mating have been well characterized [51],[52]. Biological consequences of mating in males have also been documented [23]–[25].
Both volatile chemicals and cuticular hydrocarbons signal social information in Drosophila. The gustatory receptor Gr68a, which is expressed in chemosensory cells in the male tarsi, has been implicated in tactile chemosensation during courtship [38], together with Gr32a [39]. Recognition of the courtship pheromone, 11-cis-vaccenyl acetate, is mediated via the odorant binding protein Lush (Obp76a) and the Or67d receptor [31]–[33]. The expression of transcripts for Obp76a and Or67d is highly correlated across the range of environments studied here, as is expression of transcripts for Gr32a and Gr68a. A large ensemble of chemoreceptor genes, however, is sensitive to the social environment and modulated based on social context and, especially, opposite sex group odor (Figure 7). The identities of the odorants that are instrumental in mediating social interactions are not known, neither are the mechanisms that give rise to alterations in chemosensory gene expression levels.
POU-domain transcription factors, such as acj-6, have been implicated in mediating expression of odorant receptors in Drosophila olfactory neurons [40],[53]. A phylogenetic analysis of conserved regulatory elements among sequenced genomes of 12 Drosophila species has identified regulatory elements that act combinatorially to promote or repress the expression of specific odorant receptors in the olfactory sensilla of the maxillary palp [41]. A similar array of regulatory elements acted on by various transcription factors may also regulate Or gene expression in the antenna. Similar elements that regulate expression of Obp genes or Gr genes have not yet been identified. It is not clear whether transcriptional regulators and their binding sites that fine-tune transcription of Or genes in response to environmental changes are the same as those that control Or gene expression during development. Our results show that such fine tuning is exquisite in that genes that are located in close proximity within clusters can undergo independent transcriptional regulation (e.g. Figure 3).
Elegant electrophysiological studies have provided a detailed characterization of the molecular response profiles of a large number of odorant receptors in D. melanogaster [18],[54]. We found that four odorant receptors with documented odorant response profiles that all respond to alcohols and aliphatic esters [54] are contained in module 14 (Or35a, Or47a, Or85b and Or98a). Together with the observation that two of the four genes in Module 4 (Or67d and Obp76a [Lush]) encode proteins that are known to respond to cis-vaccenyl acetate, it is reasonable to extrapolate that the observed covariance in expression may have functional significance. However, the nature of naturally occurring ecologically relevant chemical signals that are discriminated by these receptors and the functional relationships between odorant binding proteins and odorant receptors and/or gustatory receptors remain largely unknown.
Our focused analysis of the chemoreceptor gene families using cDNA microarrays that provide enhanced resolution over previously used Affymetrix GeneChips revealed that the ensemble of chemosensory genes fractionates into 20 relatively small environmentally correlated modules (Figure 8). This observation shows that plastic transcriptional responses of chemoreceptor genes to a range of environments is modular, but at the same time indicates a great capacity of groups of chemosensory genes to alter their expression levels independently under a wide range of external environmental conditions.
Isogenic Drosophila melanogaster Canton S (B) w− flies were used for all experiments and grown under standard culture conditions (cornmeal-molasses-agar-medium, 25°C, 60–75% relative humidity, 12-hr light-dark cycle) for 4–5 days, unless otherwise specified. Larvae were collected at the 3rd instar stage. Sexes were reared separately after eclosion, except where indicated otherwise.
Chemoreceptor gene expression was compared between larval and adult samples, prepared by pooling an equal number of females and males. In addition, we compared young flies (10-day old) and old flies (6 week-old), transferred to fresh food every two days.
Chemoreceptor gene expression was compared between virgin females and virgin males, between virgin and mated females, and between virgin and mated males. To ensure that males had mated, we placed single males in vials with two females and collected males for microarray analysis when they were 5 days old, if females had oviposited.
Chemoreceptor gene expression was compared between flies reared in isolation and reared in a group of 25 same sex flies. To assess to what extent modulation of gene expression was dependent on social odor cues, we exposed single males or females to the odor from groups of flies of the same sex or opposite sex. Single flies were separated from groups of flies behind a screen of two layers of cheese cloth that prevented physical interactions (visual contact does not occur as our Canton S (B) strain carries a white mutation that renders them blind).
We amplified 400–600 bp fragments from genomic DNA or cDNA corresponding to exon sequences of 50 Obp genes, 59 Or genes, 59 Gr genes, four genes encoding antennal specific proteins (a5, a10, smi21F, Os9), plus two housekeeping genes as positive controls (Gapdh1 and actin-5C), and Gal4 and LacZ as negative controls (Table S4). The identities of all amplicons were verified by sequencing and arrays were printed on a Genetix QArray2 microarray printer at the Genomic Sciences Laboratory at North Carolina State University. Experiments comparing gene expression between larvae and adult flies used arrays with four technical replicates per slide; all other experiments used arrays containing eight technical replicates per slide.
For hybridization to the arrays, fly samples were collected and frozen between 1:00 and 3:00 pm. RNA samples were extracted from 25 flies per biological replicate, subjected to one round of amplification using the MessageAmp aRNA kit from Ambion Biosystems, Inc. (Foster City, CA) and 5 µg of each RNA sample was labeled with Cy3 or Cy5 fluorescent dyes (Amersham, Pharmacia, Piscataway, NJ; cat. # PA23001 and 25001). Labeled samples were purified using the QIAquick PCR Purification Kit (Qiagen, Inc., Valencia, CA). Six biological replicates of each sample were used for each experiment and included dye swaps to control for possible dye effects. Hybridization was performed for 60 h in a water bath at 42°C in the dark.
Arrays were scanned in a GenePix 4000B scanner, and raw data gathered by GenePix Pro software.
The raw data were subjected to log2 transformation and first normalized using a mixed analysis of variance (ANOVA) model accounting for dye, array, technical replicates (nested within array), and dye×array effects, where array, rep (array) and dye×array are random effects. Residuals were then extracted from the model and used for further ANOVA analyses to assess significant differences in gene expression among the samples. We used factorial, mixed model ANOVA according to the model: Residual = μ+dye+array+rep (array)+stage/sex/condition+ε, where μ represents the overall mean value and ε the error variance, to further partition variation of transcriptional expression between dye (fixed), array (random), technical replicates nested within array (rep (array) random) and stage (or sex, or treatment) terms by gene for each experiment. We also extracted residuals from raw data across all experiments after mixed model normalization to account for technical variation for cluster analysis. We used Modulated Modularity Clustering (MMC) [29] to organize the 172 genes into modules of correlated transcripts. MMC returned 20 modules as illustrated in Figure 8. Statistically significant differences were determined following normalization of the data by mixed model ANOVA. Bar graphs in the figures show fluorescent intensities of the raw data standardized for average array intensity and dye effect by adjusting fluorescent intensities based on the overall mean fluorescent intensities across arrays and between dyes. Comparisons of chemoreceptor gene expression between virgin females, mated females, virgin males and mated males employed a loop design. The data normalization procedure and analysis were identical except for an additional post-hoc pairwise comparison Student's t-test. A detection threshold was established based on two standard deviations from the mean lacZ signal intensity of the negative lacZ control. A Bonferroni corrected significance threshold of P<5.68E-05 was established as a criterion for statistical significance.
|
10.1371/journal.pntd.0006409 | Community knowledge, attitudes and practices on Yellow fever in South Omo area, Southern Ethiopia | Yellow fever (Yf) outbreak was recently reported in South Omo of Southern Ethiopia. This area was also highly affected by Yf outbreak in the 1960s. However, there is no reliable information on the level of community knowledge attitudes and practices about the disease in the area. The objective of the current study was to assess level of community knowledge, attitudes and practices about Yf.
Between March and May 2017, a community-based cross-sectional survey was conducted in two districts of the South Omo area. During the survey, 612 randomly selected adults were interviewed about Yf using structured questionnaire.
Out of the 612 study participants, 508 (83.0%) reported that they heard about Yf which is locally known as “a disease that causes vomiting blood”. Most (90.4%) of the study participants also said that Yf is different from malaria. Two hundred thirteen (41.9%) participants said that Yf can be transmitted from a patient to another person, while only 80 (37.6%) mentioned that the disease is transmitted through mosquitoes bite. Out of 333 (65.7%) study participants who believed that Yf is a preventable disease, 280 (84.1%) mentioned vaccine as a preventive method. The majority believed that the disease is a killer (97.2%) and a newly emerging (69.4%). Among the total of 612 study participants, 221(36.1%) were considered as having a high level of overall knowledge of Yf. Having educational level above 7th grade (AOR = 3.25, 95% CI: 1.39, 7.57, p = 0.006) and being resident of Bena-Tsemay district (AOR = 1.77, 95% CI: 1.12, 2.78, P = 0.014) were significantly associated with having a high level of overall knowledge of Yf. Agro-pastoralism as an occupation compared to farming was associated with having a low level of overall knowledge of Yf (AOR = 0.51, 95% CI, 0.33, 0.79, P = 0.003).
The findings indicate that most of the study community members had a low level of overall knowledge of Yf, especially about its cause, mode of transmission and preventive methods. Thus, there is a need to increase people’s knowledge and practices regarding the cause, mode of transmission and preventive methods like avoiding mosquitoe breeding sites beside vaccination through various strategies like disseminating information through community health extension workers and community leaders in the study area.
| Yellow fever is becoming one of the most important re-emerging mosquito-borne viral diseases in many African countries despite the availability of an effective vaccine. Hence, assessing information on what a community knows about Yellow fever would contribute to the design of appropriate control strategies in addition to increasing access for vaccine. In this study, we assessed knowledge, attitudes and practices of local community about Yellow fever in South Omo area, southern Ethiopia, where outbreaks have occurred repeatedly since the 1960s. We found that the study community members had low knowledge about the cause and mode of transmission of the disease though they knew that it is a killer and affects all age groups. More than half of the study participants believed that the disease can be transmitted from a patient to another person through breathing. In the present study area, providing information to community members through community health extension workers regarding the role of mosquitoes in the transmission of this disease, and teaching what to do to minimize mosquitoes bite in understandable way would be helpful to increase their awareness about Yellow fever.
| Yellow fever (Yf) remains a major public health problem in Africa since the 1930s, especially in the endemic areas of equatorial rain forest, the moist savanna and the dry savanna areas [1], despite the availability of effective vaccine for this disease. Previously, it was estimated that Yf causes 200,000 cases and 30,000 deaths annually, of which over 90% occurred in Africa [1]. According to a recent estimate, there were 130,000 Yf cases and 78,000 deaths in Africa for the year 2013 [2]. The disease is considered as one of the common re-emerging diseases in South America and many African countries like Democratic Republic of Congo, Sudan, Cameroon, Chad, Senegal, Côte d’Ivoire, Uganda, Sierra Leone, Ethiopia and Angola, where there is a low vaccine coverage or vaccine had waned [2–5]. In Ethiopia, large and small outbreaks of Yf occurred repeatedly since the 1960s and more recently, between November 2012 and October 2013 Yf outbreak occurred and resulted in many cases and deaths in South Omo area, southern Ethiopia, the same area which was highly affected by the outbreak in the 1960s [6–8].
The virus is transmitted to humans through the bite of widely distributed different species of the Aedes and Haemagogus mosquitoes. Thus, studies suggest that all areas in Africa where environmental conditions are suitable for mosquitoes breeding can be considered as areas at high risk of transmission; and the resurgence of Yf will continue unless vaccination is supported by effective mosquitoes control [9,10]. Moreover, it is suggested that control of mosquito-borne diseases like Yf requires effective participation of the local community [11]. Hence, assessing information on what a community knows about Yf would contribute to the efforts to design appropriate control strategies in addition to increasing access for vaccine. In this study, knowledge, attitudes and practices of local community about Yf was assessed in South Omo area, Southern Ethiopia, where Yf outbreak was recently reported.
The study protocol was approved by the Institutional Review Board (IRB) of the Aklilu Lemma Institute of Pathobiology (ALIPB), Addis Ababa University. The aim of the study was explained to each of the participant and verbal consent was obtained because most of the study participants were illiterate. Each participant was interviewed independently and the collected information was kept confidential.
The study was conducted in South Omo Zone, one of the 13 zones in Southern Nations, Nationalities and Peoples' Region (SNNPR). The Zone is located at about 750 km to the south of Addis Ababa and borders with Kenya on the south, a country which reported repeated outbreaks of arboviruses. The zone has eight districts with a total population of 573,435, and the majority of the inhabitants are practicing agro-pastoralism. Detailed information on the study Zone and the study population has been described elsewhere [12].
Among the six districts where Yf outbreak occurred between 2012 and 2013, two adjacent districts (Bena Tsemay and Debub Ari) were purposely selected for the present study based on the reported number of cases and deaths from the two areas [8]. Debub Ari district is located around Jinka town, approximately at 15 km to the North of Jinka. It has 50 small administrative units (kebeles) with a total population of 237,988. Among the 50 kebeles, four kebeles namely, Arkisha, Aykamer, Geza and Shepi were purposely selected for the present study based on the recent occurrence of Yf outbreak in the area. Bena-Tsemay district is found at 42 km to southeast of Jinka. The district has 32 kebeles with a total population of 74,853. Among the 32 kebeles, three kebeles namely, Luka, Goldia and Shaba-Argemenda were purposely selected for the present study based on the report of Yf outbreak occurrence. More detailed information on the study districts including a map showing the affected areas by the Yf outbreak has been described elsewhere [8].
Between March and May 2017, a community-based cross-sectional survey was conducted in the selected kebeles of the two districts. To our knowledge, there was no previous information on the level of community knowledge about Yf in the study area. Thus, assuming that 50% of adults will have high level of knowledge about Yf with 95% confidence in the estimate, 5% degree of accuracy, design effect of 1.5 and 90% response rate, a minimum sample size of 634 study participants would be required for the study.
Prior to data collection, a list of all the households in the selected kebeles was obtained from each of the respective kebele leaders. Based on the number of households in each of the selected kebeles, the pre-estimated sample size (634) was proportionally distributed. The required number of participants from each kebele was selected using systematic random sampling. The participants were eligible if they were residents of that kebele, age over 18 years, a husband/wife (or the responsible person) in the selected households, apparently healthy and volunteer to be interviewed.
Structured questionnaire was developed in English, based on information from available literatures regarding community knowledge, attitudes and practices about arboviruses from previous studies in other countries [13,14]. The questionnaire was translated into Amharic and pre-tested for clarity and acceptability in the study districts. During pre-testing, additional information was gathered and some of the questions were modified. The participants were interviewed about the public health importance, cause, mode of transmission, clinical symptoms, treatment, and preventive methods of Yf in their own local language (Ari, Bena and Tsemay) by trained health extension workers who were selected from each of the study kebeles. Each interview was made by a house–to-house visit. Information on the socio-demographic characteristics of the participants was also included in the questionnaire.
The collected data were double-entered into a data entry file using EpiData software, V.3.1. The data were exported to Stata version 11 for statistical analysis. Pearson chi-square test was used to evaluate the statistically significant difference in the level of knowledge of signs and symptoms as well as sources of information about Yf, mode of transmission, practices of people and attitudes towards Yf between male and females. Bivariate and multivariable logistic regression analyses were performed to explore associations of socio-demographic characteristics of the study participants with increased odds of having higher levels of overall knowledge of Yf and to quantify the degrees of association using odds ratio. P-values below 5% were considered as indicators of statistical significance.
The overall knowledge of the study participants about Yf was assessed using the following eight main questions: (1) able to mention jaundice and/or vomiting blood as the severe sign/symptoms of Yf, (2) able to identify that Yf is different from malaria, (3) able to know that the treatment for Yf is different from malaria treatment, (4) able to know that Yf is transmitted from a patient to another person through mosquitoes bite, (5) able to mention that a mosquito which transmits malaria is different from a mosquitoes that transmit Yf, (6) able to mention that Yf can be transmitted from monkeys to human through mosquitoes bite, (7) able to mention that Yf is preventable by vaccine, and (8) able to know that Yf has a vaccine. Response to these questions were added together to generate overall knowledge score ranging from 0 to 8. A score of one was given to correct response, zero being used for incorrect/do not know response. Then, the response was categorized into a high,{those who scored 5 (60% cut-off point) and above} and a low,{score 4(50%) and below } overall knowledge of Yf as previously described [15].
Community’s practices regarding preventive methods of Yf were assessed by asking questions such as Yf is a preventable disease, preventive methods for Yf and are you/your family vaccinated for Yf (Table 4). Gender difference in the proportion of answering correct response to each practice of question was evaluated using Pearson chi-square test. Similarly, study participants were asked questions such as “Yf is a public health problem”, “Yf is a newly occurred disease”, “Yf affects all age groups and Yf is a killer disease” (Table 5) to assess their attitudes. The association of gender and attitudes towards the disease was evaluated using Pearson chi-square test.
Table 1 shows the socio-demographic characteristics of the study participants. A total of 612 participants (55.9% males, age range from 18 to 87 years, mean age 33.36 years) participated in the study from the two areas, with a response rate of 96.5%. Among the study participants, 388 (63.4%) were recruited from the Debub Ari and the majority (65.2%) of the participants were illiterate.
Table 2 shows communty’s knowledge of signs/symptoms of Yf and their sources of information. Out of the 612 study participants, 508 (83.0%) reported that they heard about Yf which is locally known as “a disease that causes vomiting blood”. The study participants reported that they heard about the disease from individuals who were sick from Yf (43.3%), followed from health workers (34.1%), friends (29.9%) and they have heard/seen a person who died of Yf (23.8%). A larger proportion of male participants had information about Yf compared to female participants (86.5% vs 78.5%, X2 = 6.89, p = 0.01). The most commonly mentioned signs and symptoms of the disease were vomiting blood (84.0%), fever (45%) and headache (43.8%). Regarding the cause of the disease, 112 (22.3%) responded that they knew its cause. However, only 8 (7.8%) mentioned a virus as the cause of Yf, while others mentioned bacteria/germ (42.2%) or other factors (50.0%).
Significantly higher proportion of participants from Debub Ari mentioned vomiting blood as the main symptoms of Yf compared to participants from Bena Tsemay (92.1% vs 69.0%, X2 = 44.39, P<0.001). Similarly, a significantly higher proportion of participants from Debub Ari reported jaundice as the symptom of Yf compared to those from Bena Tsemay (43.1% vs 24.6%, X2 = 16.44, P< 0.001). On the other hand, a high proportion of participants from Bena Tsemay area mentioned bloody diarrhea as the main symptom of Yf compared to those from Debub Ari (56.7% vs 18.2%, X2 = 76.08, p< 0.001). Majority (90.4%) of the study participants said that Yf is different from malaria and its treatment also different from malaria treatment (72.5%). Relatively, a higher proportion of males (76.0%) than females (67.3%) reported that the treatment for Yf is different from the treatment for malaria (X2 = 5.25, P = 0.07).
Community’s knowledge regarding mode of transmission of Yf is summarized in Table 3. Less than half (41.9%) of the study participants said that Yf can be transmitted from a patient to another person. Among the 213 individuals who said that the disease can be transmitted, only 80 (37.6%) mentioned mosquitoes bite as a mode of transmission from person to person. More than half (55.9%) of the study participants thought that the disease can be transmitted from a patient to another person through breathing. A higher proportion of participants from Bena Tsemay mentioned mosquitoes bite as mode of transmission of the disease compared to those from Debub Ari (50.6% vs 28.2%, X2 = 11.02, p = 0.001). Few study participants also reported that they ever heard that this disease can be transmitted from monkeys to a person through mosquitoes bite or through drinking water contaminated with monkeys feces.
Table 4 shows community’s practices regarding prevention of Yf as reported by the study participants. More than half (65.7%) of the participants thought that Yf is a preventable disease. Among those who believed that Yf is a preventable disease, majority (84.1%) mentioned vaccine as a preventive method. Most of the participants from Bena Tsemay believed that Yf is preventable through vaccine compared to those from Debub Ari (90.7% vs 79.9%, X2 = 6.88, P = 0.01). The participants mentioned stagnant water as the main breeding site for mosquitoes, and suggested avoiding stagnant water and use insecticide sprays as preventive methods of mosquitoes breeding. However, during the survey the research team did not ask the study participants /has not observed whether they practice avoiding stagnant water and use insecticide sprays as preventive methods of mosquitoes breeding.
Table 5 shows attitudes of the study participants about the public health importance of Yf. Majority (86.2%) reported that Yf is a public health problem in their area, and 69.4% thought that Yf is a newly occurred disease in the area. A higher proportion (78.8%) of participants from Debub Ari considered Yf as a newly emerged disease compared to those from Bena Tsemay (52.8%) (X2 = 37.24, P< 0.001). Some of the participants suggested high rain fall (13.1%) and drought (4.5%) as the factors for the occurrence of the disease, while most (65.3%) of them had no idea about the factors contributing to its occurrence. The majority also believed that the disease affects all age groups (93.3%), it is a killer (97.2%) and not easily treatable (62.3%). A larger proportion (99.1%) of participants from Debub Ari said that Yf is a killer disease compared to those from Bena Tsemay (93.9%) (X2 = 13.04, P = 0.001). Almost all (99.4%) of the participants mentioned that if they suspect themselves or their families for Yf, they will visit health facility very soon.
Table 6 shows the overall knowledge of the study participants about Yf. Among the 612 study participants, 221(36.1%) were considered as having a high level of overall knowledge about Yf. Having Educational level of at least 7th grade was significantly associated with a having high level of overall knowledge of the disease (COR = 2.6, 95%CI, 1.27, 5.34, P = 0.009 and AOR = 3.25, 95%CI, 1.39, 7.57, p = 0.006). Similarly, being resident of Bena-Tsemay district was associated with having a high level of overall knowledge of Yf compared to residents of Debub Ari (AOR = 1.77, 95% CI, 1.12, 2.78, P = 0.014). Agro-pastoralism as an occupation compared to farming was associated with having a low level of overall knowledge of Yf (COR = 0.65, 95% CI, 0.45, 0 .94, p = 0.022, and AOR = 0.51, 95% CI, 0.33, 0.79, P = 0.003).
In the present study area, Yf outbreaks have been repeatedly occurring since 1960 to the recent and caused thousands of deaths and cases [6–8].
Thus, it was assumed that at least 50% of the rural residents of the study area had a correct information about the cause, symptoms, mode of transmission and preventive methods of the disease. However, the results of this study indicated that most of community members (63.9%) had a low level of overall knowledge of the disease especially regarding its mode of transmission though the majority believed that the disease affects all age groups and it is a killer disease. The finding is in line with a previous study by Mohapatra and Aslami et al. [16] which revealed a low level of knowledge about Dengue fever among patients of a rural tertiary-care hospital in Sasaram, Bihar India, despite a good attitude. Comparable to the findings of this study, a community based study in Nepal [17], and study in India [14] also showed a low level of knowledge about Dengue fever despite the rapid spreads of the disease. In the current study, educational level above 7th grade was found to be an indicator for a high level of overall knowledge of Yf which is similar to the findings of study elsewhere regarding knowledge of Dengue fever [13].
Among others, misconception about the symptoms, mode of transmission and preventive methods of a disease could highly affect individuals attitudes towards health-seeking behaviour and prevention. Infection with Yf virus shares non-specific clinical symptoms with other febrile illness, particularly with malaria [1, 18], which could affect timely care seeking behaviour of patients, and leading to delayed diagnosis/treatment in endemic areas due to limited diagnostic facilities. In the present study area, although the majority of the study participants could identify Yf as a disease which causes vomiting blood, there were community’s members who had no clear information regarding the difference between Yf and malaria (especially they confused it with falciparum malaria). Study in Tanzania also revealed that many community members believed that most instances of fever are due to malaria and the community had a low level of awareness about other non-malaria febrile illnesses like Rift Valley fever and Dengue fever despite the endemicity of the diseases [19]. In Ethiopia, Yf is known as Bicha Woba by health professionals. In Amharic language, Bicha means yellow, and Woba means malaria. On top of this, in Ethiopia there is no laboratory based effective diagnosis of Yf, and health professionals might not be fully aware of the magnitude of the disease.Thus the diagnosis may not be even considered, or may incorrectly be interpreted as malaria. The overall effects of these poor diagnosis practices in the health facilities and the name of the disease (Bicha Woba) could also be another factor which potentially affect community’s knowledge regarding the difference between Yf and malaria in the present study area. This necessitates creating community awareness on the common clinical features of Yf and malaria in the present study area where both diseases exist.
Studies suggest that understanding of local community’s knowledge regarding the mode of transmission of mosquitoes borne emerging and re-emerging viral diseases and their attitudes/practices towards prevention are among the essential elements for designing effective control measures [11, 20, 21]. The results of the present study indicate that majority of the community’s members had no correct information on the mode of transmission of Yf, as only 80 individuals mentioned that Yf can be transmitted from a patient to another person through mosquitoes bite, while only 24 individuals thought that mosquitoes which transmit Yf are different from those mosquitoes that transmit malaria. Contrary to our finding, a recent study in Angola showed that about 44% of the study participants had correct information regarding the transmission of Yf through mosquito bites [22]. Community’s knowledge regarding the transmission of the virus from non-human primates to human was also very poor. The results of study by Mohapatra and Aslami et al. [16] also showed that few patients had knowledge about Aedes mosquitoes as vectors of Dengue fever in Sasaram, Bihar India. The finding is also comparable to the findings of study on Rift Valley fever knowledge among agropastoral communities in Tanzania, where majority of the study participants had heard of the disease, but very few knew that mosquitoes can transmit the disease [15]. Providing information to community members through community health extension workers regarding the role of mosquitoes in the transmission of Yf and other diseases like malaria, Zika virus, Dengue fever etc, and teaching what to do to protect themselves/their families from mosquito bites in a simple understandable way would be helpful in increasing their awareness about major mosquito-borne diseases.
The results of this study also showed that the members of the community in the present study area regarded Yf as one of the most deadly diseases and most of the study participants acknowledged vaccination as the main preventive method. However, some of the study participants complained that though they and their families received a vaccine during the recent Yf outbreak in the area, they did not get adequate information about the specific vaccine/for what disease they are vaccinated, duration of protection of the vaccine and whether it is a vaccine or a treatment. This could affect the positive attitudes of individuals toward vaccine, and the health professionals who deliver vaccine could provide information on the specific vaccine and for what disease during vaccination campaigns, which in turn increases community’s attitudes and the desired vaccination coverage. Marlow et al., [22] also recommended the importance of providing clear information on Yf vaccination to the target population as some individuals did not understand whether the vaccine provided prevention or treatment in Angola where Yf outbreak was recently reported.
This study would provide information on the level of community’s knowledge, practices and attitudes about Yf in the studied area where such data was not available. However, the survey was conducted in two districts which were conveniently selected among the 6 districts of the South Omo Zone where cases and deaths were reported between 2012 and 2013. This is one of the limitations of the study which could affect the generalizability of the findings to all the different communities of the South Omo Zone. In addition, this quantitative study was not supplemented by qualitative study like focus group discussion which is important to gather detailed and additional information regarding community’s knowledge, practices and attitudes of Yf.
Although Yf is becoming one of the most re-emerging mosquito-borne viral diseases in many African countries including the present study area, the findings of the present study showed that people living in endemic areas do not have adequate knowledge about its cause and mode of transmission though they consider it as one of the killer diseases. Although most of the study participants acknowledged vaccination as the main preventive method of Yf,some of the study participants complained that they did not get adequate information about the specific vaccine/for what disease they are vaccinated which affects the positive attitudes of individuals toward Yf vaccine. Thus, there is a need to increase people’s knowledge and practices regarding the cause, mode of transmission and preventive methods like avoiding mosquitoe breeding sites beside vaccination through various strategies like disseminating information through community health extension workers and community leaders.
|
10.1371/journal.pgen.1006328 | Genetic loci associated with coronary artery disease harbor evidence of selection and antagonistic pleiotropy | Traditional genome-wide scans for positive selection have mainly uncovered selective sweeps associated with monogenic traits. While selection on quantitative traits is much more common, very few signals have been detected because of their polygenic nature. We searched for positive selection signals underlying coronary artery disease (CAD) in worldwide populations, using novel approaches to quantify relationships between polygenic selection signals and CAD genetic risk. We identified new candidate adaptive loci that appear to have been directly modified by disease pressures given their significant associations with CAD genetic risk. These candidates were all uniquely and consistently associated with many different male and female reproductive traits suggesting selection may have also targeted these because of their direct effects on fitness. We found that CAD loci are significantly enriched for lifetime reproductive success relative to the rest of the human genome, with evidence that the relationship between CAD and lifetime reproductive success is antagonistic. This supports the presence of antagonistic-pleiotropic tradeoffs on CAD loci and provides a novel explanation for the maintenance and high prevalence of CAD in modern humans. Lastly, we found that positive selection more often targeted CAD gene regulatory variants using HapMap3 lymphoblastoid cell lines, which further highlights the unique biological significance of candidate adaptive loci underlying CAD. Our study provides a novel approach for detecting selection on polygenic traits and evidence that modern human genomes have evolved in response to CAD-induced selection pressures and other early-life traits sharing pleiotropic links with CAD.
| How genetic variation contributes to disease is complex, especially for those such as coronary artery disease (CAD) that develop over the lifetime of individuals. One of the fundamental questions about CAD––whose progression begins in young adults with arterial plaque accumulation leading to life-threatening outcomes later in life––is why natural selection has not removed or reduced this costly disease. It is the leading cause of death worldwide and has been present in human populations for thousands of years, implying considerable pressures that natural selection should have operated on. Our study provides new evidence that genes underlying CAD have recently been modified by natural selection and that these same genes uniquely and extensively contribute to human reproduction, which suggests that natural selection may have maintained genetic variation contributing to CAD because of its beneficial effects on fitness. This study provides novel evidence that CAD has been maintained in modern humans as a by-product of the fitness advantages those genes provide early in human lifecycles.
| It is well established that modern human traits are a product of past evolutionary forces that have shaped heritable variation, but we are far from a good understanding of whether recent natural selection has acted on these and how this process has left its imprint across the genome. While many genome-wide multi-population scans have searched for signatures of positive selection [1–9], these studies have detected relatively few adaptive candidates for common traits or diseases [10–12]. This suggests that classic ‘selective sweeps’ have been relatively rare in recent human history [13–16] and that current tools may not be appropriate for detecting and validating smaller shifts in adaptive variation, thus limiting our understanding of how natural selection acts on common diseases and traits [12]. Research in this area is also important as the combination of positive selection and significant GWAS signals at the same locus supports the existence of functional variation for disease. Here, we aimed to comprehensively identify selection signals for coronary artery disease (CAD) loci with methods designed to detect recent signals of positive selection. We compared selection signals in 12 worldwide populations (HapMap3) with CAD genetic risk (CARDIoGRAMplusC4D) to help understand how selection acts on disease variation at the genetic level and prioritize genes most likely modified in relation to CAD. We examined the association between selection signals and gene expression to further test whether adaptive candidates are functionally important for CAD in terms of gene regulation. Lastly, we tested if CAD genes are associated with reproductive fitness to try to understand why this common disease persists in modern humans.
Classic population genetics theory describes positive selection with the selective-sweep (or hard-sweep) model, in which a strongly advantageous mutation increases rapidly in frequency (often to fixation) resulting in reduced heterozygosity of nearby neutral polymorphisms due to genetic hitch-hiking [17, 18] and a longer haplotype with higher frequency. Many methods have been developed to detect these signatures [19, 20], including traditional tests that detect differentiation in allele frequencies among populations (i.e. Wright’s fixation index, Fst [21]) and more recently developed within population tests for extended haplotype homozygosity (i.e. integrated haplotype score, iHS [9]). Some of the most convincing examples of human adaptive evolution have been uncovered for traits influenced by single loci with large effects. For example, the lactase persistence (LCT) and Duffy-null (DARC) mutations affecting expression of key proteins in milk digestion [10] and malarial resistance [22] both display hallmarks of selective sweeps. Other loci that are not clearly monogenic but also show selective sweeps are associated with high-altitude tolerance (EPAS1 [23]) and skin pigmentation (SLC24A5 and KITLG [24]). These studies show that rapid selective sweeps mainly occurred for new mutations with large effects on phenotypes.
Motivated by these initial successes and the increasing availability of global population data genotyped on higher resolution arrays (i.e. HapMap Project, 1000 Genomes Project), many recent genome-wide scans for candidate adaptive loci have recently been performed [11]. These suggest that selection may have operated on a variety of biological processes [10] in ways that differ among populations (i.e. local adaptation) [25], been prevalent in genetic variation linked to metabolic processes [26], and may often target intergenic regions and gene regulatory variants rather than protein-coding regions [12]. Often only the larger signals underlying monogenic (or near-monogenic) traits are typically considered for follow-up because of losses in the statistical power needed to quantify significance for smaller candidate adaptive signals after correcting for genome-wide multiple testing [20]. The adaptive status of many smaller candidate signals also remains uncertain due to inconsistencies in results between studies that utilized the same data [14], and it is inherently more difficult to functionally validate candidate adaptive signals underlying complex polygenic traits compared to monogenic traits where only one or a few variants may have been under selection due to their influence on fitness [27, 28].
In contrast to population genetics, research in quantitative genetics has shown that rapid adaptation can often occur on complex traits that are highly polygenic [29, 30]. Under the ‘infinitesimal (polygenic) model’, such traits are likely to respond quickly to changing selective pressures through smaller allele frequency shifts in many polymorphisms already present in the population [13, 31]. Selection on such variation is generally less likely to push it towards fixation due to genetic correlations, thus producing smaller changes in surrounding heterozygosity over time that are harder to detect with most current population genetic methods [14, 28, 32]. Note that polygenic and classic sweep models are not mutually exclusive [13, 33], for alleles with small- and large-effects may both underlie a polygenic trait, which suggests that there will be some variation in the degree to which candidate alleles are modified after selective events. Because most common diseases are highly polygenic, we need to improve how we detect and classify the adaptive signatures underlying these traits.
Recent studies investigating genomic selection on polygenic traits have taken two approaches. The first scans for significant selection signals for a subset of large effect SNPs that have previously been identified as genome-wide significant. For example, Ding and Kullo [34] found significant population differentiation (Fst) for 8 of 158 index SNPs underlying 36 cardiovascular disease phenotypes, and Raj et al. [35] observed elevated positive selection scores (Fst, iHS) for 37 of 416 index susceptibility SNPs underlying 10 inflammatory-diseases. The second approach tests if aggregated shifts in genome-wide significant allele frequencies are associated with phenotypic differences by population, latitudinal, or environmental gradients, which might indicate local adaptation. For example, Castro and Feldman [36] used 1300 index SNPs underlying many polygenic traits and found elevated adaptive signals (Fst and iHS) above background variation, and Turchin et al. [37] demonstrated moderately higher frequency of 139 height-increasing alleles in a Northern (taller) compared to Southern (shorter) European populations. These approaches all assume that genome-wide significant variants are the most probable selection targets, but many if not most such variants are tags for the causal variants, which may be at lower frequencies. This suggests an approach more sensitive for detecting subtle signals of polygenic selection is needed.
We chose CAD as a model for examining polygenic selection signals for complex disease because it has (and continues to) impose considerable disease burden (and possible selection pressure) in humans [38], its underlying genetic architecture has been extensively studied [39, 40] and many of its risk factors (cholesterol, blood pressure) have been under recent natural selection [41] related to potential pleiotropic effects or tradeoffs with CAD. Antagonistic pleiotropy describes gene effects on multiple linked traits where selection on one may cause negative fitness effects (i.e. reproduction, survival, and disease) in the other due to their antagonistic genetic association [42]. Two common misconceptions are that CAD only occurs in older people and is a disease that has mainly afflicted modern humans. If either were true, selection might not have had either the opportunity or sufficient time to affect genetic variation associated with CAD. However, CAD begins to manifest during reproductive ages [43, 44] and disease origins can be detected even in adolescence through degree of atherosclerosis [44, 45] and myocardial infarction events [46]. CAD is also a product of many heritable risk factors (cholesterol, weight, blood pressure) whose variation is expressed during the reproductive period, when CAD could drive selection directly or indirectly. Furthermore, CAD has impacted human populations since at least the ancient Middle Kingdom period, with atherosclerosis detectable in Egyptian mummies [47]. This suggests that there has been enough time for evolutionary responses to CAD to have occurred, genomic signatures from which may be detectable in modern humans.
By combining several 1000 Genomes-imputed datasets including HapMap3 and Finnish SNP data, a large genetic meta-analysis of CAD, HapMap3 gene expression data and lifetime fitness data from the Framingham Heart Study, we sought to address the reason(s) why CAD exists in humans by answering the following questions: 1) Has selection recently operated on CAD loci? 2) How do selection signals underlying CAD loci vary among populations and are they enriched for gene regulatory effects? 3) Do candidate adaptive signatures overlap directly with CAD genetic risk and is this useful for highlighting disease-linked selection signals? 4) Do CAD-linked selection signals display functional effects and evidence of antagonistic pleiotropy, in that they are also linked to biological processes or traits influencing reproduction?
To test for selection signals for variants directly linked with CAD, we utilized SNP summary statistics from 56 genome-wide significant CAD loci in Nikpay et al. [40], the most recent and largest CAD case-control GWAS meta-analysis to date, to identify 76 candidate genes for CAD (see Methods). Nikpay used 60,801 CAD cases and 123,504 controls from a mix of individuals of mainly European (77%), south (13% India and Pakistan) and east (6% China and Korea) Asian, Hispanic and African American (~4%) descent with genetic variation imputed to a high-density using the 1000 Genomes reference panel. By investigating all SNPs in CAD genes, we aimed to improve detection of smaller polygenic selection signals for the range of functional genic variants and short-range intergenic regulatory variants that would be missed with approaches that only consider genome-wide significant SNPs.
We utilised the integrated Haplotype Score (iHS) to estimate positive selection for each SNP underlying CAD genes within each population separately. Because iHS is typically used to detect candidate adaptive SNPs where the selected alleles may not have reached fixation [9], this estimate is well suited for detecting recent signals of selection as opposed to other measures [20]. iHS is also better suited for detecting selection acting on standing variation in polygenic traits [20, 48].
Candidate selection signals were found for many of the 76 CAD genes within each of the 12 worldwide populations (11 HapMap3 populations and Finns; Fig 1A for top 40 based on their association with CAD log odds genetic risk, S1 Fig for all 76). These were defined as ‘peaks’ of significantly elevated iHS scores across SNPs within each gene-population combination, with the apex approximating the likely positional target of positive selection.
The results for the largest iHS score per gene and population (Fig 1A) show that most candidate selection signals were relatively small, but a few larger signals were detected. For example, out of the 912 gene-by-population combinations (S1 Fig), 354 (38%) contained weak-moderate candidate selection signals (significant iHS between 2–3), 84 (9%) contained moderate-strong signals (significant iHS between 3–4), and 6 (0.6%) had very strong signals (significant iHS > 4). The 6 largest candidate signals were found in the following gene-population combinations: BCAS3 in GIH (iHS = 4.45), MEX (iHS = 4.23) and CEU (iHS = 4.86), PEMT in MKK (iHS = 4.24), ANKS1A in LWK (iHS = 4.03), and CXCL12 in JPT (iHS = 4.10), with all iHS p values <0.0001. Six genes (BCAS3, SMG6, PDGFD, KSR2, SMAD3, HDAC9) exhibited candidate selection signals consistently within all populations (Fig 1A), and many genes also contained consistent selection signals for all populations within similar ancestral groups (e.g. African, European etc, Fig 1A).
Within CAD genes, multiple candidate selection signals were sometimes present (particularly within larger genes, within separate linkage disequilibrium (LD)-blocks); these varied between and sometimes within a population. For example, eleven (of the twelve) populations had candidate selection signals in PHACTR1 introns 4, 7 or 11 (Table 1; see also S2 Fig, comparing cross-population selection signals in PHACTR1) that were in separate LD-blocks (see S2 Fig, LD plots). For eight populations, there was a broad and relatively weak set of candidate selection signals in intron 4 (the largest PHACTR1 intron, ~300kb in length). Intron 4 is also the location of the published CAD index SNP (rs12526453) for PHACTR1. Other interesting candidate selection signals present in other CAD genes (S1 Fig) are not discussed here. Such patterns suggest that candidate selection signals are sometimes complex and often do not correspond to the SNPs with largest effect on disease.
For each CAD gene within each population, we used a mixed effects linear model to regress SNP-based estimates of CAD log odds genetic risk (ln(OR), obtained from cardiogramplusc4d.org) against iHS selection scores (see Methods). We accounted for LD structure by including the first eigenvector from an LD matrix of correlations (r2) between SNPs within each gene as a random effect.
For a subset of CAD loci, we found significant quantitative associations between disease risk and selection signals and for each of these the direction of this association was often consistent between populations (Fig 1B). Furthermore, when compared to a null distribution of genes selected randomly from the genome, the strength of the CAD log odds versus selection signal at most loci was statistically significant (Fig 1C). Fig 1B shows 40 genes ranked based on those with the most consistent number of significant associations across the 12 populations, with those that showed fewer than four significant associations excluded. Positive and negative associations indicate elevated selection signals present in regions with higher or lower CAD log odds genetic risk, respectively.
In the comparison across populations, directionality of significant selection-risk associations tended to be most consistent for populations within the same ancestral group (Fig 1B). For example, in PHACTR1, negative associations were present within all European populations (CEU, TSI, FIN), and in NT5C2 strong positive associations were present in all East Asian populations (CHB, CHD, JPT). Other negative associations that were consistent across all populations within an ancestry group included five genes in Europeans (COG5, ABO, ANKS1A, KSR2, FLT1) and four genes (LDLR, PEMT, KIAA1462, PDGFD) in East Asians.
Additional consistent positive associations included four genes (CNNM2, TEX41, NT5C2, MIA3) in East Asians, three (BCAS3, RAI1, KCNK5) in Europeans, and one (PPAP2B) in Africans. In comparison to other ancestral groups, African populations showed fewer significant selection-risk associations (27.9% of all 76-gene x 12-population combinations) than Asians (31.5%) or Europeans (32.8%). Some associations were consistent in all but one population (e.g. CNNM2, ABCG8 in Europeans; BCAS3, KCNK5 in Asians; CNNM2, TEX41 in Africans) or unique to one population within an ancestral group (e.g. TEX41 in FIN, COG5 in ASW).
Below we focus on BCAS3 (Fig 2) and PHACTR1 (Fig 3), two of the strongest selection-risk associations which, when adjusting for LD (see Methods), displayed varying directionality between at least two populations.
The genetic risks of CAD for variants in BCAS3 were positively correlated with an extremely large candidate adaptive signal in all European and two of three East Asian populations (Fig 1B). For example in CEU, the largest iHS score was 4.85 and highly significant, and was elevated across most of BCAS3 (Fig 2B CEU, spanning introns 1–18 and various LD-blocks, Fig 2C), which matched the approximate trends in CAD log odds giving rise to a highly significant positive correlation (Fig 2A CEU). In contrast, in YRI there was no detectable selection signal close to the index SNP (Fig 2B YRI), but weak-moderate signals were present towards the end of BCAS3 (Fig 2B YRI, introns 18–19, smaller LD-blocks Fig 2C), which also corresponded with lower CAD log odds (Fig 2B, YRI) thus giving rise to a significant negative correlation in Fig 2A.
For all European populations, PHACTR1 (see CEU example, Fig 3A) selection peaks were typically located within regions of consistently lower CAD log odds (Fig 3B). This contrasted with most other non-European populations where the highest candidate selection peaks were located within regions with elevated CAD log odds (including the index CAD SNP rs12526453, intron 4). The largest selection peak in GIH (Fig 3B) overlapped the CAD log odds peak in PHACTR1 giving rise to the strong positive association seen in Fig 3A. The two distinctive selection peaks in both CEU and GIH were separated by different LD-blocks (Fig 3C), suggesting that these may have developed independently within PHACTR1. Interestingly, the negative association found for the MKK population was due to the location of the selection peaks more closely matching those of the European populations in intron 11 (S2 Fig).
To establish whether variants with evidence of selection in CAD genes also showed evidence of function, we performed an eQTL scan in 8 HapMap3 populations with matched LCL gene expression. We compared all SNPs in each CAD locus against expression for each focal gene within each population.
We found that SNPs with significant integrated Haplotype Scores (iHS) were often also involved in gene regulation, compared to SNPs with non-significant selection scores (Fig 4, Kolmogorov-Smirnov test p value <0.001). To assess which biological pathways were enriched for the highest-ranked genes according to Fig 1B, i.e. those where selection scores were most closely associated with CAD log odds genetic risk, we included the top 10 genes into the Enrichr analysis tool [49] and found that these genes are especially enriched in pathways related to metabolism, focal adhesion and transport of glucose and other sugars. More interestingly, we found connections to reproductive phenotypes in the associations of these genes with pathways, ontologies, cell types and transcription factors. For example, we found links to ovarian steroidogenesis and genes expressed in specific cell types and tissues including the ovary, endometrium and uterus (see S1 Table for Enrichr outputs).
To test whether CAD genes are directly associated with human lifetime reproductive success (LRS or total number of children born across reproductive lifetimes), a prerequisite for responses to selection, we examined their association with LRS for women in the Framingham Heart Study (FHS). Out of the 76 CAD genes (representing 20,254 SNPs in total; a minimum, average and maximum of 18, 266 and 2121 SNPs tested per gene, respectively), 51 genes contained SNPs that were significantly nominally associated with LRS (p<0.05), 30 genes contained SNPs associated at p<0.01 and 12 genes contained SNPs associated at p<0.001, based on both nominal p values from FaST-LMM and permuted p values (see S2 Table). For example, the most significant associations per gene included rs56152906 in PPAP2B (p = 5.23E-06, permuted p<0.0001), rs7896502 in LIPA (p = 0.0002, permuted p = 0.0001) and rs2479409 in PCSK9 (p = 0.0003, permuted p = 0.0001) including a further 9 (COL4A2, FLT1, HDAC9, KSR2, LPA, MIA3, PDGFD, PLG, SMAD3) genes with significant LRS associations at permuted p<0.001. The two previous studies that have investigated genome-wide SNP associations with LRS found associations with similar levels of evidence to our study. For example, the leading SNP in Kosova et al. [50] for completed family size was rs10966811 with p = 5.57E-06. The top two leading SNPs in Aschebrook-Kilfoy et al. [51] for LRS were rs10009124 (p = 7.65E-08) and rs1105228 (p = 2.16E-06).
When we considered these associations using fastBAT that combines SNP associations within a gene (accounting for LD-redundancy) into single gene-level p value, similar results were obtained with 8 genes significantly associated with LRS (e.g. PPAP2B, p = 0.0004, permuted p = 0.001, SMAD3, p = 0.0061, permuted p = 0.007, MIA3, p = 0.008, see S2 Table).
To test the null hypothesis that CAD variation is no more significantly associated with LRS than is variation in the rest of the genome, we used a permutation approach. We sampled 20,254 non-CAD related SNPs (matched within MAF bins to the CAD SNPs) randomly (without replacement) across the genome 100 times. The permuted p value was based on the number of times each random sample of 20,254 non-CAD SNPs shared significantly more associations with LRS than did the 20,254 CAD SNPs. The total sample of randomly selected SNPs (n = 2,025,400) was also compared against the 20,254 CAD SNPs with a Kolmogorov-Smirnov (K-S) test. We found that CAD genetic variation was significantly (p = 9.49E-08 and p = 1.90E-07 based on one- and two-sided K-S tests, respectively; permuted p<0.01) more enriched for LRS compared to the rest of the genome (see S2 Table for other fitness-related traits), providing strong evidence in the FHS for shared fitness effects at CAD loci. This was also the case when we tested this at the gene-level using fastBAT results (permuted p = 0.026, S2 Table).
To test whether effects between CAD loci and LRS were antagonistic, we cross-referenced the genome-wide significant index SNPs for CAD from Nikpay [40] with significant SNPs for LRS from the FaST-LMM analysis. Of the 56 CAD index SNPs in Nikpay [40], 53 were genotyped or imputed in the FHS to a high confidence. In FHS, six of those SNPs (11.3%) were significantly associated with LRS (FaST-LMM p < 0.05), with 5 out of those 6 antagonistic, i.e. the allele that increases LRS also increases risk for CAD (see S3 Table). For example, in FLT1, rs9319428-A significantly increases both LRS (ß = 0.041, p = 0.0143) and CAD risk (ß = 0.039, p = 7.13E-05), and similarly, rs2048327-C in LPA significantly increases both LRS (ß = 0.041, p = 0.00894) and CAD risk (ß = 0.057, p = 2.46E-09). This suggests that antagonistic effects occur in some loci, but the power to detect and define this for smaller effect variants on LRS is limited in the FHS (e.g. see S3 Fig for power estimates). Compared to the CARDIoGRAMplusC4D study [40] where the 56 genome-wide significant CAD index SNPs were obtained using a meta-sample of ~184,000 individuals, SNP effects on LRS were based on 1,579 women from the FHS. Given that power to detect small effects (i.e. |ß|<~0.3–0.4 or OR <~1.2–1.3) in these studies is poor when n is small (i.e. ~1000 individuals [52]) suggests that larger samples of women and men with completed reproduction are needed to test for antagonistic effects comprehensively to avoid false negatives.
We further tested whether SNPs are associated with both LRS and CAD due to potential confounding effects rather than antagonistic pleiotropy, i.e. confounding effects would occur if CAD SNPs influence LRS, which in turn cause significant changes in CAD risk due to physiological, hormonal or social changes related to childbearing/rearing. We tested the association between CAD SNPs and CAD in FHS females, stratified by LRS (see S3 Fig for full analysis). We found no significant effect of LRS modifying SNP effects on CAD (see S3 Fig), which supports the antagonistic pleiotropy hypothesis, however we caution that larger, better powered studies may show some level of attenuation.
Extending this investigation to understand why CAD genes are significantly enriched for LRS, i.e. what possible underlying reproductive processes are contributing, we performed an extensive systematic literature search on the 40 top-ranked genes in Fig 1 and a random set of 20 non-CAD genes. While gene set enrichment had been performed (above) suggesting some connections to reproductive phenotypes, such tools cannot capture the full range of possible effects on multiple fitness traits, some that are themselves rarely tested in other mammalian (non-human) species due to ethical limitations. We found evidence for direct links between CAD genes and fitness (S4 and S5 Tables) including genes associated with reproductive (PPAP2B, [53]) or twinning (SMAD3, [54]) capacity and number of offspring produced (e.g. KIAA1462, [55], SLC22A5, [56]). PHACTR1, LPL, SMAD3, ABO and SLC22A5 may contribute to reproductive timing (menarche, menopause) in women [57–59] and animals [60]. Expression of PHACTR1 [61], KCNK5 [62], MRAS and ADAMST7 [63] appear to regulate lactation capacity. Some gene deficiencies also cause pregnancy loss (e.g. LDLR, [64], COL4A2, [65]). Evidence for other pleiotropic links related to fitness included 25 genes that shared links with traits expressed during pregnancy (S4 and S5 Tables), i.e. variation that can negatively influence the health and survival outcomes of both the fetus and mother [66]. For example, a variant of CDKN2B-AS1 significantly contributes to risk of fetal growth restriction [67], both FLT1 [68] and LPL [69] are significantly differentially expressed in placental tissues from pregnancies with intrauterine growth restriction (IUGR), and preeclampsia and LDLR-deficient mice had litters with significant IUGR [70]. A further 29 and 19 genes were linked to traits that can directly influence female and male fertility, respectively (13 influence both) (S4 and S5 Tables). For example, BCAS3 and PHACTR1 are highly expressed during human embryogenesis [71, 72], SWAP70 is intensely expressed at the site of implantation [73], and PHACTR1 may play a role in receptivity to implantation [74]. For ABCG8 and KSR2, animal models provide further support as gene expression deficiency can cause infertility in females (ABCG8, [75]) and males (KSR2, [76]).
Pleiotropic connections were also apparent in the classification of specific disorders or from studies investigating single-gene effects. For example, women with polycystic ovarian syndrome (PCOS) have higher rates of infertility due to ovulation failure and modified cardiovascular disease risk factors (i.e. diabetes, obesity, hypertension [77]). While reduced fecundity associated with PCOS might suggest it would not fit the model of antagonistic pleiotropy, some hypothesize that it is an ancient disorder and may have provided a rearing advantage in ancestral food-limited environments [78]. A number of CAD genes in this study (e.g. PHACTR1, LPL, PDGFD, IL6R, CNNM2) are found differentially expressed in PCOS women [79–83], suggesting possible links between perturbed embryogenesis and angiogenesis. In males, this can be demonstrated with a mutation in SLC22A5 that causes both cardiomyopathy and male infertility due to altered ability to break down lipids [84, 85]. More generally, many recent studies link altered cholesterol homeostasis with fertility, which is most apparent in patients suffering from hyperlipidemia or metabolic syndrome [86, 87].
For the random set of non-CAD genes that were approximately the same size as the top 20 genes in Fig 1, we were only able to find three (out of 20) with at least one potential link with fitness (S6 Table) using the same systematic literature search further demonstrating the relative abundance of CAD loci effects on fitness earlier in life.
This study identified many candidate adaptive signals suggesting that selection on CAD loci is much more widespread than previously appreciated (also see S1 Discussion). It has previously been suggested [12] and demonstrated [88] that selection on gene expression levels has been an important element of human adaptation in general. We confirm this result for CAD associated loci. Positive selection signals within CAD loci were more likely than random SNPs to be associated with gene expression levels in cis (Fig 4).
We found evidence that some of these signals may be a result of selection pressures induced directly by CAD itself. This finding is important for highlighting genes that may have been modified directly by selection on disease phenotypes and also for our general understanding of how quickly human genomes can respond to selection induced by changing environments. Subsequent fitness and biological process analyses and a thorough literature review demonstrated that CAD loci are enriched for lifetime reproductive success in women and also linked to other male and female reproductive phenotypes, which suggests both their potential to respond to natural selection and their possible role via antagonistic pleiotropy in the reproductive tradeoffs that would help to explain why CAD is common in modern humans. While the connection between cardiovascular disease and lifetime parity is not novel (e.g. see [89, 90]), it is not known whether this connection is due to hormonal, physiological, social or selective processes. The current study provides the first evidence for a selective and antagonistic mechanism.
One of our most interesting findings was the significant association between selection signals and CAD log odds genetic risk. This approach of integrating genome scans of positive selection with genome-wide genotype-phenotype data has been promoted previously as a tool to uncover biologically meaningful selection signals of recent human adaptation [12, 88] but has rarely been applied. Among the exceptions, Jarvis et al. [91] found a cluster of selection and association signals coinciding on chromosome 3 that included genes DOCK3 and CISH, which are known to affect height in Europeans.
For highly-ranked genes (according to the number of significant associations present within the 12 populations) in Fig 1B such as BCAS3, CNNM2, TEX41, SMG6 and PHACTR1, the consistent overlap between selection and genetic risk of CAD suggests that many of these may have been modified by CAD-linked selective pressures. If so, then two conditions must have been met. Firstly, CAD was present for long enough to be involved in these genetic alterations, an evolutionary process which generally takes thousands of years. Indeed, precursors of CAD (i.e. atherosclerosis) are detectable in very early civilizations [47]. Secondly, the effects of CAD were directly or indirectly expressed during the reproductive period and trait variation was under natural selection due to its effects on reproductive success.
It is only possible for natural selection to directly act on CAD if those outcomes modify individual fitness relative to others in the same population. As outlined in the introduction, this is possible as CAD outcomes (i.e. myocardial infarction) do occur in young adults. However, early-life CAD outcomes are relatively rare, suggesting selection is more likely to operate indirectly on CAD via its risk factors (or other pleiotropically linked traits, discussed below), which provides a more likely explanation for the close associations we found between positive selection and genetic risk. Supporting this, phenotypic selection has been found operating on CAD risk factors [41], suggesting that these selection pressures are still present in modern humans.
Some genes had large signals of selection but showed weak or no consistent overlap with CAD genetic risk. For example HDAC9 (Histone Deacetylase 9) shows extensive evidence for having undergone recent selection within most populations, especially those of European or Mexican decent, but little or no overlap with CAD risk was evident in most populations. This suggests positive selection has operated on this gene due to its effects on a trait unrelated to CAD, which may not be surprising given HDAC9’s broad biological roles (as a transcriptional regulator, cell-cycle progression) and association with other very different phenotypes including ulcerative colitis [92] and psychiatric disorders [93]. This further demonstrates that this approach is useful for separating candidate selection signals important for the disease or phenotype of interest from those that aren’t.
We found direct evidence in the Framingham Heart Study for shared fitness effects at CAD loci, which were specifically significantly enriched for their effects on female lifetime reproductive success relative to the rest of the genome. This novel finding shows a connection between direct fitness and later disease expressed through CAD loci. An extensive literature review supported this conclusion. All 40 CAD genes from Fig 1 shared at least one (often more) connection with fitness (S4 and S5 Tables). Some appear to directly influence fitness (offspring number, age at menarche, menopause, survival), while many were associated with early-life reproductive traits that are likely to correlate with fitness, including variation in ability to fertilize/conceive or fetal growth, development and survival. This suggests further pleiotropic links between CAD and early-life fitness-related traits. Directly testing for antagonistic effects between fitness and CAD, we found evidence at specific loci for the leading CAD index SNPs, where the allele that significantly increased LRS also significantly increased CAD. We further found no evidence that this link between CAD and LRS was due to confounding (e.g. physiological, hormonal) effects of LRS on CAD risk. While this is promising, the Framingham study is limited in its power to detect small fitness and CAD effects; better powered studies may yet be needed to definitively establish antagonistic effects at all loci. Fitness traits collected on genotyped populations are currently rare, but this is likely to change as more biobank-scale studies come online.
To facilitate interpretation of selection occurring on early-life traits or CAD phenotypic risk factors that share pleiotropic connections and possible evolutionary tradeoffs with coronary artery disease, we present a conceptual figure (Fig 5). These pleiotropic effects are important because many of them affect traits expressed early in life, some extremely early in life. Any allele that increases reproductive performance enough early in life to more than compensate for a loss of associated fitness late in life will be selected [42]. Such a mechanism has been recently suggested to help explain the maintenance of polymorphic disease alleles in modern human populations [94]. While such tradeoffs have been previously tested for in humans using genotypes, LRS and lifespan (e.g. [95]), there is not yet much evidence that such a mechanism influences human disease. A 2017 study by Rodríguez et al. [96] that used an indirect measure of fitness provides support for antagonistic pleiotropy acting on general early health and later life disease. Our study demonstrates that CAD genes are significantly and directly enriched for fitness with evidence that some of the leading CAD effect SNPs share an antagonistic relationship with fitness through significant positive effects on LRS. This provides support for such a mechanism influencing CAD and may help to explain our vulnerability to this disease.
There are also some limitations to our approach. We utilized CAD genetic risk estimated from a meta-analysis based on predominantly European (77%) with smaller contributions from south/east Asian (19%), Hispanic and African American (~4%) ancestry [40]. Genetic risk variation for CAD might be different in the un-represented (i.e. Mexican) or less-represented (i.e. African) populations in this meta-analysis. If that were the case, it would reduce the usefulness of comparing selection and risk estimates in those populations. We also saw fewer significant selection-risk associations in the African populations (Fig 1B), however this may be due to selection signals in the African populations being less obvious than those in East Asian and European populations, perhaps due to lesser linkage disequilibrium, as is consistent with results from previous studies [99]. Calculating disease risk and selection variation from populations within the same ancestral group might help resolve this, however it only represents a potential shortcoming for our cross-population analyses and not observations of antagonistic pleiotropy.
In this study, we found evidence that natural selection has recently operated on CAD associated variation. By comparing positive selection variation with genetic risk variation at known loci underlying CAD, we were able to identify and prioritize genes that have been the most likely targets of selection related to this disease across diverse human populations. That selection signals and the direction of selection-risk relationships varied among some populations suggests that CAD-driven selection has operated differently in these populations and thus that these populations might respond differently to similar heart disease prevention strategies. The pleiotropic effects that genes associated with CAD have on traits associated with reproduction that are expressed early in life strongly suggests some of the evolutionary reasons for the existence of human vulnerability to CAD.
We started with the 56 lead index SNPs from Supplementary Table 5 in Nikpay et al. [40] corresponding to 56 CAD loci. When the index SNP was genic, all SNPs within that gene were extracted (using NCBI’s dbSNP) including directly adjacent intergenic SNPs ±5000bp from untranslated regions (UTR) in LD r2>0.7 (with any respective genic SNP). When the index SNP was intergenic, that SNP and other directly adjacent SNPs ±5000bp and in LD>0.7 (with the index SNP) were extracted and combined with SNPs from the respective linked gene listed in Nikpay including SNPs ±5000bp from UTR regions in LD r2>0.7 with that gene. This resulted in SNP lists for 56 genes. To further explore other genes not directly connected with lead index SNPs, but that were within the CAD loci identified by the two most recent CARDIoGRAMplusC4D studies–including Deloukas et al. [39] (i.e. 46 loci and 61 genes listed in their Tables 1–2) and Nikpay et al. [40] (i.e. 10 loci and 15 genes listed in their Table 1)—we extracted SNPs within each of those genes (plus SNPs ±5000bp from UTR regions in LD r2>0.7 with that gene). This resulted in SNP lists for a further 20 genes, bringing the total number of candidate genes for CAD to 76.
The per-SNP log odds (ln(OR)) values for the 76 genes were obtained for the additive model from Nikpay et al. [40] available at http://www.cardiogramplusc4d.org/downloads and used in the analysis described below.
Genotype data (1,457,897 SNPs, 1,478 individuals) were downloaded for 11 HapMap Phase 3 (release 3) populations (http://www.hapmap.org [100]) including: Yoruba from Ibadan, Nigeria (YRI), Maasai in Kinyawa, Kenya (MKK), Luhya in Webuye, Kenya (LWK), African ancestry in Southwest USA (ASW), Utah residents with ancestry from northern and western Europe from the CEPH collection (CEU), Tuscans in Italy (TSI), Japanese from Tokyo (JPT), Han Chinese from Beijing (CHB), Chinese in Metropolitan Denver, Colorado (CHD), Gujarati Indians in Houston, TX, USA (GIH), and Mexican ancestry in Los Angeles, CA, USA (MEX). We also included another HapMap3 population, the Finnish in Finland (FIN) sample (ftp://ftp.fimm.fi/pub/FIN_HAPMAP3 [101]). These data had already been pre-filtered, i.e. SNPs were excluded that were monomorphic, call rate < 95%, MAF<0.01, Hardy-Weinberg equilibrium p<1x10-6.
Before phasing and imputation, we performed a divergent ancestry check with flashpca [102] to check accuracy of population assignments, converted SNP data from build 36 to 37 with UCSC LiftOver (https://genome.ucsc.edu/cgi-bin/hgLiftOver), checked strand alignment in Plink v1.9 [103] to ensure all genotypes were reported on the forward strand, and kept only autosomal SNPs. To speed up imputation, data were first pre-phased with Shapeit v2 [104] using the duoHMM option that combines pedigree information to improve phasing and default values for window size (2Mb), per-SNP conditioning sates (100), effective population size (n = 15000) and genetic maps from the 1000 Genomes Phase 3 b37 reference panel (ftp.1000genomes.ebi.ac.uk/vol1/ftp/release/20130502/).
Phased data were imputed in 5 Mb chunks across each chromosome with Impute v2 [105]. We then removed any multiallelic SNPs (insertions, deletions etc) from the imputed data and excluded SNPs with call rate < 95%, HWE p<1x10-6 and MAF<1%. The final dataset was then phased with Shapeit v2, and alleles were converted to ancestral and derived states using python script. Ancestral allele states came from 1000 Genomes Project FASTA files and derived 6-primate (human, gorilla, orangutan, chimp, macaque, marmoset) Enredo-Pecan-Ortheus alignment [106] from the Ensembl Compara 59 database [107].
Integrated Haplotype Score (iHS): Using the package rehh [108] in R version 3.1.3, per SNP iHS scores were calculated within each population (after excluding non-founders) using methods described previously [9]. iHS could not be calculated for SNPs without an ancestral state, or whose population minor allele frequency is <5%, or for some SNPs that are close to chromosome ends or large regions without SNPs [9]. Rehh was also used to standardize (mean 0, variance 1) iHS values empirically to the distribution of available genome-wide SNPs with similar derived allele frequencies. For analyses in the main text, we considered a SNP to have a candidate selection signal if it had an absolute iHS score > 2, a permuted p value <0.05, and was within a ‘cluster’ of SNPs that also had elevated iHS scores. Although permuting p values is computationally more intensive, it provides more flexibility to detect smaller selection signals that may be incorrectly classified with the more stringent Bonferroni correction that is often applied to these estimates. For the analyses described below, even though we only used iHS estimates for the SNPs defined in the CAD genes (and additional SNPs for permutation purposes), we calculated per-SNP iHS scores genome-wide (rather than locally, i.e. within 1MB regions around focal SNPs), for this provides more accurate estimates because final adjustments are made relative to other genome-wide SNPs of similar sized derived allele frequency classes. P values for iHS scores were permuted based on comparison of nominal p values against 10000 randomly selected estimates from within the same derived allele frequency classes.
We first tested the null hypothesis that there is no association between CAD genetic risk and signals of positive selection for CAD genes. For each gene within each population, we used a mixed effects linear model to regress SNP-based estimates of CAD log odds (ln(OR)) genetic risk against selection scores (iHS) resulting in 912 separate regressions. To account for LD structure (and potential confounding of highly correlated SNPs) within each gene, we also included the first eigenvector derived from an LD matrix of correlations (r2) between SNPs within each gene as a random effect. We chose to model LD structure with mixed-effects models rather than LD-prune because for many genes, the SNP samples would have been too small for regression analyses. Also, it would be very difficult to properly capture both selection and the CAD log odds peaks needed to compare these variables. We did however investigate alternative models to validate our approach (i.e. running the same models without the LD structure variable; using smaller multiple LD-pruned subsets of SNPs per gene) with consistent results suggesting our approach was largely robust to LD effects and likelihood of false positive associations. We accounted for multiple testing by permuting p values for each regression based on comparing each nominal p value against 10000 permuted p values derived from shuffling iHS scores.
Genes were then ranked based on the number of significant associations summed across the 12 populations. The 40 genes with at least four or more significant associations are shown in Fig 1B. To illustrate the positional architecture of these selection-risk associations, plots for selected highly-ranked genes are shown in Figs 1 and 2. By demonstrating how CAD genetic risk peaks and valleys correspond to variation in the magnitude of selection scores (iHS), this allowed visual assessment of potential modifications made to the phenotype-genotype map by selective pressures imposed directly or indirectly by CAD. It also helped us localize selection peaks within genes and compare them between populations. Similar peaks suggested similar selection and different peaks suggested local adaptation. This way of presenting the results also allowed us to detect the smaller adaptive shifts in allele frequencies typically expected to underlie selection on polygenic traits.
We then tested a second null hypothesis: that the selection-risk associations using the CAD genes are not unique compared to non-CAD associated loci. For each of the 76 CAD genes, we randomly (without replacement) chose 100 genes of similar length across the genome and performed the same mixed effects regression procedure described above for each gene by population combination using both CAD log odds values from Nikpay et al. [40], iHS scores estimated from the SNP data, and the first LD eigenvector from SNPs within a gene. Permuted p values were derived by comparing the nominal p value for each CAD gene against the 100 null distribution p values from the non-CAD associated genes. Results are shown in Fig 1C.
To examine whether candidate adaptive signals within each gene corresponded to a gene’s regulatory variation, we regressed SNPs within focal genes and gender against that gene’s probe expression levels, which had previously been quantified in lymphoblastoid cell lines from circulating peripheral blood using Illumina’s Human-6 v2 Expression BeadChip for eight of the 12 populations [109]. Given gene expression in peripheral blood is known to be an important marker for cardiovascular disease, we therefore might expect this cell type a good candidate to search for association between selection signals and regulatory variants important for these genes. The raw gene microarray expression data had previously been normalized on a log2 scale using quantile normalization for replicates of a single individual then median normalization for each population [109]. P values for each SNP-probe association were permuted using 10000 permutations by randomly shuffling gene probes expression. P values were then extracted for the most significant iHS score for each gene-population combination and compared to the same number of p values randomly drawn from different LD blocks underlying SNPs with non-significant iHS scores across each gene-population combination. A Kolmogorov-Smirnov test was used to compare the distribution of p values from each. To examine what biological processes were associated with the top ranked genes from Fig 1, we uploaded the top 10 genes into Enrichr (http://amp.pharm.mssm.edu/Enrichr/) to define associated pathways (i.e. KEGG 2016, kegg.jp/kegg), ontologies (MGI Mammalian phenotypes, informatics.jax.org), cell types (Cancer cell line Encyclopedia, broadinstitute.org/ccle) and transcription factors (ChEA 2015, amp.pharm.mssm.edu/lib/chea.jsp).
We tested whether CAD SNPs were directly associated with human fitness. For a trait to evolve, this is one of the main prerequisites, but it also helps demonstrate whether alleles that influence disease also influence reproduction, which in the case of CAD suggests there may be antagonistic trade-offs between early versus late life.
We used the Framingham Heart Study dataset because it has completed reproductive outcomes (lifetime reproductive success (LRS) or number of children ever born), genotypes, pedigree data, cardiovascular outcomes and demographic and socioeconomic data. LRS was derived from clinical questionnaires and further validated with pedigree data. We did not include other datasets for validation here, as it is extremely hard to find others that include all these variables. There were 1,579 women from the Original and Offspring cohorts who had genotypes and all phenotypes available also after excluding non-founders. FHS 500k Affymetrix genotypes were 1000-Genomes imputed using the same pipeline described above bringing the total number of SNPs available (at MAF>1%) to 7,486,901. LRS was adjusted to deal with secular demographic change where data was broken into six groups based on year women were born and divided by the mean reproductive success of women in that group (same as described in [41]).
We examined the association of all SNPs available in the FHS for the 76 CAD genes (20,254 SNPs) with LRS using linear mixed models implemented in FaST-LMM [110] that account for potential confounding effects of genetic similarity by including a k-spectral decomposition variable derived from the realized relationship matrix (RRM) of an LD-pruned subset of SNPs. SNPs used for RRM were not in LD with CAD SNPs to avoid proximal contamination. Factors that may affect LRS—education, smoking status, whether the person was born in the US, and estrogen usage (hormone therapy or contraceptive use)—were included as covariates. Permutations with 10,000 iterations were also run for each SNP in order to validate nominal p values obtained directly from FaST-LMM, where permuted p values were based on the number of times nominal p values for 10,000 randomly chosen SNPs (within a similar MAF bin) were greater than the target SNP nominal p value. Bonferroni and FDR adjustment was also applied to p values based on 20,254 tests.
To test the null hypothesis that CAD SNPs are collectively no more enriched for fitness compared to non-CAD SNPs, we randomly sampled without replacement 20,254 non-CAD SNPs (matched within MAF bins to the CAD-SNP sample) 100 times. The permuted p value was based on the number of times (out of 100) that the number of significant p values in the random sample exceeded that for the CAD SNP sample. We also compared the distribution of p values between all randomly chosen SNPs (2,025,400) from this analysis to the 20,254 CAD SNPs with a Kolmogorov-Smirnov test. One- and two-sided tests were run. The one-sided test specifically tested whether the distribution of p values for CAD SNPs was stochastically larger compared to non-CAD SNPs.
We then used fastBAT [111] to test whether CAD is enriched for fitness at the gene-level. fastBAT combines SNP-based summary-level data from GWAS and LD reference data to give locus-based estimates of association. We also ran 100 permutations for each of the 76 genes in order to estimate a permuted p value for each locus. Permuted p values were based on the number of times p values for the 100 randomly chosen similarly-sized genes were greater than that for each CAD gene. We also tested whether CAD genes were collectively more enriched for fitness at the gene-level, relative to non-CAD genes. We randomly chose 76 non-CAD genes of similar size with 300 permutations (sampling without replacement) and asked how many times the number of p values for the non-CAD genes exceeded that of the CAD genes.
Other traits that may influence fitness were also tested (using the same analysis/permutation tests as above) for FHS women including age at first and last birth, interbirth interval, menarche and menopause. Menarche and menopause were derived from questionnaires, while birth timing/spacing were estimated from pedigrees. We did not consider reproductive outcomes for men as that data was only available from pedigrees, which is less reliable than clinical records. Age at first and last birth and interbirth interval were also adjusted for secular demographic changes (same as above).
Alternative simplified models for LRS, AFB and ALB were run in FaST-LMM where fitness measures were not adjusted for temporal effects and no covariates were included. This boosted sample sizes (S2 Table) due to avoiding missing values associated with covariates, however results were largely comparable (S2 Table) suggesting no power gain from unadjusted models.
We only tested for antagonistic effects for LRS as that is the most direct measure of fitness and was the only fitness trait where CAD SNPs were significantly and consistently enriched across (un)adjusted models (S2 Table). To assess whether antagonistic effects were present between LRS and CAD, genome-wide significant CAD index SNPs were taken from Nikpay et al. [40] and cross-referenced with significant LRS SNPs from the FaST-LMM regression results. In an extended analysis (results not shown), we also included any SNP in high-LD (r2> = 0.8) and proximal (±1MB) to the index SNP to boost the number of SNPs available for comparison: results were virtually identical for the significance of LRS and CAD effects and the consistency of antagonistic effects. An antagonistic effect was defined as an allele that significantly increased LRS and significantly increased CAD risk. We also tested whether CAD SNPs were associated with both LRS and CAD due to other confounding (rather than pleiotropic) effects (see S3 Fig for methods and findings).
|
10.1371/journal.pntd.0003989 | A Randomized Controlled Trial of a Brief Intervention for Delayed Psychological Effects in Snakebite Victims | Snakebite results in delayed psychological morbidity and negative psycho-social impact. However, psychological support is rarely provided to victims.
To assess the effectiveness of a brief intervention which can be provided by non-specialist doctors aimed at reducing psychological morbidity following snakebite envenoming.
In a single blind, randomized controlled trial, snakebite victims with systemic envenoming [n = 225, 168 males, mean age 42.1 (SD 12.4) years] were randomized into three arms. One arm received no intervention (n = 68, Group A), the second received psychological first aid and psychoeducation (dispelling prevalent cultural beliefs related to snakebite which promote development of a sick role) at discharge from hospital (n = 65, Group B), while the third received psychological first aid and psychoeducation at discharge and a second intervention one month later based on cognitive behavioural principles (n = 69, Group C). All patients were assessed six months after hospital discharge for the presence of psychological symptoms and level of functioning using standardized tools.
At six months, there was a decreasing trend in the proportion of patients who were positive for psychiatric symptoms of depression and anxiety from Group A through Group B to Group C (Chi square test for trend = 7.901, p = 0.005). This was mainly due to a decreasing trend for symptoms of anxiety (chi-square for trend = 11.256, p = 0.001). There was also decreasing trend in the overall prevalence of disability from Group A through Group B to Group C (chi square for trend = 7.551, p = 0.006), predominantly in relation to disability in family life (p = 0.006) and social life (p = 0.005). However, there was no difference in the proportion of patients diagnosed with depression between the three groups (chi square for trend = 0.391, p = 0.532), and the intervention also had no effect on post-traumatic stress disorder.
A brief psychological intervention, which included psychological first aid and psychoeducation plus cognitive behavioural therapy that can be provided by non-specialist doctors appeared to reduce psychiatric symptoms and disability after snakebite envenoming, but not depression or post-traumatic stress disorder.
Sri Lanka Clinical Trials Registry: SLCTR/2011/003
| Snakebite is an important health problem in many rural communities in tropical countries. However, little is known about the lasting physical and mental health effects following a bite. We recently reported that mental problems, with harmful social outcomes, can occur in many people after they are bitten by a snake. As the affected are often poor farmers or manual workers, this may affect their livelihoods. We, therefore, performed a trial which looked at the effectiveness of short psychological interventions, lasting about 15 and 20 minutes, which can be provided by even non-specialist doctors, in reducing these mental and social problems in people bitten by snakes. Our results show that such interventions may indeed be helpful to reduce some of these problems, but more research is needed to improve these interventions, especially so that they that can reduce post-traumatic stress disorder and depression after snakebite.
| Snakebite causes significant morbidity and mortality and the highest burden exists in the poorer rural populations of South Asia, Southeast Asia, and sub-Saharan Africa. Globally, it has been estimated that snakebite results in as many as 1.8 million envenomings and 94 000 deaths each year [1]. In Sri Lanka, about 35,000 to 40,000 persons are treated in hospital for snakebite each year [2]. The actual number of bites is likely to exceed this number, as some victims seek traditional forms of treatment and do not come into contact with mainstream health services. The affected are often poor farmers or manual workers in whom the snakebite may result in loss of livelihood and good health [3].
Though the physical impacts of snakebite are well documented and researched, the long term psychological impact of snakebite remained largely unexplored. We recently reported that significant delayed psychological morbidity occurs in victims of snakebite, including increased rates of depression, anxiety and post-traumatic stress disorder (PTSD), with associated negative psychosocial impact [4]. This study showed that snakebite victims had more depressive symptoms than controls based on the modified Beck Depression Scale and more symptoms of depression and anxiety measured by the Hopkins Symptoms Checklist. Fifty four per cent of the cohort met criteria for depressive disorder compared to only 15% of the controls. PTSD occurred in 22% of patients and 27% claimed that the snakebite caused a negative change in their employment with 10% stopping work and 17% reporting residual physical disability. This is hardly surprising, as the affected are often poor farmers or manual workers in whom the snakebite may result in loss of livelihood [4].
There are many myths surrounding snakebite. In many Asian and African cultures snakes are considered deities and hence snakebite can be misconstrued as punishment from the gods. These myths are not merely confined to Asia and Africa as is evident by the snake receiving prominence in most western health emblems due to its perceived mythical prowess and impact on health [5]. To this day in Sri Lanka, a ritualistic dance called the “sanninatuma” is performed to exorcise the demons causing eighteen common illnesses, and one of these dances specifically targets “Naga Sanniya” which literally translates to “snake madness” and is characterized by nightmares involving snakes and inability to perform activities of daily living [6]. “Naga sanniya” may be the first recorded description of psychological disability following snakebite. The physical impact of snakebite can be coloured by cultural beliefs, and it is not uncommon for people to believe that snakebite will lead to sapping of strength, diminished physical abilities, blindness, physical disfiguration and an overall inability to function at their premorbid level leading to avoiding work, withdrawing from social life and resigning themselves to a life of suffering [4].
Improved emergency care, availability of antivenom, and increasing numbers of victims seeking hospital based care has led to a reduction in mortality rates and physical complications of snake bite [7]. However, until recently, delayed psychological morbidity following snakebite was not recognized and psychological support is rarely offered to victims, although its inclusion into snakebite management protocols has been recommended [7,8].
Psychoeducation, which refers to the education offered to individuals with a mental health condition and their families to help empower them and deal with their condition in an optimal way, has been shown to be of benefit in the treatment of many mental conditions [9]. Although there are no precedents for the use of psychoeducation following animal bites, there is evidence for the usefulness of psychoeducation following traumatic situations [10]. Trauma based cognitive behavioural therapy is also recommended in the treatment of mental illness following stressful life events [11]. There is evidence to suggest that brief interventions based on trauma focused cognitive behavioural therapy are effective in the prevention of PTSD [12].
The aim of this study was to assess the effectiveness of a brief psychological intervention provided by non-specialist doctors in reducing delayed psychological morbidity and negative social impact associated with snakebite envenoming.
This was a single blind, randomized, controlled, parallel design trial of a brief psychological intervention in snakebite victims to determine its effectiveness in reducing psychiatric symptoms, depression, psychosocial disability and PTSD.
Ethical approval for the study was obtained from the Ethics Review Committee of the Faculty of Medicine, University of Kelaniya, Ragama. The study was registered as a clinical trial with the Sri Lanka Clinical Trials Registry (SLCTR/2011/003). All subjects provided written informed consent.
The study was conducted in the Polonnaruwa District General Hospital in Sri Lanka from August 2011 to April 2014. Polonnaruwa is situated in the northeast of the country and has a predominantly rural agricultural population. The area has one of the highest rates of snakebite envenomings in the country [13]. All snakebite victims admitted to hospital identified as being envenomed and requiring treatment with antivenom were eligible for inclusion. Exclusion criteria were those under 18 years of age, those with known mental illness, and those without basic fluency in the Sinhala language.
After snakebite patients had received standard medical treatment for their snake envenoming, they were randomized to one of three study arms before discharge from hospital, after obtaining written informed consent. Group A received no psychological intervention, Group B received a psychological intervention based on psychological first aid and psychoeducation at time of discharge from hospital, and Group C received psychological first aid and psychoeducation at time of discharge and were subsequently recalled one month following discharge from hospital and provided with a psychological intervention based on trauma based cognitive behavioural therapy principles. All participants were assessed six months following discharge from hospital.
A sample size of 195 (65 in each arm of the study) was calculated in order to detect a 50% reduction in rates of depressive disorder in the intervention group assuming an incidence of 54% as detected in our previous study [4], a power of 80% and a significance level of 0.05. A 10% loss to follow up rate was assumed, resulting in an increase in the sample size to 216. We selected depression for sample size calculation as it was an important disability that was identified in our previous study [4].
The brief interventions were administered by non-specialist doctors. The non-specialist doctors involved in the study were trained by a specialist psychiatrist, initially over a period of one week with continued support over the course of the study. They were trained on communication skills and counseling via practical demonstration.
Patients were assessed for presence of psychological morbidity and functional status six months following discharge from hospital by a specialist psychiatrist blind to intervention status and trained in the usage of the study tools. Psychological distress was quantified using a number of measures: the Hopkins symptoms checklist– 25(HSCL-25) [15,16], a modified Sinhala version of the Beck depression inventory [17], the Sheehan Disability Inventory [18], and the Post-traumatic Stress Symptom Scale—Self Report (PSS-SR) [19]. All have been previously validated and used in Sri Lanka [20].
The Hopkins Psychiatric Symptom Checklist measures a combination of depressive and anxiety symptoms. It does not provide a diagnosis of illness but based on the score, classifies subjects as positive or negative for psychiatric symptoms. The Beck's modified depression scale scores were categorized into no depression (0–15), mild depression (16–24), moderate depression (25–32) and severe depression (>32) in terms of accepted figures. The established clinically significant item-average cut-off score of ≥1.75 for each sub-scale was used for the Hopkins somatic symptoms checklist. An overall cut-off of 15/30 and a domain specific cut-off of 5/10 were used for the Sheehan Disability Inventory. The generally accepted cut off score ≥20 on the PSS-SR was taken as compatible with post-traumatic stress disorder.
The outcomes assessed were the proportions of patients with psychiatric symptoms overall, positive symptoms of depression, positive symptoms of anxiety (based on the Hopkins Somatic Symptoms Checklist), depression (based on Beck’s Modified Depression Scale), disability in family life, social life and work (based on Sheehan’s Disability Inventory) and PTSD.
Analysis of quantitative data was done using SPSS version 16 on an intention to treat basis. Chi square test for trend and Fisher’s exact test were used to assess the differences between the study groups. After adjustment for multiple testing, a p<0.0125 was considered significant.
There were 225 snakebite victims [168 males, mean age 42.1 (SD 12.4) years] who were randomized into one of the three study arms (n = 75 each). Of these, 202 (89%) (Group A, n = 68; Group B, n = 65; Group C, n = 69) completed the study and were assessed at 6 months after discharge from hospital (Fig 1).
Male farmers of working age were highly represented in the study sample. There were no differences in age, sex or occupation between the three groups (Table 1).
Overall, the biting species was identified in 24.1%; the proportion of species identified was similar in the three groups. The proportion of patients who developed severe reactions to AVS [21] and who were treated in intensive care were similar in the three groups. Four patients developed tissue necrosis, but none required amputation.
At six months, the overall proportion of patients who were positive for psychiatric symptoms of depression and anxiety was 18/68 (26.5%) in Group A, compared to 9/65 (13.8%) in Group B and 6/69 (8.7%) in Group C. This decreasing trend was statistically significant (Chi square test for trend = 7.901, p = 0.005). On further analysis, this decreasing trend was seen for both symptoms of anxiety (Chi square test for trend = 11.256, p = 0.001), and symptoms of depression (Chi square test for trend = 5.793, p = 0.016) (Table 2).
Depression was diagnosed in 21/68 (30.9%) patients in Group A, 17/65 (26.2%) patients in Group B and 18/69 (26.1%) patients in Group C. These rates did not show a statistically significant trend (chi square for trend = 0.391, p = 0.532) (Table 3).
However, on further analysis, the rate of severe depression was significantly higher in Group A (10.3%) compared to Group B (1.5%) and Group C (0) (Fisher’s exact test p = 0.004) (Table 4).
The overall prevalence of disability was 18/68 (26.5%) in Group A compared to 11/65 (16.9%) in Group B and 6/69 (8.7%) in Group C with a statistically significant decreasing trend from Group A through Group B to Group C (chi square for trend = 7.551, p = 0.006; Table 5). On further analysis, this decreasing trend was seen in relation to disability in family life (p = 0.006) and social life (p = 0.005), but not in relation to disability at work (p = 0.056).
The overall prevalence of PTSD was (17/202) 8.4%. The proportion of patients with PTSD was 7/68 (10.3%) in group A, compared to 8/65 (12.3%) in group B and 2/69 (2.9%) in Group C which was not statistically significant (Chi-square for trend = 2.448; p = 0.118) (Table 6).
A sensitivity analysis was conducted assuming the worst possible scenario (worst clinical outcome) in the patients who dropped out after randomization without receiving the intervention and outcome assessment. The decreasing trend observed from Group A through Group B to Group C in the results of Hopkins Psychiatric Symptoms Checklist and Sheehan Disability Inventory remained statistically significant.
We found that brief psychological interventions by non-specialist doctors, which included psychological first aid and psycho-education at discharge and a single cognitive behavioural therapy based intervention one month after discharge, appeared to reduce psychiatric symptoms of anxiety and depression, and improve overall functionality, especially those related to family and social life in victims of snake bite envenoming. The interventions did not reduce the proportion of patients with PTSD or overall depression, but appeared to reduce severe depression. The apparent discrepancy in the fact that overall depression was not reduced though there was a reduction in severe depression in the intervention groups, may reflect the fact that the interventions are simply effective in converting severe depression to milder forms. Our findings have important implications given the potential the socio-economic burden that may result from psychological disability following snakebite envenoming in this predominantly subsistence farming population. The proportion of patients who developed severe reactions to AVS, who developed tissue necrosis and who were treated in intensive care were similar in the three groups. These factors are, therefore, unlikely to have influenced the differences in outcome between the three groups.
Psychological interventions following trauma aimed at preventing psychiatric illness have generally shown mixed results [14]. Trauma based cognitive behavioural therapy is recommended in the treatment of psychological disorders following stressful life events [11] because there is evidence to suggest that brief interventions based on trauma focused cognitive behavioural therapy is effective in the prevention of PTSD [12]. Recent studies also support single interventions such as ours as being useful [22].
A study on psychological interventions following trauma from Chile has shown reduction in rates of depression and improvement of functional levels, although PTSD rates did not improve [23]. Cognitive errors are known to dominate the thoughts of a person afflicted with secondary depression [24]. Therefore, they may be more easily targeted by providing psychological first aid and psychoeducation and cognitive behavioural therapy. However PTSD constitutes a constellation of symptoms less under cognitive control. This may not respond to brief psychological interventions. Rates of PTSD were not reduced by the interventions conducted in our study and this reflects the findings of studies conducted elsewhere [12,25], and underscores the view that one intervention may work better for a particular symptom but not for another. The challenge would be to design a brief intervention that is able to reduce all types of symptoms to some extent, although it would possibly not be the ideal intervention for each individual symptom.
Although there was a trend towards reduction of psychiatric symptoms as well as improved functionality in both intervention groups, the best outcome was seen in the group which received both psychological first aid and psychoeducation and the cognitive behavioral therapy based intervention. Though the interventions were designed on existing models and delivered in a structured manner, the factors that played a role in reducing psychological morbidity remain to be determined. Common therapeutic factors such as confidence in the therapist, explaining illness factors and interaction with the therapist have a positive effect regardless of type of psychotherapy provided [25]. We are unable to ascertain if it was the specific therapy provided or the interaction and counseling provided by a health care professional that caused the beneficial effect seen in our study. Further analysis of the interventions will be required to improve them further. Nevertheless, the fact that psychological interventions provided by non-specialist doctors seemed to help in reducing some psychological morbidity in snakebite victims is encouraging and warrants further exploration.
Psychological therapies are provided by health care professionals other than psychiatrists and psychologists. These include general practitioners, counsellors, therapists, social workers and psychiatric nurses. The lack of qualified psychotherapists is a worldwide problem [26]. Snakebites occur predominantly in poorer social settings in Asia and Africa, and so it is unlikely that specialist mental health services will be optimal in these settings. Training non-specialist doctors to provide brief psychological interventions may be a viable alternative option. Having a non-specialist doctor providing the psychological intervention as opposed to a psychiatrist or psychologist may even improve compliance, given the stigma associated with mental illness which remains a significant barrier towards accessing mental health care [27].
Our study has limitations. As ours was an intention to treat analysis, the dropout rate after randomization may have affected the results. However, there was no change in the overall results after a sensitivity analysis was conducted assuming the worst possible clinical outcome in the patients who dropped out after randomization without receiving the intervention and outcome assessment. Also, as more than one outcome of the intervention was assessed, multiple comparisons had to be made. However adjustments were made for multiple comparisons when calculating significance. Psychiatric caseness was not confirmed by a psychiatrist or by using detailed protocols, as we used only screening instruments for depression and anxiety. However, all of the instruments we used have been previously validated and used in Sri Lanka [20]. We did not have any information on pre-event rates of depression or anxiety in the study group which may have affected our results. In an attempt to minimize such an effect we excluded patients with known mental illness from the study. It is commonly believed that some snakes induce more fear and could be psychologically more traumatic than bites of other snakes. Although of potential interest, we were unable to perform a subgroup analysis of our results based on biting species as the offending snake was identified in a minority of cases. This is the usual situation in rural Sri Lanka where the offending snake is brought to hospital very infrequently, venom antigen detection is not routinely available, and bites often occur in situations where the snake is not even seen clearly—eg. in scrub jungle, paddy fields.
In conclusion, this study is the first attempt to treat the recently recognized problem of psychological morbidity following snakebite envenoming. The results clearly suggest a role for brief psychological interventions post-snake bite, provided by non-specialist doctors that should be achievable even in settings in which mental health services are sub-optimal. However, further research is required to refine the types of interventions, focusing on PTSD and depression.
|
10.1371/journal.pcbi.1006770 | A numerical approach for a discrete Markov model for progressing drug resistance of cancer | The presence of treatment-resistant cells is an important factor that limits the efficacy of cancer therapy, and the prospect of resistance is considered the major cause of the treatment strategy. Several recent studies have employed mathematical models to elucidate the dynamics of generating resistant cancer cells and attempted to predict the probability of emerging resistant cells. The purpose of this paper is to present numerical approach to compute the number of resistant cells and the emerging probability of resistance. Stochastic model was designed and developed a method to approximately but efficiently compute the number of resistant cells and the probability of resistance. To model the progression of cancer, a discrete-state, two-dimensional Markov process whose states are the total number of cells and the number of resistant cells was employed. Then exact analysis and approximate aggregation approaches were proposed to calculate the number of resistant cells and the probability of resistance when the cell population reaches detection size. To confirm the accuracy of computed results of approximation, relative errors between exact analysis and approximation were computed. The numerical values of our approximation method were very close to those of exact analysis calculated in the range of small detection size M = 500, 100, and 1500. Then computer simulation was performed to confirm the accuracy of computed results of approximation when the detection size was M = 10000,30000,50000,100000 and 1000000. All the numerical results of approximation fell between the upper level and the lower level of 95% confidential intervals and our method took less time to compute over a broad range of cell size. The effects of parameter change on emerging probabilities of resistance were also investigated by computed values using approximation method. The results showed that the number of divisions until the cell population reached the detection size is important for emerging the probability of resistance. The next step of numerical approach is to compute the emerging probabilities of resistance under drug administration and with multiple mutation. Another effective approximation would be necessary for the analysis of the latter case.
| Drug therapies for cancer have dramatically succeeded since molecular-targeted drugs have been introduced in medical practice; however, drug treatment often fails owing to the emergence of drug-resistant cells. A variety of approaches, including mathematical modeling, has been undertaken to clarify the mechanism of resistance and subsequently avoid resistance to therapy. This paper proposes one of the mathematical approaches that uses a stochastic model and provides the emerging probabilities of resistance at detection size.
| Oncogenic pathways have been investigated using molecular biology techniques, which have helped elucidate the molecular mechanism of cancer growth, invasion, and metastasis among other aspects. The findings from these investigations have encouraged the development of anti-cancer drugs that inhibit specific oncogenic pathway and have helped improve clinical outcomes dramatically [1] [2]. But there exists some percentage of patients who have no response to these kinds of drugs. One of the reasons for the lack of response is the point mutation of a specific gene. Cancer cells mutate and acquire resistance to the anti-cancer drug, posing a significant obstacle for curing cancer [3] [4].
Several recent studies have attempted to understand the proliferation of cancer cells by employing mathematical models for the process of biological evolution [4] [5]. The mutation occurs randomly in the cell population and the number of resistant cells in the population increases as the cancer cells grow. Because expanding process of resistant cells in cancer cell population is like the dynamics of biological evolution, this process of expanding mutation cells can be viewed as an evolutionary process within the body occurring within a short span of time [5]. Mathematical models are often used to elucidate the dynamics of evolutionary process and have been studied to understand the mechanism through which cancer cells develop drug resistance [4] [6] [7] [8] [9] [10].
Iwasa et al. [11] analyzed the dynamics of resistant mutants in the exponential growth of cancer cells. The authors used a continuous-time branching process to calculate the probability of resistance, and the probability distribution of resistant cells when the population of cancer cells attains a certain detection size in the absence of therapy. They observed that the probability of resistance is an increasing function with the product of detection size and mutation rate. They concluded that the probability of resistance and the average number of resistant cells increase with the number of cell divisions over the course of the cancer.
Haeno et al. [12] extended Iwasa’s model to cancer cells carrying two mutations. They also used a continuous-time branching process to calculate the probability of formation of at least one cell carrying two mutations, and the probability distribution when the population reaches a certain size. Their findings were similar to those from Iwasa’s study.
Foo et al. [13] [14] modelled the cancer cell population during treatment with a continuous-time birth and death process. They measured the effect of drugs in reducing the proliferation rate of drug-sensitive cells. They studied resistance dynamics during therapy under a general time-varying treatment schedule. They coupled their stochastic framework with pharmacokinetic models incorporating the processes of drug absorption and elimination within the body. They calculated the probability of resistance arising during continuous and pulsed administration strategies. They used their estimates of probability of resistance and population size of drug-resistant cells to determine an optimum drug administration schedule that would minimize the risk of resistance.
The mathematical approach used in these models were analytical ones. The probability of resistance was obtained by solving differential equations and confirmed by computer simulations. To obtain the analytical solution, various ways to derive the solution of equations were devised. For example, the authors in [11] [12] [13] [14] handle the cell size as continuous variables in their calculations, though it is considered appropriate for addressing the discrete state space when the dynamics of cell size is discussed.
The purpose of this paper is to present numerical approach for computing the emerging probability of resistant cells. We model the process of cancer progression through a discrete state space, continuous time Markov chain. Then, we transform it into an embedded Markov chain to reduce the computational effort for computing the emerging probabilities of resistant cells.
This paper first explains our models for computing the number of resistant cells. Then the two ways of computation are presented: Exact analysis and aggregation approximation. The approximation method is introduced as efficient way to compute. Second, the computed values of exact analysis and those of approximate are compared. At the same time the computed values of exact analysis also are compared with those of previous study. The relative errors are used for evaluating the accuracy of these computed values. Third, values of computer simulation are compared with those of approximation. This comparison is performed at detection size over 10000. The 95% confident intervals are used to confirm and evaluate the accuracy of computed values of approximation. The execution time of both methods are also compared. Lastly, the parameter dependency on the computed values of emerging probabilities of resistant cells. The parameter dependency of division rate, death rate and mutation rate are investigated by computing the probabilities of resistance with changing these values of parameters. Factors of effecting the emerging probabilities of resistance are discussed.
Consider an expanding cancer cell population. There are two types of cancer cells: drug-sensitive and drug-resistant. The sensitive cells divide and die at a rate of λ and μ, respectively. The probability of mutation when a sensitive cell divides (i.e., the probability of formation of a resistant cell) is γ, and the probability of formation of a sensitive cell is 1 − γ. The resistant cells divide and die at a rate of α and β, respectively. Our objective is to obtain the distribution of resistant cells when the total cell population reaches detection size, which is denoted by M. Let us denote the total number of cells (both sensitive and resistant) and the number of resistant cells by m and n, respectively. Then, the number of sensitive cells increases at the rate of λ(1 − γ)(m − n) and decreases at the rate of μ(m − n). The number of resistant cells increases at the rate of λγ(m − n) + αn and decreases at the rate of β n. Let us consider the two-dimensional Markov chain with states (m, n), where m = 0, 1, 2, ⋯, M; n = 0, 1, 2, ⋯, m. Let us consider the process starting at the state (m, n) = (1, 0). The process can end either at extinction (m = 0) or when the cell population reaches detection size (m = M).
Let us denote the state of the process (m, n) after the t-th event (cell division or death) as (mt, nt). The transition probabilities of this process are given as follows:
P r { ( m t + 1 , n t + 1 ) = ( i + 1 , j ) | ( m t , n t ) = ( i , j ) } = λ ( 1 − γ ) ( i − j ) / Γ i , j (1)
P r { ( m t + 1 , n t + 1 ) = ( i + 1 , j + 1 ) | ( m t , n t ) = ( i , j ) } = { λ γ ( i − j ) + j α } / Γ i , j (2)
P r { ( m t + 1 , n t + 1 ) = ( i − 1 , j ) | ( m t , n t ) = ( i , j ) } = μ ( i − j ) / Γ i , j (3)
P r { ( m t + 1 , n t + 1 ) = ( i − 1 , j − 1 ) | ( m t , n t ) = ( i , j ) } = j β / Γ i , j (4)
Here Γi,j = (i − j)(λ + μ) + j(α + β) is the sum of the rates, normalizing the total probability to 1. Note that the Markov chain (m, n) is homogeneous.
Let us define the set of states in which the total number of cells is i, {(i, 0),(i, 1) ⋯ (i, i)} as level i. Let us denote the transition probability sub-matrix from level i to level i + 1, and from level i to level i − 1 as Pi and Qi, respectively. The element of Pi in the k-th row and i-th column is the transition probability from (i, k) to (i + 1, k). The corresponding element of Qi is the transition probability from (i, k) to (i − 1, k).
Pi and Qi are expressed as follows:
P i =( i λ ( 1 − γ ) Γi 0 i λ γ Γi 0 0 ⋯ ⋯ ⋯ ⋯ 0 0 ( i − 1 ) λ ( 1 − γ ) Γi 1 ( i − 1 ) λ γ + α Γi 1 0 ⋯ ⋯ ⋯ 0 0 0 ( i − 2 ) λ ( 1 − γ ) Γi 2 ( i − 2 ) λ γ + 2 α Γi 2 ⋯ ⋯ ⋯ 0 ⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯ 0 0 ⋯ ⋯ ⋯ λ ( 1 − γ ) Γi i − 1 λ γ + ( i − 1 ) α Γi i − 1 0 0 0⋯⋯⋯⋯⋯⋯0 i α Γi i )
Q i =( i μ Γi 0 0 0 ⋯ ⋯ 0 β Γi 1 ( i − 1 ) μ Γi 1 0 ⋯ ⋯ 0 0 2 β Γi 2 ( i − 2 ) μ Γi 2 ⋯ ⋯ 0 ⋯⋯⋯⋯⋯⋯⋯ 0 0 0 ⋯ ( i − 1 ) β Γi i − 1 μ Γi i − 1 0 0⋯⋯0 i β Γi i )
where Pi is the (i + 1) × (i + 2) matrix, and Qi is the (i + 1) × i matrix. We can now express the transition probability matrix of the process (m, n) by using Pi and Qi as follows:
S = ( 1 0 0 0 0 ⋯ ⋯ ⋯ 0 Q 1 0 P 1 0 0 ⋯ ⋯ ⋯ 0 0 Q 2 0 P 2 0 ⋯ ⋯ ⋯ 0 0 0 Q 3 0 P 3 ⋯ ⋯ ⋯ 0 ⋯⋯⋯⋯⋯⋯⋯⋯ 0 ⋯ ⋯ ⋯ ⋯ ⋯ Q M − 1 0 P M − 1 0 ⋯ ⋯ ⋯ ⋯ ⋯ ⋯ 0 I M )
Note that the states in levels 0 and M are absorbing states.
We take the submatrix T from S as follows:
T = ( 0 P 1 0 0 ⋯ ⋯ ⋯ 0 Q 2 0 P 2 0 ⋯ ⋯ ⋯ 0 0 Q 3 0 P 3 ⋯ ⋯ ⋯ 0 ⋯⋯⋯⋯⋯⋯ 0 ⋯ ⋯ ⋯ ⋯ ⋯ Q M − 1 0 )
The submatrix corresponds to the transition probability matrix from transient states to transient states.
Then, mt denotes the total number of cancer cells and nt denotes the number of resistant cells after the t-th event (cell division or death). Here, we define (mt, nt) as the state of the number of total cells and resistant cells after the t-th event. In addition, the probability being at the state is expressed as π ( i , j ) t = P r { ( m t , n t ) = ( i , j ) }.
Now, we define the probability distribution vector:
π i t = ( π t ( i , 0 ) , π t ( i , 1 ) , π t ( i , 2 ) , ⋯ , π t ( i , i − 1 ) , π t ( i , i ) )
Here, we consider the process starting at one sensitive cell and no resistant cell, so the initial state of the process is (m0, n0) = (1, 0) i.e.,
{ π0i=(1,0)(i=1)π0i=0(i≠1)
Our objective is to obtain the probability distribution and the emerging probabilities of resistant cells when the total number of cells reaches M. As level 0 and level M are absorbing states, the emerging probability of resistant cells is expressed by lim t → ∞ π ( M , j ) t 1 − π ( 0 , 0 ) t (j = 0, 1, 2, 3, ⋯, M).
In general, to calculate the distribution of probability in the absorbing states, we should obtain the fundamental matrix as shown below:
I + T + T 2 + T 3 + ⋯ = ( I − T ) − 1 (5)
The matrix T is large, and the calculation of the fundamental matrix involves high computational complexity as the matrix becomes large. Therefore, we propose another algorithm to reduce complexity.
Let πi = (π(i,0), π(i,1), π(i,2), ⋯, π(i,i−1), π(i,i)) be the probability distribution of the process at first arrival to level i, where ∑ k = 0 iπ ( i , k ) = 1. The discrete-time chain {πi} is said to be embedded in {π i t}, so it is referred to as embedded Markov chain. Then the probability distribution when the total number of cells reaches M is equivalent to πM. Let us now denote the probability matrix that the state at first arrival to level i + 1 is (i + 1, ⋅) under the condition that the state at first arrival to level i is (i, ⋅) by Fi. There are two paths of transition for the state in level i to reach level i+1. In one, the cells transition directly from level i to level i + 1 in a single step. In the other, the cells first transition from level i to level i − 1 and then reach level i + 1 through level i. Using the transition probability matrices Pi and Qi, we obtain the recurrence formula as follows:
F i = P i + Q i F i − 1 F i ( i = 2 , 3 , ⋯ , M ) (6)
Thus, we have
F i = ( I i + 1 − Q i F i − 1 ) − 1 P i (7)
where F1 = P1, and Ii+1 is the (i + 1) × (i + 1) identity matrix. When the total number of cells reaches M, the probability distribution πM is calculated by the following formula:
{ π2=(1,0)F1πi+1=πiFi(i=3,4,⋯,M−1)
Here, Fi is an i × (i + 1) matrix. Thus, the complexity of calculation is much lower in this case than in the case of the fundamental matrix.
The algorithm proposed in the previous section still includes the inverse of the (M − 1) × M matrix. We propose an approximate aggregation to further reduce the complexity of the calculation.
If the level of states is greater than m, we aggregate the states in which the number of resistant cells is greater than m to a single state (Fig 1). The level m + k is the set of states as follows:
{ ( m + k , 0 ) , ( m + k , 1 ) , … , ( m + k , m − 1 ) , ( m + k , m ) , … , ( m + k , m + k ) }
From among these, we aggregate the states (m + k, m), (m + k, m + 1), …, (m + k, m + k) to one state and denote the aggregated state by (m + k, m*). Then, state (m + k, m*) can transit to (m + k + 1, m*), (m + k − 1, m*), or (m + k − 1, m − 1).
Let us denote the rate of transition from state (m + k, m*) to each of state (m + k + 1, m*), (m + k − 1, m*), and (m + k − 1, m − 1) by R A k, R B k, and R C k, respectively. These rates can now be expressed as follows:
R A k = ∑ x = 0 k { ( k − x ) λ ( 1 − γ ) + ( k − x ) λ γ + ( m + x ) α } × π t ( m + k , m + x ) ∑ v = 0 k π t ( m + k , m + v ) = [ 1 2 k ( k + 1 ) λ + { ( k + 1 ) m + 1 2 k ( k + 1 ) } α ] × π t ( m + k , m + x ) ∑ v = 0 k π t ( m + k , m + v ) (8)
R B k = ∑ y = 0 k { ( k − y ) μ × π t ( m + k , m + x ) ∑ v = 0 k π t ( m + k , m + v ) + ∑ y = 1 k { ( m + y ) β × π t ( m + k , m + x ) ∑ v = 0 k π t ( m + k , m + v ) = [ 1 2 k ( k + 1 ) μ + { ( k m + 1 2 k ( k + 1 ) } β ] × π t ( m + k , m + x ) ∑ v = 0 k π t ( m + k , m + v ) (9)
R C k = m β × π t ( m + k , m + x ) ∑ v = 0 k π t ( m + k , m + v ) (10)
Here, we assume that the following probabilities are the same:
π t ( m + k , m ) = π t ( m + k , m + 1 ) = ⋯ = π t ( m + k , m + k ) = 1 k + 1 (11)
Now we have
R A k = ∑ x = 0 k { ( k − x ) λ ( 1 − γ ) + ( k − x ) λ γ + ( m + x ) α } × 1 k + 1 = [ 1 2 k ( k + 1 ) λ + { ( k + 1 ) m + 1 2 k ( k + 1 ) } α ] × 1 k + 1 (12)
R B k = ∑ y = 0 k { ( k − y ) μ × 1 k + 1 + ∑ y = 1 k { ( m + y ) β × 1 k + 1 = [ 1 2 k ( k + 1 ) μ + { ( k m + 1 2 k ( k + 1 ) } β ] × 1 k + 1 (13)
R C k = m β × 1 k + 1 (14)
Using the sum of the rates Γ(k), which is described as
Γ ( k ) = R A k + R B k + R C k = [ 1 2 k ( k + 1 ) ( λ + μ ) + { ( k + 1 ) m + 1 2 k ( k + 1 ) } ( α + β ) ] , (15)
we can obtain the probabilities of transition from state (m + k, m*) to each of (m + k + 1, m*), (m + k − 1, m*), and (m + k − 1, m − 1) through the following equations:
R A k Γ( k )= k λ + ( 2 m + k ) α k ( λ + μ ) + ( 2 m + k ) ( α + β ) (16)
R B k Γ( k )= k ( k + 1 ) μ + { 2 k m + k ( k + 1 ) } β ( k + 1 ) ( λ + μ ) + { 2 m ( k + 1 ) + k ( k + 1 ) ( α + β ) } (17)
R C k Γ( k )= 2 m β ( k + 1 ) ( λ + μ ) + { 2 m ( k + 1 ) + k ( k + 1 ) ( α + β ) } (18)
After the aggregation, the transition probability submatrix from level m + k to level m + k + 1, and from level m + k to level m + k − 1, which are denoted by P ˜ m + k and Q ˜ m + k, respectively, become:
P ˜ m + k = ( i λ ( 1 − γ ) Γi 0 i λ γ Γi 0 0 ⋯ ⋯ ⋯ 0 0 ( i − 1 ) λ ( 1 − γ ) Γi 1 ( i − 1 ) λ γ + α Γi 1 0 ⋯ ⋯ 0 0 0 ( i − 2 ) λ ( 1 − γ ) Γi 2 ( i − 2 ) λ γ + 2 α Γi 2 ⋯ ⋯ 0 ⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯ 0 0 0 ⋯ ⋯ λ ( 1 − γ ) Γi i − 1 λ γ + ( i − 1 ) α Γi i − 1 0 0 ⋯⋯⋯⋯ ⋯ 0 R A k Γ( k ) )
Q ˜ m + k = ( i μ Γi 0 0 0 ⋯ ⋯ 0 β Γi 1 ( i − 1 ) μ Γi 1 0 ⋯ ⋯ 0 0 2 β Γi 2 ( i − 2 ) μ Γi 2 ⋯ ⋯ 0 ⋯⋯⋯⋯⋯⋯⋯ 0 0 0 ⋯ ( i − 1 ) β Γi i − 1 0 0 0 ⋯⋯ R C k Γ( k ) R B k Γ( k ) )
where P ˜ m + k and Q ˜ m + k are (m + k) × (m + k) matrices. Then we can approximately calculate Fi (i = m, m + 1, m + 2, ⋯) using
F i = ( I i + 1 − Q ˜ i F i − 1 ) − 1 P ˜ i (19)
The largest size of the matrix Fi is (m + 1) × (m + 1), so we can obtain the probability more easily than through exact analysis.
In this section, we present the numerical results. First, we computed the emerging probabilities of resistance by exact analysis and approximate aggregation. Then we computed relative errors between these numerical results to evaluate the accuracy of approximation. Numerical results computed by the formula in previous study [11] were also compared with the approximation. Second, we computed the emerging probabilities by our approximation method when the detection size was over 10000. Computer simulation was performed to compare the results with approximation. The averages and the 95% confidential intervals of simulation results were computed. Last, we examined the effect of parameter change on emerging probabilities of resistance.
Fig 2 shows the results of emerging probabilities of resistance using the exact analysis method and the aggregation approximation method. We set the detection size M at 500, 1000, and 1500 because the exact analysis took a considerable time when M was greater than 1500. We computed each relative error at variable aggregation size; m = 10 ⋯ 100. Computations using the aggregate approximation and the formula in previous study. [11] were executed and computed the relative errors. The relative errors of approximation method were no lower than 10−6 order regardless of aggregation size. For comparison, the relative error of the previous study. [11] was no less than the order of 10−4. The relative error became smaller as the detection size becomes larger because they regarded the number of cells as continuous valuables when they calculated the probability of resistance. These results are shown in Figs 2 and 3 and Supporting information S1 Table.
In this study, we modeled the cell progression process by a two-dimensional Markov process that was characterized by the total number of cells and the number of resistant cells. We calculated the emerging probability of resistance when the total number of cells reached detection size M, starting from one drug sensitive cell. This probability was equivalent to the probability of being absorbed in the absorbing state wherein the total number of cells was M. To calculate this probability, we needed to inverse matrix of size (M + 2)(M − 1)/2 × (M + 2)(M − 1)/2 and the complexity of calculation was O(M6).
We employed the embedded Markov analysis approach and observed only the timepoint at which the total number of cells change. We also derived the recurrence formula for state transition probabilities of the first visit to the set of states wherein the total number of cells was n + 1 from the set of states wherein the total number of cells was n. Using this approach and the formula, we proposed an efficient calculation method for emerging probabilities of resistant cells that required M times calculation of the (M + 1) × (M + 1) inverse matrix only. Then, we calculated the emerging probabilities of resistance when the number of cells reached M = 1000. However, it took significant execution time for realistic detection sizes such as M = 10000 or 100000; thus, we designed a more practical method for calculation.
The approximation approaches inverting a matrix of large dimension have been intensively studied, and this approach may be useful to reduce the execution time for emerging probabilities of resistance. However, as shown above, once the number of resistant cells reached 100, the probability of extinction of resistant cells is under 10−5, the information of probability distribution of over 100 resistant cells was not very valuable from the viewpoint of treatment strategy. Hence, we aggregated the states (m + k, m), (m + k, m + 1), … ., (m + k, m + k) to a single state for each k(k = 1, 2, …, M − m). This approximation method required computing M times calculation of (m + 1) × (m + 1) inverse matrix to obtain the approximate solutions for the emerging probabilities of resistant cells. We set m = 100 in the numerical analysis, because the calculation was completed in a practical execution time for a realistic detection size such as100000.
The numerical analysis showed that approximation errors of emerging probability are negligibly small when M = 500 and M = 1000 and that our approximation method demonstrated the same level of accuracy when M = 1000 or the higher level of accuracy when M = 500 compared to the result of a previous study. We also performed a stochastic computer simulation to confirm the results of the approximation method. The simulation performed 1000000 × 10 runs to obtain a 95% confidence interval of the emerging probability of resistance. The results of emerging probability of resistance by the approximation method fell within the 95% confidence interval, and the execution time of our approximation method was considerably shorter than that obtained for our simulation.
The numerical results demonstrated that the probability of resistance was chiefly dependent on the number of cell divisions until the cell population reached the detection size. It is reasonable as drug-resistant cells are generated by mutation in the process of cell division. A large population of cancer cells would have a greater likelihood of generating resistant cells via mutation. As only a few resistant cells exist in the early stage of cancer, if any, the possibility of extinction of resistant cells owing to natural causes cannot be disregarded.
The proposed method in this paper was able to track the transition of cell size. By applying this method, we would be able to follow the transition of the number of cells under drug administration. The next step in our study is to design the treatment strategy based on the analysis under drug administration.
In many cases, multiple different mutations can confer resistance and the mutagenic processes leading to such mutations may be different. Then, clones may have different growth and death rates in accordance depending on the types of mutations. It is of interest to compute the number of resistant cells of multiple types when the cell population reaches the detection size. However, it would be more difficult to extend the framework in this paper to the case of multiple resistance mutations. For example, it is necessary to analyze a three-dimensional Markov process in the case of two types of mutations. Thus, another effective approximation would be necessary for the analysis.
|
10.1371/journal.ppat.1000352 | De Novo Synthesis of VP16 Coordinates the Exit from HSV Latency In
Vivo | The mechanism controlling the exit from herpes simplex virus latency (HSV) is of
central importance to recurrent disease and transmission of infection, yet
interactions between host and viral functions that govern this process remain
unclear. The cascade of HSV gene transcription is initiated by the
multifunctional virion protein VP16, which is expressed late in the viral
replication cycle. Currently, it is widely accepted that VP16 transactivating
function is not involved in the exit from latency. Utilizing the mouse ocular
model of HSV pathogenesis together with genetically engineered viral mutants and
assays to quantify latency and the exit from latency at the single neuron level,
we show that in vivo (i) the VP16 promoter confers distinct regulation critical
for viral replication in the trigeminal ganglion (TG) during the acute phase of
infection and (ii) the transactivation function of VP16 (VP16TF) is uniquely
required for the exit from latency. TG neurons latently infected with the VP16TF
mutant in1814 do not express detectable viral proteins following stress, whereas
viruses with mutations in the other major viral transcription regulators ICP0
and ICP4 do exit the latent state. Analysis of a VP16 promoter/reporter mutant
in the background of in1814 demonstrates that the VP16 promoter is activated in
latently infected neurons following stress in the absence of other viral
proteins. These findings support the novel hypothesis that de novo expression of
VP16 regulates entry into the lytic program in neurons at all phases of the
viral life cycle. HSV reactivation from latency conforms to a model in which
stochastic derepression of the VP16 promoter and expression of VP16 initiates
entry into the lytic cycle.
| Herpes simplex virus (HSV) establishes life-long latent infections in sensory
neurons of the human host. Periodically, HSV exits latency in an infected neuron
and is transported to the body surface where it replicates, leading to recurrent
disease and infection of new hosts. We do not currently understand how entry
into the lytic cycle is blocked in neurons and latency is established. Nor do we
know how, at some time in the future, the lytic program becomes activated in the
one or two latently infected neurons which characterize a reactivation event. In
tissue culture cells, and by analogy in cells at the body surface, the HSV
replication program is initiated by the interaction of a virion protein, VP16
(brought in with the virus as a protein), with host cell factors. Here we show
that the de novo synthesis of VP16 is required for efficient viral replication
during the acute phase of infection in neurons. This indicates that latency is
favored because VP16 may not be transported efficiently to the nerve cell
nucleus. Once latency is established, the de novo expression of VP16 is an
absolute and very early requirement for the exit from the latent state. Our data
support a model of HSV reactivation in which the stochastic derepression of the
VP16 promoter and resulting expression of VP16 starts the viral lytic
program.
| Primary infection with herpes simplex virus (HSV), universally the result of close
contact with an infected individual, is accompanied by dissemination of viral
genomes into the host nervous system. Although symptoms of the primary infection
usually resolve, large numbers of viral genomes remain in a transcriptionally
repressed state within neurons of sensory ganglia and the brain for the life of the
infected individual [1]. Periodically, stimulated by various stressors,
latency is exited and infectious virions are generated in a small number
(<0.05%) of latently infected neurons [2]–[5] which
transport virus back to the body surface through innervating axons. Although
individual neurons supporting lytic viral replication do not survive this process
[4]–[6], the large reservoir of
latently infected neurons allows this cycle to occur repeatedly which is the
mechanism of transmission and the cause of serious sequellae including blindness and
encephalitis. That 70–90% of the human population worldwide is
now infected is a testament to the efficacy of this strategy [7]. There is currently no
way to either eliminate latent virus or to prevent the exit from latency and no
effective vaccine to protect the uninfected, thus transmission rates remain high. To
date, the molecular mechanisms regulating reactivation from latency remain unclear.
Identifying the interactions between the neuron and latent viral genome that result
in the exit from latency is critical toward progress in understanding and ultimately
controlling this complex process.
In a striking case of parallel evolution, most DNA viruses employ strong enhancers to
promote the transcription of the earliest viral genes [8]. HSV differs from other
DNA viruses including most other herpesviruses in that transcription of its
immediate early (IE) genes is principally dependent on a protein component of the
virion that is a potent transcriptional activator [9],[10]. This
multifunctional late gene protein, VP16 (VMW65, α-TIF, UL48), interacts with
host cell proteins including HCF-1, a cell cycle regulator, and Oct-1, a POU domain
transcription factor, to form the VP16 induced complex (VIC) that binds to TAATGARAT
elements present in the five HSV-1 immediate early gene promoters [11]–[14]. Considering the complex
in vivo life cycle of HSV, the dependence on a structural protein produced late in
the infectious cycle to initiate transcription from the viral genome presents a
conundrum. How can the latent viral genome initiate the transcription of lytic phase
genes in the absence of this crucial transcriptional activator? Studies in the early
1990's led to the dogma that VP16 is simply not involved in reactivation
[15]–[17] and that its
function in initiating the lytic cycle is fulfilled by another viral function or a
host cell factor.
There have been two long-standing hypotheses regarding the initiation of the lytic
cycle during reactivation from latency. The first hypothesis proposes that the viral
IE gene ICP0 initiates reactivation from latency [18]–[21]. The
second proposes that viral early gene expression and DNA replication precedes and is
required for efficient IE gene expression during reactivation from latency [22],[23]. In these studies, reactivation was evaluated
using axotomized and explanted ganglia. Although this assay has been widely
utilized, it has become increasingly clear that explant reactivation does not model
HSV reactivation as it occurs in vivo [24],[25]. In hindsight, this
is not a surprising finding, in that axotomized and explanted neurons rapidly
undergo radical transcriptional changes, including apoptosis [24],[26],[27]. Recent reexamination
of these hypotheses using in vivo reactivation and single neuron level approaches
have demonstrated that the exit from latency does not require either a viral DNA
pre-amplification step [28] or functional ICP0 [25]. An important clue as
to how exit from latency is regulated came from the analysis of a viral mutant
termed ΔTfi in which a 350 bp region of the ICP0 promoter, which includes
the TAATAGARAT element through which VP16 transactivates this IE gene, is deleted.
Although this mutant reactivates with wild type kinetics in explant assays [25],[29], in
vivo it is severely impaired in its ability to reactivate, suggesting that
transactivation by VP16 may indeed be critical in the regulation of reactivation in
vivo [25].
Here we report results from experiments designed to test the hypothesis that VP16
regulates the exit from latency. Our studies support the hypothesis that in elegant
simplicity, the major coordinator of IE gene expression and tegument protein, VP16,
functions to regulate entry into the lytic program at all phases of the viral life
cycle. We find that in vivo (i) the VP16 promoter confers distinct regulation
critical for viral replication in the trigeminal ganglion, and (ii) VP16
transactivating function is required for reactivation from latency. Importantly,
that VP16 transactivation function (VP16TF) is required very early in the exit from
latency is supported by (i) failure of latent viral genomes to enter the lytic cycle
(as defined by expression of lytic viral protein) uniquely in the absence of VP16TF
(ICP0 null, viral thymidine kinase null, and tsICP4 mutants do exit latency), and
(ii) the restoration of reactivation competency of ΔTfi by replacement of
the TAATGARAT element. In the nervous system, de novo expression of VP16 from the
latent viral genome allows VP16 to coordinate the expression of the viral IE genes
and thereby initiate the productive lytic cycle.
HSV initiates the viral lytic cycle under two distinct conditions, (i) following
infection of a cell by the virion, and (ii) from the latent viral genome. In the
first case, the lytic cycle is engaged through coordinated activation of the
viral IE genes by the virion associated transactivator, VP16 [11]–[14]. How the lytic cycle
is initiated from the latent genome remains unknown, although it is reasoned
that VP16, expressed with late kinetics during the lytic cycle, does
not supply this function [15], [16],
[18],
[30]–[33].
ICP0 null mutants can exit latency (demonstrated by the detection of lytic viral
protein expression), however, progression to lytic virus production
(reactivation) does not occur [25]. In addition, a mutant in which the VP16
binding site has been deleted from the ICP0 promoter also fails to reactivate in
vivo. Together, these findings raise the possibility that VP16 may play an
unexpected role in the regulation of IE genes very early in the exit from
latency. If this were the case, the regulation of VP16 must be distinct in this
context, with the protein expressed as a very early event and not as a standard
leaky late gene. To test this, we asked whether another viral promoter of
equivalent strength and kinetic class [34]–[36] could
confer “proper” regulation of VP16 in vivo. The VP5 promoter
was selected since replacement of this viral promoter with that of VP16 has been
reported previously and no measurable effect on the ability of the virus to
replicate in vivo, either at the surface or in the nervous system was observed
[37]. Thus the converse mutant in which the VP16
promoter/5′utr was replaced with that of VP5 was generated as detailed
in methods. A diagram of this mutant is
shown in Figure 1A. Three
independently derived viral mutants were characterized in vitro and in vivo.
Levels of VP16 mRNA in rabbit skin cells (RSC) infected with mutant VP5p/VP16
were not reduced compared to 17syn+ as quantified by northern blot
analysis at 6, 8 and 12 hr pi (not shown). Standard single (not shown) and
multi-step replication kinetic analysis in RSC revealed no alterations when
compared to the parental strain 17syn+ or the genomically restored
mutant VP5p/VP16-1R (Figure
1B). In order to determine the effect, if any, of this promoter exchange
on viral replication in vivo, five groups of 16 mice each were inoculated on
scarified corneas with 1×105 pfu of either VP5p/VP16-1,-3,
-5, VP5p/VP16-1R, or 17syn+. Titers of infectious virus in the eyes and
trigeminal ganglia (TG) were determined independently in three mice from each
group on days 2, 4, 6, 8, and 10 pi. The total amount of infectious virus
detected in the eyes during the acute stage of infection was not different among
the viruses compared, (area under the curve
(AUC) = 292, 261 vs. 250,013 or 281, 982,
respectively) (Figure 1C).
Note also that the peak viral replication occurring in the eyes on day 4 pi was
not different (p = 0.65; ANOVA ). In contrast,
total infectious virus detected during the acute stage in the TG was more than
200 fold reduced for the VP5p/VP16 mutants compared to the parental strain or
the genomically restored isolate (AUC = 628 vs.
155,237 or 148, 810) and the peak viral titers detected on day 4 in VP5/VP16
infected TG was more than 2 orders of magnitude lower than those detected in
17syn+ or VP5/VP16-1R infected TG (Figure 1D). The viral feedback loop between
the surface and the ganglion is well documented [38],[39].
The decline in viral titers in the eyes of VP5p/VP16 infected mice (days
6–8) most likely resulted from the absence of significant replication
in the ganglia and transport of virus back to the eye as described previously
[39]. Importantly, the high viral titers generated by
mutant VP5p/VP16 in RSC and on the corneal surface in vivo confirm the
infectious nature of the virions, which would not be the case if levels of VP16
in the tegument were deficient [30],[40],[41].
The VP16 protein produced during infection with this mutant is fully functional
and would be anticipated, if it were indeed efficiently transported to the
neuronal cell body, to initiate lytic viral infection in the neuron. However,
the replication of the VP5/VP16 mutants in TG is severely impaired, although
viral DNA is transported to the ganglion as determined by real-time PCR assay
(not shown). This strongly suggests that viral replication in neurons is in fact
not initiated by VP16 protein transported from the surface,
but rather by its synthesis in the infected neuron de novo. The profound
selective loss of replicative capacity in the TG of mice infected with the
VP5p/VP16 mutant provides the first evidence that the VP16 promoter is unique in
its ability to regulate gene expression in the nervous system and supports the
hypothesis that VP16, through distinct regulation in the TG neurons, could play
an important role in exiting latency.
The possibility that de novo expression of VP16 may be required in neurons during
both the acute stage of infection and during reactivation is suggested when
considering collectively (i) the well documented requirement for VP16
transactivating function during the acute infection in TG [16],[30]
(presumably for entry into the lytic cycle) , (ii) the inadequacy of leaky late
expression of VP16 from the VP5 promoter to support lytic viral replication
(reported here) and (iii) evidence from another α-herpes virus that
viral nucleocapsids arrive at the neuronal cell body largely devoid of VP16
[42],[43]. Framed within
conventional understanding of HSV gene regulation, the question to be asked is
straightforward, namely is VP16 expressed as a late gene, as demonstrated in
tissue culture or is VP16 expressed with distinct kinetics in neurons? The
concept of cascade gene regulation [44] and the kinetic
class of viral promoters during viral lytic cycle are fundamental to how we view
this process. However, these criteria were developed from en masse analyses of
synchronously infected cells of uniform type in the presence of drug blockades.
This experimental format cannot be recapitulated in vivo. One approach to
evaluating promoter activity in vivo is through the generation of viral
promoter/reporter mutants [45]–[55]. For this
purpose, a VP16 promoter/beta-galactosidase gene (LacZ) reporter mutant was
generated as detailed in methods and utilized to ask whether activation of the
VP16 promoter in neurons is consistent with conventional leaky late gene
expression. If this were the case, then VP16 promoter activity would be
anticipated only in neurons expressing lytic viral protein.
TG from mice inoculated with 2×105 PFU of 17VP16pLZ were
harvested on days 4 and 5 pi and processed sequentially for in situ E. coli
beta-galactosidase (b-gal) activity and for HSV proteins as detailed previously
[25],[28],[55].
Figure 2A shows two
populations of neurons evidencing activity from the viral genome. In the
majority of these neurons (464/551, 86%), VP16 promoter activity was
co-localized with lytic viral proteins. However, in 13–16%
of positive (infected) neurons, the VP16 promoter was active in the absence of
detectable lytic viral proteins. Even if very low and undetectable levels of
viral proteins are present in these neurons, the findings indicate that in
neurons, activation of the VP16 promoter can precede expression of significant
levels of viral proteins, an expression pattern inconsistent with our
understanding of late gene expression.
When examined in infected RSC, the pattern of expression of the VP16 promoter was
consistent with late kinetics in that at either high or low multiplicity of
infection (moi), b-gal activity was detected only in cells in which viral
proteins were also detected. The asynchrony of low moi infection more closely
represents infection in vivo and in this case plaques formed by 17VP16pLZ were
ringed by cells expressing viral proteins but little or no b-gal activity (Figure 2B). We examined the
expression of an IE gene promoter/reporter virus, 17-0pZ56gJ [55]
using this same assay and observed that plaques were now ringed by cells
expressing b-gal with very low levels of viral proteins present, as would be
expected for a promoter activated at the initiation of the lytic cycle. These
findings support the hypothesis that the regulation of VP16 in vivo is dependent
on cell type and different from that seen in vitro.
We have reported previously that two viral functions (ICP0 and viral DNA
synthesis) considered to play critical roles in the initiation of reactivation
from latency, are in fact not required for lytic viral protein expression
following a reactivation stimulus in vivo [25],[28].
These functions are, however, required for progression to infectious virus
production. This knowledge provides the opportunity to ask whether VP16 is
expressed in the absence of ICP0 function and in the absence of viral DNA
replication following a reactivation stimulus. If VP16 is not present, it would
suggest that this protein is not likely to be initiating entry into the lytic
cycle. If, however, VP16 is detected, it would be consistent with an early role
and reveal that in the context of reactivation, the expression of VP16 is not
dependent upon either ICP0 function and/or viral DNA replication, both of which
play a role in the regulation of late gene expression in cultured cells.
The exit from latency in vivo is highly a controlled process, restricted to a
very small percentage of those neurons latently infected per event. Despite
this, the number of neurons exiting latency and the number of neurons expressing
VP16 can be quantified using whole ganglion immunohistochemistry (WGIHC), an
assay that has been validated to provide a precise quantitative readout on the
number of neurons expressing lytic viral proteins within a ganglion [56]
Groups of mice were inoculated with either dl1403 (an ICP0 null mutant [57]) or
17tBTK- (a thymidine kinase negative mutant [24]). In the absence
of the viral thymidine kinase (TK) function, viral DNA synthesis and replication
in neurons are severely impaired. This gene is required for reactivation [58]–[61] but not for entry
into the lytic cycle from the latent viral genome [28]. The deficit in
each of these mutants results in significantly reduced total latent viral DNA
[62]–[64] and numbers of
latent infections in the TG, [25],[28],[65],[66], which in turn,
reduces the number of neurons which exit latency [25],[28],[55],[56],[66]. Nevertheless,
VP16 protein was detected at 22 hrs post hyperthermic stress (HS) in neurons in
ganglia from mice latently infected with both of these mutants (3/27 and 9/20 in
17tBTK- and dl1403 infected ganglia, respectively). Analysis of the second TG of
each pair with the anti-HSV antibody revealed no difference compared to the
number of neurons in which VP16 was detected (4/27 and 8/20 in 17tBTK- and
dl1403 infected ganglia), and numbers similar to our previous reports [25],[28]. Viral protein
expressing neurons were not detected in uninduced latently infected ganglia
(0/18). The number of neurons expressing VP16 within a positive ganglion ranged
from 1–3, and these numbers were not different than those detected
when using the anti-HSV antibody, which detects lytic viral proteins from IE,
early (E) and late (L) kinetic classes. Likewise, following HS of mice latently
infected with tsK+, a 17syn+ based mutant with a temperature
sensitive mutation in the essential viral IE transactivator ICP4 [67], VP16
was expressed in rare neurons. The number of neurons in which latency was
established with this mutant was very low (2.7%), yet VP16 was
detected in rare neurons post stress (2 neurons in 2/30 ganglia were positive).
This number is consistent with the frequency of reactivation observed in ganglia
infected with wild type 17syn+ in which there were similar low levels
of latency (1/31 positive) [56]. These findings indicate that independently,
neither ICP0, ICP4, nor viral DNA replication is required for VP16 expression
during reactivation in vivo.
In order to examine a larger population of neurons exiting latency, a chemical
blockade of viral DNA replication was used to investigate the influence of viral
DNA replication on the expression of VP16 in 17syn+ infected neurons
following a reactivation stimulus. As shown previously, acyclovir (ACV) blocked
detectable infectious virus production during reactivation in vivo [28],[68]. Infectious virus
was not detected at 22 h post HS in the 14 TG tested (0/14). Ganglion pairs from
an additional 15 mice from this group were harvested and examined using WGIHC.
The number of neurons expressing lytic viral proteins of diverse kinetic classes
was quantified in one ganglion from each pair and VP16 expression was quantified
in the second ganglion from each pair. The number of neurons exiting latency,
whether detected by the anti-HSV antibody (31 neurons/15 ganglion) or the
antibody specific for VP16 (33 neurons/15 ganglion) was not different and
similar to the numbers of neurons exiting latency previously reported [4],[24],[25],[55],[56],[66],[69].
As observed in 17tBTK- infected TG, blockade of viral DNA replication did not
alter the expression of VP16.
VP16 is an essential multifunctional protein. However, mutations which impair the
transactivation function of VP16 have been generated and characterized in vitro
and in vivo. Mutant in1814 contains a 12 bp insertion that disrupts a domain
required for the VP16 induced complex formation and thus the transactivation
function of the protein [40],[70]. The carboxy-terminal acidic activation
domain has been deleted in two mutants, V422 [41] and RP5 [30],
both built in HSV-1 strain KOS. While these three mutants are phenotypically
similar in vitro, important differences have been reported in their in vivo
phenotypes. Despite the impaired replication reported for both in1814 and RP5 in
mouse eyes and TG, in1814 established latent infections efficiently and
reactivated in explant assays [16],[71],[72]. RP5 failed to accomplish either of these
outcomes [30]. The in vivo phenotypic differences between
HSV strains 17syn+ and KOS is a confounding issue [69],[73].
Therefore, mutant 17VP16Δ422 was constructed as detailed in methods. We utilized mutants in1814 and
17VP16Δ422 to evaluate the role of VP16 transactivation on reactivation
in vivo as outlined in Figure
3A.
Groups of male Swiss Webster mice were inoculated as described in methods. Viral replication was evaluated on
days 2,4,6,8 and 10 pi in tissues harvested from 3 mice from each inoculation
group. On day 4 pi, 3–4 logs fewer pfu were detected in the eyes and
TG of in1814 and 17VP16Δ422 infected mice compared to those infected
with in8141R, 17VP16Δ422R, and 17syn+ (Figure 3C and 3D, and not shown). These
results are in general agreement with previous reports [16],[30].
Interpretation of this result is complicated by the fact that at low moi (such as
a plaque assay), mutants lacking the transactivation function of VP16 enter the
lytic cycle inefficiently, leading to an underestimate the amount of virus
present [40],[41],[74].
Several strategies have been utilized to overcome this problem, including VP16
expressing cell lines, and superinfection with a replication impaired virus
[16],[30]. The addition of
the cell differentiating agent, hexamethylene bisacetamide (HMBA), to cell
cultures increases the plaquing efficiency of in1814 [74]. As shown in Figure 3C and 3D, the addition
of HMBA to the culture medium revealed the presence of 100 and 500 fold more
virus in in1814 and 17VP16Δ422 eye homogenates (day 2 pi), respectively.
As anticipated, this compound had little effect on the plaquing efficiency of
the parental strain, 17syn+ (1.8 fold increase in virus detected in
homogenates from 17syn+ infected eyes). Differences between the two
VP16 mutants, in1814 and 17VP16Δ422, to replicate in the TG were
dramatic. Plaque assays performed in the presence of HMBA revealed that in1814
did replicate within the TG, although maximum titers are 17 fold lower than
those achieved by wild type virus or 1814R (Figure 3D and not shown). In contrast, even
in the presence of HMBA, infectious virus was not detected in 17VP16Δ422
infected TG, although ∼200 pfu were detected in the TG on day 4 pi when
input titers of this mutant were increased (not shown). The importance of viral
replication within the TG for achieving maximum numbers of latently infected
neurons has been demonstrated [66]. That in1814 actually does replicate within
TG is consistent with its ability to efficiently establish latent infections as
well as reports that this mutant may retain some residual transactivating
function [75].
In preliminary studies, 17VP16Δ422 was determined to establish latent
infections, but at very low levels compared to 17syn+ (not shown). Thus
it becomes impractical to study in vivo reactivation with this mutant because
the efficiency of reactivation in vivo following HS is directly correlated with
the number of latently infected neurons in the ganglion
(r2 = 0.99) [56],[76].
Likewise quantification of the number of latently infected neurons in in1814
infected TG compared to the parental strain (17syn+) and rescue (1814R)
is critical for interpreting the outcome of experiments to quantify viral
reactivation. The number of latently infected neurons and the number of viral
genomes within individual infected neurons in TG from 3 mice from each group was
quantified using a single neuron PCR assay termed CXA [25],[28][66][77]. In this assay,
ganglia stabilized by fixation are enzymatically dissociated and individual
neurons from enriched neuronal fractions are harvested and analyzed by QPCR,
providing information on both the frequency of latently infected neurons and the
number of viral genome copies in the individual neurons analyzed. As anticipated
from the results of preliminary experiments, similar numbers of latently
infected neurons were observed in in1814, 1814R, and 17syn+ infected
ganglia, 28%, 25% and 26%, respectively (Figure 4A). The number of
viral genomes detected within individual latently infected neurons is shown in
the scattergram (Figure 4B).
No significant difference among the viral genome copy number profiles was
observed (mean copy number = 56.5, 51.8 and
44.8, respectively p = 0.94; ANOVA).
At 40 days pi, mice from each group of infected mice were induced to reactivate
in vivo using HS [5],[24]. At 22 hrs post
treatment, mice were euthanized, ganglia were removed and homogenized and the
homogenates assayed for infectious virus in the presence of 3 mM HMBA. In
striking contrast to previous reports in which in1814 reactivated in explant
reactivation assays, infectious virus could not be detected in any ganglia
(0/20) from in 1814 infected mice induced to reactivate in vivo. However,
infectious virus was detected at 22 hr post treatment in 17/20 (85%;
p = 0.0002) and 16/20 (80%;
p = 0.0002) of the TG pairs from mice infected
with 1814R and 17Syn+ (Figure 5A).
HS has been utilized extensively and shown to reproducibly induce viral
reactivation which peaks at ∼22 hrs post treatment using several
laboratory strains [69] as well as 10 low passage clinical isolates
(unpublished). However, it was possible that reactivation of in1814 was delayed
compared to wild type virus. Infectious virus could not be detected in ganglia
of in1814 infected mice 48 hrs post HS (0/20). These data demonstrate that
in1814 did not reactivate to detectable levels in vivo in response to HS.
In order to confirm that ganglia from this group of in1814 infected mice would
produce infectious virus when axotomized and explanted as previously reported
[16],[17], the 10 TG from
5 mice latently infected with in1814 were either directly homogenized and
assayed for infectious virus in the presence of HMBA or explanted and cultured
for 5 days and then tested for infectious virus in the presence of HMBA (see
methods). No virus was detected in TG
homogenized directly upon removal but infectious virus was detected in
100% (5/5) of explanted TG, a finding similar to previous reports in
which neurons were axotomized [16],[17] (Figure 5B). Similarly, the
ability of ganglia from 17VP16Δ422 infected mice to reactivate in
explant was tested. The presence of eGFP in this virus was used to monitor exit
from latency and spread within the explanted ganglia over time. No GFP
expression was detected in ganglia (0/6) at the time of explant, however within
4 days, a single GFP positive neuron was detected in 1/8 ganglia and by day 6
post explant, virus had spread within this TG (Figure 5C and 5D). After 15 days in explant,
17VP16Δ422 exited latency in 4/6 ganglia (Figure 5B). Ganglia were homogenized and
plated in the presence of HMBA and infectious virus was recovered which was GFP
positive and confirmed by southern blot to have the expected genomic structure
(not shown).
Reactivation from latency is functionally defined by the detection of infectious
virus. To expand our understanding of the process of reactivation, we have
developed a strategy for quantifying at the single neuron level the number of
neurons which exit latency as evidenced by detectable lytic viral protein
expression [4],[78]. This method is
the first to allow us to partition the process of reactivation into stages, and
to begin the assignment of viral and host cell functions critical for either
entry into the lytic cycle or for progression to infectious virus production
[25],[28]. In addition, this
approach obviates the inherent problem of detection of reactivated virus when a
mutant with low plaquing efficiency (such as in1814) is employed. At 40 days pi,
additional mice from the groups detailed above were induced to reactivate in
vivo using HS. Latently infected control mice and treated mice (at 22 hrs post
treatment) were euthanized, the ganglia were removed and processed for the
detection of lytic viral proteins using WGIHC as detailed previously [4],[78]. This method can
reliably detect a single neuron exiting latency among the 10's of
thousands in the ganglion. In the uninduced groups of animals, lytic viral
protein expressing neurons were not observed (0/10, 0/8, and 0/9, in1814, 1814R
and 17syn+ infected ganglia, respectively). This was not unexpected as
we have previously shown that the level of “spontaneous”
reactivation of strain 17syn+ in the latently infected Swiss Webster
mouse TG is very low [4],[78]. Consistent with
the detection of infectious virus above, ganglia from mice latently infected
with either 1814R or 17syn+ contained neurons expressing lytic viral
proteins at 22 hrs post induction, a total of 60 and 55 neurons, respectively in
the ganglia examined (Figure 6A and
6B). In contrast, no lytic viral protein expressing neurons were
detected in in1814 ganglia post induction (0/40). These findings indicate that
in vivo, the VP16 transactivating function is required for HSV to exit the
latent state and produce detectable viral proteins.
If VP16 functions at the earliest stages of the initiation of reactivation, its
role is likely to be the coordinated activation of immediate early genes through
the TAATGARAT promoter elements. We reported previously that mutant
ΔTfi, which contains a 350 base pair deletion in the ICP0 promoter
including the TAATGARAT element, reactivates with wild type frequency and
kinetics from ganglia axotomized and explanted [25],[29],
but exhibits severely impaired reactivation in vivo [25]. To test the
importance of the TAATGARAT to this phenotype, a mutant,
ΔTfi+TAATGARAT, in which the TAATGARAT regulatory element was
added back in its proper context to the ICP0 promoter of ΔTfi, was
generated as detailed in methods and tested for its ability to reactivate in
vivo.
Mice were inoculated with 4×106 pfu of ΔTfiR or
ΔTfi+TAATGARAT and at 40 days pi groups of latently infected
mice were utilized to quantify the number of latent infections and to determine
the in vivo reactivation frequency. There was no difference in either the number
of neurons latently infected or the viral genome copy number profile in TG
latently infected with ΔTfiR or ΔTfi+TAATGARAT (not
shown). Infectious virus was not detected in latently infected uninduced TG from
either ΔTfiR or ΔTfi+TAATGARAT infected mice, 0/5 and
0/5, respectively. However at 22 h post HS, infectious virus was detected in
41% (7/17) of TG from ΔTfiR and 44% (8/18) of TG
from ΔTfi+TAATGARAT infected mice. Taken together, this finding
and the clear requirement for VP16 transactivating function for the exit from
latency in vivo, support the hypothesis that as during acute lytic infection,
VP16 operates through the TAATGARAT element to regulate immediate early gene
expression during reactivation.
The preceding experiments suggest an amplification feedback loop between VP16 and
the IE gene products is required to exit latency. If this is true, expression of
VP16 must occur very early, requiring that the VP16 promoter be activated de
novo in contrast to the standard cascade of viral functions facilitating the
activation of the viral leaky late promoters. With wild type HSV, once
reactivation is initiated, the production of the viral IE transactivator
proteins induces transcription from all viral promoters in a cascade fashion.
That in1814 fails to produce detectable viral proteins following HS of latently
infected ganglia affords a unique opportunity to directly examine VP16 promoter
function in the absence of other viral proteins. Thus activation of the VP16
promoter following a reactivation stimulus in neurons latently infected with
in1814 would provide strong support that transcription of the VP16 gene can be
upregulated in the absence of other viral proteins during the earliest stages of
exit from latency. To test this hypothesis mutant in1814VP16pLZ was generated as
detailed in methods. The in vivo phenotypes of this mutant (replication,
establishment of latency and reactivation) were not different than in1814 (not
shown). Mice were infected with in1814VP16pLZ and maintained for 20 days pi.
Groups of mice were subjected to HS and at 22 hr post HS ganglia were removed
and processed to detect b-gal activity. No neurons were positive in 20 TG from
infected untreated animals. However, 5/20 of the ganglia from treated mice
contained one or more neurons positive for b-gal post HS, demonstrating that the
VP16 promoter had been activated in these neurons. In additional studies it was
observed that activation of this promoter post HS becomes increasingly rare as
time progresses which parallels the ability to detect VP16 protein (not
shown).
Successful completion of the complex in vivo life cycle of both herpes simplex virus
type 1 and type 2 requires activation of the viral lytic cycle from the latent viral
genome. Since HSV latency is characterized by the absence of all detectable lytic
viral proteins, the lytic cycle must start de novo, i.e. without the viral proteins
normally carried into a cell. These proteins, among other functions, facilitate
activation of transcription from the viral genome. Studies performed nearly two
decades ago led to the formative conclusion that VP16 does not
coordinate the exit from latency [15],[16],[72]. Using new
approaches we have revisited this important issue and now demonstrate that the
transactivating function of VP16 is indeed requisite for HSV reactivation in vivo.
The functional requirement for VP16 is very early in the transition from the latent
to the lytic cycle, as in its absence the latent viral genome cannot advance to the
production of lytic viral proteins. In contrast, elimination of “immediate
early” and “early” viral functions previously
considered critical for the initiation of reactivation has no measurable effect on
lytic viral protein expression at this early stage in reactivation 25,28. We show that replacing the VP16 promoter
with another leaky late gene promoter exposes unique regulatory properties of the
VP16 promoter in the context of infection in neurons in vivo. The VP16 promoter is
responsive to reactivation stimuli in the absence of other viral functions and
robust expression of VP16 from the latent viral genome is observed in neurons in the
absence of viral functions normally required for efficient late gene expression.
These data argue that the expression of VP16 during reactivation diverges from the
established late expression pattern of this gene. Our findings lead to the
hypothesis that in the context of the latently infected neuron, the expression of
VP16 is a critical initiating event, coordinating the activation of the viral IE
genes which results in productive entry into the viral lytic cycle.
The differing outcomes and thus conclusions regarding the role of VP16 in
reactivation stem primarily from the models of reactivation utilized. In previous
studies utilizing in1814 in the mouse ocular or footpad models reactivation was
evaluated ex vivo by removing ganglia from the latently infected animal and
culturing the tissue (explanting) for >21 days [16],[17]. Recent studies
demonstrate that axotomy and explant result in rapid and progressive changes in the
gene expression and physiological states of neurons and other cells not observed in
ganglia following an in vivo stress resulting in reactivation [24]. These changes can
obviate the need for and obscure the roles of viral genes important for the
reactivation process [25]. O'Hare suggested that VP16 might indeed
play a pivotal role during the early stages of reactivation, proposing that
explantation of ganglia into culture might overcome the requirement for VP16 [26], thus
explaining reports that in1814 reactivated ex vivo [15]–[17].
Our results confirm this hypothesis.
Quantitative analyses of both latency and reactivation at the single neuron level
have increased understanding of the relationships between viral replication, the
number of latently infected neurons, and the probability of reactivation [4],[25],[27],[28],[56],[69],[77],[79].Viral
replication within the trigeminal ganglia (TG) is required for maximizing the number
of latent infections [66] and mutants that replicate poorly in the TG
generally establish latency very inefficiently [25],[28]. In previous studies
mutant in1814 seemed to be an exception as no replication of in1814 was detected in
sensory ganglia [16],[72] and yet latency was
established very efficiently as determined by the detection of the latency
associated transcripts [16],[72]. However, with the
addition of HMBA to the plaque assay, it is clear that in1814 does replicate in the
TG. The mutant 17VP16Δ422, however, is severely restricted in TG and
establishes very low levels of latency as would be predicted by its limited
replication. It seems likely that mutated VP16 protein produced by in1814 still
retains some ability to transactivate IE genes, perhaps through elements other than
the TAATGARAT. VP16 has been shown to transactivate through the GCGGAA element in IE
gene promoters and this activity is independent of transactivation through the
TAATGARAT element [80],[81]. Despite the low
levels of latency established by 17VP16Δ422 this mutant did reactivate in
67% of the ganglia placed into explant culture within 15 days, a finding
that further emphasizes the differences inherent in the in vivo and ex vivo
reactivation models.
In vivo reactivation is a tightly regulated event in that following induction the
lytic cycle is engaged in ∼0.05% of neurons harboring the latent
genome [4]. Viral infection is confined to the individual neurons
undergoing reactivation (usually 1–5 neurons per TG) with no spread of
infection within the ganglion and consistent with this, it is characterized by low
levels of infectious virus in the ganglion [3],[5],[82]. Reactivation of HSV in
vivo has been characterized most extensively following hyperthermic stress, however
other induction triggers, including ultraviolet light irradiation [3],[6] show a
similar outcome. In contrast, in explanted ganglia virus infection spreads
unchecked, accounting for the extremely high titers of virus recovered [24]. The
limited production of virus in vivo raises the issue of sensitivity and the
possibility that the differing reactivation outcomes observed merely reflect
differences in detection of the infectious virus produced in the ganglia, a
consequence of reduced plaquing efficiency of in1814 compared to 1814R or
17syn+ [40]. This question could not be resolved without a
different approach for detecting the exit from latency, one that did not depend on
the detection of infectious virus. The need for such an assay has long been
recognized [31],[83],[84] and an in situ immunohistochemical method in
whole ganglia that permits the detection of lytic phase viral proteins in the very
rare neurons that exit latency was employed for this purpose [4],[24],[25],[28],[55],[78]. This assay makes it
possible to evaluate the ability of viruses that are blocked or impaired at later
steps in the replicative cycle, to exit latency. This assay revealed that the
failure to detect infectious virus in the ganglia latently infected with in1814
following induction was not merely an issue of sensitivity but an actual failure of
latent in1814 genomes to exit latency. The implications of this observation are
profound in that our attention is directed toward the regulation of VP16 from the
latent genome as a critical interface between neuronal responses to stress and entry
into the lytic cycle.
An important question relating to the mechanism of reactivation is whether this
phenotype is unique to VP16 transactivating function or whether other viral
functions are required at this earliest stage in the reactivation process. Two
hypotheses framed thinking about how the lytic cycle was initiated from the latent
genome without VP16 function. The first suggested that ICP0 served this function
[18]–[21] and the second proposed
that limited viral DNA replication is required for and precedes efficient expression
of IE genes [22],[23]. Using the same approaches utilized for the
analysis of in1814, the role of these viral functions in reactivation were examined.
In the absence of ICP0 function, initiation of reactivation as evidenced by lytic
viral protein expression was not measurably different from that in the rescue or
parental strain. Thus in contrast to VP16 transactivating function, ICP0 is not
essential for these very early events. A similar result was observed when viral DNA
replication were blocked either pharmacologically or genetically. In both cases,
viral proteins but not infectious virus was detected, confirming the roles of these
viral functions for the progression to infectious virus production [25],[28]. These
studies emphasize that VP16 appears to play a unique role in the earliest stages of
reactivation to coordinate IE gene expression and thereby entry into the lytic
cycle. Further evidence that VP16 functions early in the process of reactivation to
upregulate IE genes through the TAATGARAT motif comes from our finding that adding
back the TAATGARAT sequences to the ICP0 promoter in mutant ΔTfi restored
the ability of this mutant to reactivate in vivo. This is in keeping with a
requirement for ICP0 for progression to infectious virus production and the
TAATGARAT element for proper expression of ICP0 during reactivation in vivo.
The extreme rarity of viral reactivation at the neuronal level might be explained if
stress induced changes occur only in very rare neurons, or if only a few latently
infected neurons contain viral genomes capable of reactivation, but neither is the
case. Stress does induce global changes in the chromatin structure of latent viral
genomes [85]–[90]. To be detected such
changes must occur on the majority of latent viral genomes, while only a very few
viral genomes exit the latent state. Thus, the alterations in chromatin measured may
or may not be necessary but are not sufficient to precipitate viral
reactivation. Many latently infected neurons are capable of reactivation as
evidenced by semi-quantitative assays performed on dissociated latently infected
ganglia [91]. Likewise, over a period of a few days, 100s of
neurons produce viral proteins (exit the latent state) in TG axotomized and
explanted into culture in the presence of the antiviral drug acyclovir (which
prevents virus replication and spread within the TG) [24],[28]. Further, we have
determined that viral reactivation can be induced repeatedly (30 times over 10
months) in the same mice in vivo without reduction in the frequency of reactivation
(unpublished).
Reactivation from latency has been traditionally thought to be the result of changes
in host neuronal transcriptional regulators induced by systemic stress which then
directly stimulate the activation of one or more viral promoters. However, it is
difficult to reconcile the extreme rarity of these reactivation events with this
simple genetic model. For this model to be correct the stress induced signals that
initiate reactivation from latency must only occur in very rare neurons at any given
time, and since neurons do not survive viral reactivation [3],[5],[78], these same rare
changes would have to occur in a new rare subset of neurons as each virus
reactivation event occurs through time. The extremely low frequency of reactivation
suggests a model of stochastic derepression of VP16 in the presence of positive
transcription factors [92]. This type of model would predict that the
transcription and/or translation of viral genes is extremely repressed and that
viral reactivation is essentially an extremely rare and seemingly random event
precipitated when the amount of VP16 protein present in a given neuron reaches a
level adequate to initiate the cascade of lytic viral gene expression.
The hypothesis that VP16 functions in conjunction with host cell proteins as a
regulatory switch, promoting the lytic cycle when VP16 is present and latency in its
absence has been proposed by many investigators [13],[15],[26],[32],[83],[93]. Sears and Roizman first
proposed that the VP16 protein in the tegument may be inefficiently transported the
distance through the axon to the cell body, thereby promoting latency [15].
Although their early attempts to support this hypothesis experimentally were not
successful, there is support for the notion that the VP16 equivalent in pseudorabies
virus is dissociated from the nucleocapsid prior to transport to the neuronal cell
body [43].
Importantly, our results imply that VP16 generated and packaged into virions during
surface replication is not sufficient to efficiently activate IE genes in TG
neurons.
We propose that neuron driven regulation of VP16 orchestrates its de novo synthesis
which is central to coordinated entry into the lytic cycle in neurons and the
balance between the lytic and latent viral programs (Figure 7). During acute infection, virus
replication at the body surface feeds virions into the axons of innervating sensory
neurons (Figure 7A). Data
support the notion that VP16 is not transported efficiently to the neuron nucleus
and in its absence the latent transcriptional pathway is engaged. De novo expression
of VP16 is then required for the virus to enter the lytic pathway in these neurons
and VP16 is only expressed when repressors are overcome. Repressors that operate
during the acute stage of infection are likely to include riboregulators encoded by
the latency associated transcript locus (LAT) [94]–[96].
Expression of the LAT locus is required for ∼65% of the latent
infections established [27],[79], and in the absence
of LAT transcription, half of the neurons destined to be latently infected enter the
lytic pathway and die [27]. If sufficient VP16 is expressed to coordinate IE
gene expression, a positive feedback loop overcomes repression, the neurons becomes
lytically infected and the virus produced spreads both within the ganglion as well
as back down to the surface. Efficient virus replication at the surface and within
TG is required to maximize the number of latent infections established [66].
Acute virus replication ends by about 10 days pi and during the period between 10
and 40 days pi latency becomes consolidated (Figure 7B), perhaps through chromatinization of
the viral genome [85], [88]–[90],[97],[98], the recently proposed
immune mediated non-lethal inhibition of virus production once the exit from latency
has been initiated in neurons [99], and/or the build up of riboregulators [94]–[96],[100],[101], and fewer latently
infected neurons respond to stresses that induce viral reactivation [4]. By 40
day pi about 0.05% of latently infected neurons show evidence of viral
reactivation following HS whereas latency is maintained in the other
99.95% [4],[56],[69]. We hypothesize that stress induces the de novo
production of VP16, which only rarely reaches levels sufficient to coordinate
activation of the IE genes that overcome repressive factors and initiate the lytic
transcription program.
Stocks of HSV-1 strain 17syn+ and the mutants employed in this study
were generated in rabbit skin cell (RSC) monolayers and the viral titers were
determined by serial-dilution plaque assay [5],[102].
The wild type HSV-1 strain 17syn+ was originally obtained from John H.
Subak-Sharpe at the MRC Virology Unit in Glasgow, Scotland. The generation and
characterization of the VP16 transactivation deficient mutant in1814 and its
genomically restored counterpart, 1814R, have been described [16],[40]. Where indicated, the plaquing efficiency of
in1814 was enhanced by the addition of 3 mM hexamethylene bisacetamide (HMBA,
Sigma) as described [74],[103]. The mutant
ΔTfi and its genomically restored counterpart, ΔTfiR were a kind
gift of David Leib, Washington University, and have been described in detail
[29],[104].
All restriction enzyme sites and base pair numbering are referred to as the
corresponding positions in the published HSV-1 sequence of strain
17syn+ [105],[106] as present in
Genbank (NID g1944536).
A 649 bp DNA molecule was synthesized (Blue Heron) in which the VP16 promoter and
5′UTR (bp 105,435 to 105,107) was converted to the promoter and
5′UTR of VP5 (bp 40,659 to 40,812), flanked by appropriate sequences
and recombined into strain 17Syn+ The promoter structures of 5
independently derived mutants were confirmed by Southern blot analysis and DNA
sequencing (not shown). Wild type UL49-UL48 sequences were recombined with
VP5p/VP16-1 to generate the genomically restored variant VP5p/VP16-1R as
described [27],[79].
Mutant 17VP16Δ422 was generated in strain 17syn+ to be similar
to mutant V422, in which the carboxy-terminal acidic transactivation domain of
VP16 was deleted in strain KOS [107]. The eGFP gene driven by the b-actin
promoter was inserted at the SacI site at bp 103,808 in the VP16 ORF in the
orientation opposite to VP16, truncating the protein at amino acid 422 and
facilitating selection of the mutants. 3 mM HMBA was added to the culture medium
to increase the plaquing efficiency of the VP16 mutants [103]. Following low
MOI infection of RSC, mutants displayed a reduced capacity to replicate and
plaque which was increased 40 fold in the presence of HMBA. Wild type UL49-UL48
sequences were recombined with 17VP16Δ422 to generate the genomically
restored variant 17VP16Δ422R as described [27],[79].
Truncation of the ICP0 promoter at −145 just prior to the TAATGARAT
element resulted in an ICP0 promoter that drives expression efficiently in cells
and tissues in vivo with the exception of TG neurons, where this promoter fails
to function [55]. Using this information, a construct was
generated that restored sequences to −172 in the ΔTfi ICP0
promoter including 27 additional bases (CGTGCATGCTAATGATATTCTTTGGGGG)
that contain the functional TAATGARAT of ICPO (underlined). This construct was
recombined into both ICP0 promoters in the mutant ΔTfi to generate the
ICP0 promoter mutants termed ΔTfi+TAATGARAT. Independently
derived mutants were generated and characterized and all replicated as well as
wild type in RSC and in mice in vivo (not shown).
Mutants 17VP16pLZ and in1814VP16pLZ express the E. Coli beta-galactosidase gene
(b-gal) from a 423 bp promoter/5′UTR fragment of the VP16 gene
(105,108-105,534 bp) inserted into the intergenic region between glycoprotein J
(gJ) and gD in strain 17Syn+, or the VP16 transactivation mutant in1814
respectively. A single base mutation (G>A) was introduced at 138,045 on
the viral genome to generate an EcoRV site in the intergenic region between gJ
and gD. The promoter/reporter cassette (terminated by bi-directional SV40
polyadenylation signals) was cloned in the orientation opposite that of the
viral gJ and recombined in to the genome [55]. Six
independently derived mutants replicated as well as wild type in RSC and in mice
in vivo (not shown). Wild type levels of gJ and gD mRNA of the expected sizes
were produced by the mutants (not shown).
All procedures involving animals were approved by the Children's
Hospital or University of Cincinnati Institutional Animal Care and Use Committee
and were in compliance with the Guide for the Care and Use of Laboratory
Animals. Animals were housed in American Association for Laboratory
Animal Care-approved quarters. Male, outbred, Swiss Webster mice
(22–25 grams in weight, Harlan Laboratories) were used throughout
these studies. Prior to inoculation, mice were anesthetized by intraperitoneal
injection of sodium pentobarbital (50 mg/kg of body weight). A 10 ul drop
containing the amount of virus as detailed in the text was placed onto each
scarified corneal surface. In some experiments the inoculum titer was altered to
achieve uniform levels of latent infections as previously detailed [56],[66]. In preliminary
experiments we determined that inoculation of mice with
2×105 pfu of strain 17syn+ or in1814R and
8×105 pfu of in1814 resulted in similar levels of
latency as shown in Figure
4B. In these preliminary studies mice infected with in1814 at
2×105 pfu were examined. In these mice as in those
receiving the higher imput titer no exit from latency was detected.
Mice infected as above were euthanized at the indicated times post infection and
tissues from three mice from each inoculation group were individually assayed
for virus as previously detailed [49].
Additional mice from the groups infected as above were maintained for at least 40
days pi. Enriched neuron populations were obtained and individual neurons were
assayed for the presence of the latent viral genome as described [77],[79].
Isolation and quantification of total DNA from TG and quantification of total
viral genomes by real time PCR was performed as detailed previously [28].
Latent HSV was induced to reactivate in the ganglia of mice in vivo using
hyperthermic stress (HS) and at 22 hours post induction TG were assayed for
infectious virus as detailed previously [5].
Latently infected ganglia were aseptically removed and placed into Minimum
Essential Media (MEM) supplemented with 5% newborn calf serum and
incubated at 37°C in a 5% CO2 incubator. At the
indicated times post explant, ganglia were homogenized and assayed for
infectious virus as for reactivation in vivo [24].
Colocalization of b-gal activity and HSV lytic viral proteins was carried out by
first histochemically staining whole TG by incubation in x-gal (Sigma) followed
by paraffin embedding of TG, sectioning and immunohistochemical detection of
viral proteins. Where indicated, HSV proteins were also detected in whole
ganglia using whole ganglia immunohistochemistry (WGIHC). Primary antibodies
utilized include rabbit anti-HSV (AXL237, Accurate), rabbit anti-VP16 antibody
(clonetech), and secondary antibody utilized was HRP labeled goat anti-rabbit
(Vector). These methods and the dilutions and characterizations of antibodies
utilized have been detailed extensively in previous reports [4],[55].
|
10.1371/journal.ppat.1002331 | The SARS-Coronavirus-Host Interactome: Identification of Cyclophilins as Target for Pan-Coronavirus Inhibitors | Coronaviruses (CoVs) are important human and animal pathogens that induce fatal respiratory, gastrointestinal and neurological disease. The outbreak of the severe acute respiratory syndrome (SARS) in 2002/2003 has demonstrated human vulnerability to (Coronavirus) CoV epidemics. Neither vaccines nor therapeutics are available against human and animal CoVs. Knowledge of host cell proteins that take part in pivotal virus-host interactions could define broad-spectrum antiviral targets. In this study, we used a systems biology approach employing a genome-wide yeast-two hybrid interaction screen to identify immunopilins (PPIA, PPIB, PPIH, PPIG, FKBP1A, FKBP1B) as interaction partners of the CoV non-structural protein 1 (Nsp1). These molecules modulate the Calcineurin/NFAT pathway that plays an important role in immune cell activation. Overexpression of NSP1 and infection with live SARS-CoV strongly increased signalling through the Calcineurin/NFAT pathway and enhanced the induction of interleukin 2, compatible with late-stage immunopathogenicity and long-term cytokine dysregulation as observed in severe SARS cases. Conversely, inhibition of cyclophilins by cyclosporine A (CspA) blocked the replication of CoVs of all genera, including SARS-CoV, human CoV-229E and -NL-63, feline CoV, as well as avian infectious bronchitis virus. Non-immunosuppressive derivatives of CspA might serve as broad-range CoV inhibitors applicable against emerging CoVs as well as ubiquitous pathogens of humans and livestock.
| Broad-range anti-infective drugs are well known against bacteria, fungi, and parasites. These pathogens maintain their own metabolism distinctive from that of the host. Broad-range drugs can be obtained by targeting elements that several of these organisms have in common. In contrast, target overlap between different viruses is minimal. The replication of viruses is highly interweaved with the metabolism of the host cell. A high potential in the development of antivirals with broad activity might therefore reside in the identification of host factors elemental to virus replication. In this work we followed a systems biology approach, screening for interactions between virus and host proteins by employing an automated yeast-two-hybrid setup. Upon binding of a viral protein to cyclophilins the screen led to the identification of the Calcineurin/NFAT pathway possibly being involved in the pathogenesis of SARS-Coronavirus. Secondly, cyclophilins were suggested to play an elemental role in virus replication since cyclosporin A inhibited replication of all Coronavirus prototype members tested. This large range of viruses includes common cold viruses, the SARS agent, as well as a range of animal viruses. For the first time this work shows that an undirected, systems-biology approach could identify a host-encoded, broad-range antiviral target.
| Five distinct CoVs (SARS-CoV, hCoV-NL63, hCoV-HKU-1, hCoV-OC43, hCoV-229E) cause respiratory tract illness in humans, ranging from mild common cold to deadly virus-associated pneumonia [1]. At least seven different animal CoVs cause economically significant epizootics in livestock, and deadly disease in companion animals [1]. The agent of SARS was a novel CoV introduced into the human population from an animal reservoir, resulting in a highly lethal epidemic in 2002/2003 [1], [2]. A tremendous diversity of CoVs exists in complex mammalian and avian reservoirs [1], [3], [4]. Host switching is a common feature in CoV evolution, and novel epidemic CoV can emerge anytime [1], [3], [5]. Because the large diversity of CoVs complicates the design of vaccines, the identification of broad-range anti-CoV drug targets might indicate alternative approaches against CoV epidemics [1]. Broad range anti-CoV drugs would also be desirable to treat severe infections caused by known human and animal CoVs.
The SARS-CoV genome is predicted to encode 14 functional open reading frames, leading to the expression of up to 29 structural and non-structural protein products [1]. The functions of many of these proteins are poorly understood or unknown. To study the interplay of viral proteins with the host cell and to identify new targets involved in viral replication we have performed a genome-wide analysis of protein - protein interactions between the SARS-CoV and human proteins via a High-Throughput Yeast Two Hybrid Screen (HTY2H) [6], [7]. Within this framework we identified redundant interactions between SARS-CoV non-structural protein Nsp1 and a group of host proteins with peptidyl-prolyl cis-trans-isomerase activity, including the cyclophilins/immunophilins PPIA, PPIG, PPIH and FKBP1A, FKBP1B. These modulate the Calcineurin/NFAT pathway that plays an important role in immune cell activation [8], [9]. The NFAT family of transcription factors encodes four calcium-regulated proteins of which three (NFAT1, -2, -3) are expressed in a variety of cell types including T-cells, B-cells, mast cells, natural killer cells and eosinophils [8], [9]. NFAT activation regulates pivotal immune processes like apoptosis, anergy, and T-cell development. An essential activation step for NFAT is its dephosphorylation by the phospatase calcineurin A (CnA), resulting in the translocation of NFAT into the nucleus. Cyclosporin A (CspA) forms complexes with cyclophilins that bind to CnA, preventing its activity. This effect is used in transplant patients to prevent organ rejection by suppression of the immune system. Here we show that SARS-CoV nonstructural protein Nsp1, as well as full replicating SARS-CoV, enhance the CnA/NFAT pathway and induce NFAT-responsive promoters. Because interactions with upstream elements of the pathway were redundantly identified in a hypothesis-free virus-host interaction screen, the pathway is likely to play a significant role for virus replication. Indeed, an extensive panel of CoVs covering all three relevant virus genera was strongly inhibited by manipulation of cyclophilins using CspA.
All SARS-CoV ORFs and a number of subfragments lacking transmembrane regions were cloned into eukaryotic expression vectors. Using HTY2H, these were screened against a cDNA library of very high complexity (1.4×107) derived from human brain, as well as an additional library of individually-cloned full-length ORFs encoding 5000 human proteins. Inserts from positive yeast clones were sequenced and compared against GenBank. BLAST searches on 2287 DNA sequences yielded 942 different human gene hits. These were divided into four confidence categories: category A (highly confident interaction partners found more than once in one or several screens), category B (single hits), category C (sticky preys interacting with several to many bait proteins) and category D (3′-UTR cDNA regions or inserts in reverse orientation coding for unnatural peptides). We found 132, 383, 245, and 282 hits in categories A – D, respectively. For validation, the cDNAs of 86 category A and category B interaction candidates were cloned in-frame with the Renilla reniformis luciferase and overexpressed in HEK 293 cells. SARS-CoV ORFs were cloned in-frame with N-terminal protein A domains and co-expressed in the same cells. Protein A-directed immunoprecipitates retained on IgG-coated magnetic beads were identified by measuring in-vitro Luciferase activity. About 48% of category A candidates and 36% of category B candidates were confirmed positive with a Z-score >1 (Figure 1, see Materials and Methods for definition), corresponding to previous observations [10]. A list of validated category A and B HTY2H interactor candidates is provided in Table S1.
For an overall estimate of plausibility, more than 5,000 Medline abstracts mentioning “SARS” or “Coronavirus” were screened using the text mining program syngrep, scanning for the mentioning of human protein designations and synonyms. Abstracts mentioning YTH or co-immunoprecipitation assays were specifically sought. Twenty-eight CoV-/host protein interactions were identified in the literature, as listed in Table S2. It was then determined how these literature hits overlapped with the lists of candidate interactors as identified by HTY2H screening in different confidence levels. Using a hypergeometric test, the fractions of overlap were compared to the fraction of literature hits in the list of search terms (31,941 human proteins used for text-mining). Abstracts were enriched for proteins identified as SARS-CoV interaction partners both in the high confidence and the complete data sets (Table S3 and Table S4). Figure 2 summarizes highly confident interactions identified in the overall screen and GO [11], [12] analysis. SARS-CoV proteins were found to preferentially target protein complex subunits (Table S5 and Table S6). Of 9 complexes which were targeted through ≥4 subunits, 4 complexes were found to be significantly enriched: The respiratory chain complex I (7 subunits targeted by SARS-CoV, p-value ∼0.036), the cytoplasmic ribosome (10 subunits targeted by SARS-CoV, p-value ∼0.036), in particular the 60S ribosomal subunit (7 subunits targeted by SARS-CoV, p-value ∼0.036) and the LCR-associated remodeling complex which is involved in DNA conformation modification (4 subunits targeted by SARS-CoV, p-value ∼ 0.039). Furthermore, the analysis of the centrality of SARS targets within the human interaction network (Figure S1) indicated that SARS-CoV proteins target both highly interactive proteins (hubs) as well as so-called bottleneck proteins which are central to many of the shortest paths in their networks [13] (Figure S2). Table S4 summarizes GO results for SARS-CoV nonstructural protein Nsp1, a protein yielding particularly interesting candidate interaction networks. Interactions between Nsp1 and several members of the class of immunophilins (PPIA, PPIG, PPIH, FK506-binding proteins FKBP1A and -B) and calcipressins (RCAN1 and -3) were selected for experimental confirmation.
The immunophilin proteins (cyclophilins and FK506-binding proteins) are all known to bind to CnA in combination with inhibitory molecules, and to influence the CnA/NFAT pathway that plays a major role in the establishment of T-cell immune response [14]. For a more detailed mapping of HTY2H hits, PPIA, PPIB, PPIG, PPIH, FKBP1A and RCAN3, and three versions of Nsp1 [Nsp1(aa 1–180), Nsp1(aa 1–93) and Nsp1(aa 119–180)] were cloned into LUMIER assay vectors to yield luciferase and protein A fusion proteins, respectively. Although PPIB was not identified as an interactor by Y2H it was included in the experiment as it is known to bind to the HIV-1 gag protein [15] and to the HCV NSB5 protein [16]. All tested proteins interacted with Nsp1(aa 1–93), suggesting redundant interactions of SARS-CoV with the CnA/NFAT pathway via the N-terminal part of Nsp1 (Figure 3). To examine the functional consequences of Nsp1 expression on NFAT activity, NFAT and CnA cDNAs were overexpressed in HEK 293 cells. Parallel experiments in Jurkat cells were done without overexpression due to their constitutive activity of the CnA/NFAT pathway. The CnA/NFAT pathway was stimulated by addition of PMA (40 ng/ml) and ionomycin (2 µM) to the culture medium. In both cell lines treated this way, expression of Nsp1 did not induce NFAT activity directly, but increased significantly the stimulatory effect of PMA/ionomycin on NFAT activation (Figure 4A). The increase in NFAT activity could be blocked by CspA, an inhibitor of the NFAT pathway (Figure 4A and B). Coexpression of the calcipressin RCAN3 as shown in Figure 4B attenuated the overall stimulating effect on the NFAT activity. In contrast, coexpression of other CoV proteins or coexpression of PPIA, PPIH, FKBP1A did not impact NFAT activity (data not shown). Experiments up to this point employed overexpression of NFAT3. As different NFAT species are expressed depending on cell type [17], NFAT1 and NFAT2 were alternatively expressed and compared in the same assay. For both species essentially the same influence of Nsp1 on PMA/ionomycin-dependent stimulation was seen (Figure 4C and D). Altogether this suggested a broad effect of Nsp1 on NFAT activation that is mediated via the canonical NFAT activation pathway including CnA.
In order to determine the extent of PMA/ionomycin-dependent NFAT activation during virus infection, HEK 293 lp cells (lp = low passage) with a short passage history were infected with SARS-CoV at an MOI = 1. These cells had been previously demonstrated to support SARS-CoV replication, in contrast to common HEK 293 cells [18], [19]. Figure 5 shows that the CnA/NFAT pathway was induced in the context of SARS-CoV infection at considerable extent, and in a PMA/ionomycin-dependent way.
Viruses may interfere with cytokine induction, but on the other hand, may also induce cytokine genes directly. To examine whether the Nsp1-mediated, PMA/ionomycin-dependent activation of NFAT may cause specific induction of relevant cytokines, HEK293 cells were co-transfected with the plasmids described above, except that the NFAT reporter plasmid was replaced by luciferase reporter plasmids carrying the IL-2, IL-4 and IL-8 promoters, respectively (Figure 6). Expression of Nsp1 induced the IL-2 promoter significantly by a factor of about 2.5 (Figure 6A). This effect was inhibited by CspA and RCAN-3, suggesting dependence on the CnA/NFAT pathway. The IL-4 promoter activity was not significantly elevated by Nsp1 expression in the presence of PMA/ionomycin (Figure 6B). Its activity was decreased in the presence of CspA but not RCAN-3. The IL-8 promoter was induced by PMA/ionomycin alone, but significantly downregulated by a factor of about 1.8 in additional presence of Nsp1 (Figure 6C). Expression of RCAN-3 reduced IL-8 promoter activity levels to about half, while CspA inhibited the promoter completely. In Jurkat cells, which express endogenous NFAT3 and CnA, the Nsp1 protein did not induce the IL-2 promoter. The slight induction of IL-4 and the downregulation of the IL-8 promoter activities in presence of Nsp1 (about twofold) were similar to effects seen in HEK 293 cells. These results suggested that Nsp1 expression had the strongest influence on the IL-2 promoter.
Next to NFAT, transcription factors NFκB and Activating Protein 1 (AP-1) determine IL-2 regulation [20]. NFAT, AP-1 and NFκB binding sites are juxtaposed in the IL-2 promoter, and it has been shown that NFAT and AP-1 act in a cooperative manner on the promoter while NFkB has enhancing function [17]. Simultaneously, NFκB induces the IFN-beta gene by binding to the PRDII DNA element [21]. The latter is a more sensitive assay of NFκB nuclear translocation upon viral infection. To examine potential direct effects of Nsp1 on NFκB nuclear translocation and AP-1, HEK 293 and Jurkat cells were cotransfected with SARS-CoV Nsp1fl and p55A2luc containing repeated PRDII elements or pAP-1-luc containing the AP-1 binding site of the IL-2 promoter (Figure 7). Overexpression of Nsp1fl as well as treatment with PMA/ionomycin, respectively, caused small but significant luciferase increases in both cell lines. The combined expression of Nsp1fl with PMA/ionomycin treatment led to significant induction of PRDII by a factor of about 6 in both cell lines. The AP-1 promotor was only slightly upregulated in HEK293 and downregulated in Jurkat cells. This indicated a co-involvement of NFκB but not of AP-1 in the induction of IL-2 by Nsp1, suggesting dependence mainly on the NFAT pathway. In summary, SARS-CoV caused relevant and specific induction of IL-2 by activating the NFAT pathway via Nsp1.
CspA is a highly efficient antagonist of NFAT activation, interacting with cyclophilins. Due to the high specificity of Nsp1-dependent activation of NFAT and due to the high redundancy of SARS-CoV interactions with upstream elements of the CnA/NFAT pathway, we suspected an essential function for the virus. It was therefore investigated whether CspA might influence viral replication (Figure 8). Vero cells were inoculated with a low dose of SARS-CoV (MOI = 0.0001) and growth of virus replication was determined by real-time RT-PCR and plaque titration. In parallel cell cultures treated with the same concentrations of CspA, cell viability was measured with a highly sensitive assay based on ATP provision in metabolically active cells. A profound and dose-dependent inhibition of replication of SARS-CoV strain Frankfurt-1 in Vero E6 cells was seen in absence of cytopathic effects conferred by the compound. The 50% effective inhibitory concentration was 3.3 µM.
Because Nsp1 proteins of group I and SARS coronaviruses share structural and functional similarities [22], it was tested whether the inhibitory effect of CspA could be extended to other pathogenic CoVs. These included members of the genera Alphacoronavirus (human CoV-NL63 and -229E, Feline CoV serotypes I and II [strains Black and 791146], porcine transmissible gastroenteritis virus [TGEV]), Betacoronavirus (SARS-CoV isolates Frankfurt and Hongkong) and Gammacoronavirus (avian infectious bronchitis virus [IBV]). All tested CoVs were inhibited by CspA; replication of TGEV and IBV in the tested range (up to 25 µM) was diminished close to background by CspA. HCoV-NL63 and -229E and the two Feline CoV serotypes were completely inhibited, with 50% effective concentrations of 2.3 µM, 2.3 µM and 2.7 µM, respectively (Figure 8). Figure S3 shows reduction of virus replication in a log scale.
In order to determine the principal stage of the CoV replication cycle inhibited by CspA, a novel SARS-CoV replicon carrying a secreted Metridia luciferase reporter construct instead of the major structural proteins S, E, and M was used (Figure 9A). The replicon RNA together with an mRNA for the nucleocapsid protein was electroporated in BHK cells. Replicon activity in parallel reactions was controlled to be at the same level after 16 h of incubation (data not shown), and increasing amounts of CspA were added to cells after repeated washing. As shown in Figure 9B, accumulated luciferase activities in supernatants were decreased in a CspA dose-dependent manner after 24 h. Two different specific inhibitors of the CoV main protease, Cinanserin [23] and XP17 (R. H., own unpublished observations), inhibited replicon activity to a comparable extent as CspA, at comparable substance concentrations (Figure 9B). To control against any influence of the nucleocapsid protein that is co-electroporated for maximal replicon efficiency [24], [25], [26] and that is also contained in the replicon RNA, this protein was expressed from a eukaryotic expression vector in the same cells and an NFAT induction assay was conducted as described above. No N-dependent effect on the assay was seen (Figure 9C). These results suggest an action of CspA on genome replication and/or transcription, rather than other stages such as virus entry or egress.
Various genomic and proteomic methods have been utilized to identify protein-protein interactions in the context of viral replication [6], [27], [28], [29], [30], [31]. HTY2H is among the most direct approaches to identify interactions between members of viral ORFeomes and large host cDNA libraries. Major advantages of the method include its potential for high throughput testing and automation, as well as its high sensitivity. The latter facilitates investigation of proteins expressed at low levels and of those causing weak and transient interactions [32], [33]. Drawbacks include the inability to control and confirm expression of genes of interest, other than by positive selection of yeast expression clones containing nutritional markers. Moreover, some proteins need posttranslational modifications not provided by the yeast cell, in order to interact with binding partners. Also, since interactions of bait and prey proteins take place in the nucleus of the yeast cell, the assay is influenced by hydrophobic and transmembrane regions affecting the nuclear membrane. It is well known that a considerable fraction of interacting proteins in HTY2H represent false positive findings, making it absolutely necessary to validate interactions by independent eukaryotic assays. We have implemented a version of the Lumier assay that is amenable for screening in mammalian cells at a medium scale of parallelity [34]. Our modified version using a protein A tag instead of a Flag tag enables automated capture of precipitates on IgG Fc-coated magnetic beads. Throughput is mainly limited by the requirement to subclone Y2H plasmid inserts, as the assay does not involve any cell-based imaging or other readouts going beyond in-vitro Luciferase assays. In our analysis of 86 Y2H-positive interaction partners we achieved a positive confirmation rate of about 42% in category A and B interactors, which is in good agreement with a recent standardized comparison of five different interaction assays in which the LUMIER pull-down assay showed the highest sensitivity (36%) on a positive reference set of human proteins [10]. It has to be mentioned that all interaction assay systems carry intrinsic limitations. In the case of our modified Lumier method the Renilla and protein A tags are rather long as compared to His or HA tags. Therefore, true interactions might be missed, and it is possible that the interaction of PPIA with full-length nsp1 is sterically prevented by the length of these tags, as compared to the N-terminal fragment of Nsp1. But the chief attraction of our method is its applicability in high thoughput assays.
The range of interactors identified in this study defines an unprecedented resource for future investigations into pathogenetic mechanisms and antiviral applications against CoVs. In order to demonstrate that HTY2H can afford a direct identification of novel antiviral targets, we have chosen one promising group of interactors for further investigation in the present study. The interaction of Nsp1 with the cyclophilins PPIA, PPIB, PPIG, PPIH, the FK506-binding proteins FKBP1A/B, and the CnA (calcipressin) regulators RCAN1 and RCAN3 represented a highly redundant virus-host interaction involving critical elements of the same regulatory network immediately upstream of the CnA/NFAT pathway. Nsp1 is a virulence factor in-vivo whose action has been linked with early stages of the immune response, including antagonism against IFN signaling and inhibition of host protein synthesis [35], [36], [37], [38]. Our findings add an important new dimension to Nsp1′s role in pathogenicity, identifying this protein as a strong and specific activator of NFAT enhancing the induction of the IL-2 promoter. The increase of NFAT activation extended to all three major NFAT species, suggesting a potential for induction of broad and systemic cytokine dysregulation by affecting several types of immune cells. The pattern of cytokine dysregulation in severe SARS cases differed from the cytokine burst seen in other acute viral diseases in its delayed occurrence, manifesting beyond the second week of symptoms. Interestingly, it was noted upon clinical observations that late aggravation was correlated with severe clinical outcome, triggering parallel efforts in several centers to treat patients empirically with steroids [39], [40], [41]. Our results suggest an influence on fundamental triggers of immune cell activation, contributing an explanation for the cytokine dysregulation and immune-dependent pathogenesis observed in severe cases of SARS.
The detected interactions with immunophilins caught our attention because peptide inhibitors such as CspA and Tacrolimus (FK506) are available that bind to immunophilins and cause dramatic effects on the cellular phosphatase CnA, suppressing the CnA/NFAT immune-regulatory pathway [42]. Cyclophilins are essential cofactors for replication of HCV, HIV and of some parasites [43]. Incorporation of PPIA into HIV-1 particles via binding to gag is paralleled by its binding to the N protein of SARS-CoV by an educated guess finding using surface plasmon resonance biosensor technology [44]. Furthermore, in a recent proteomics study in which viral and cellular proteins incorporated into SARS-CoV virions were spectrometrically profiled PPIA was also found in purified virus particles [31]. We have indeed shown that the inhibition of cyclophilins by CspA exhibited strong and specific inhibitory effects on members of all three genera of CoVs. This broad antiviral action is supported by a recent characterization of Nsp1 structural conservation that extends beyond the limits of CoV genera [22]. Inhibition took place in the low micromolar range, indicating prospects for future investigation of similar (non-immunosuppressive) drugs as broad-range antivirals. Such drugs have already been employed against HCV, whose nonstructural NS5A and NS2 proteins interact with cyclophilins and whose replication can be inhibited by CspA and non-immunosuppressive derivatives thereof [45].
Our study is limited in that it does not clarify the biological functions of the interaction between Nsp1 and its cellular partners, nor does it confirm the involvement of Nsp1 in full virus context by knock out experiments in recombinant viruses. Brockway and Denison showed that deletion of residues in the amino-terminal half of nsp1 is not tolerated for a productive infection [46]. Thus it seems not possible to construct viruses with mutations in the N- terminal half of nsp1 which do not bind to cyclophilins anymore. It will thereofore be difficult to exactly delineate the role of nsp1 in virus-host cooperation. It should also be appreciated that CspA-inhibitable NFAT induction by SARS-CoV nsp1 may be independent of the replication inhibition by CspA. CspA specifically binds to cyclophilins and this complex binds to calcineurin phosphatase preventing the dephosphorylation of NFAT. Binding of nsp1 to cyclophilins and the induction of NFAT is obviously inhibited by Csp A. On an independent level, formation of the Cyclophilin-CspA complex might prevent those cyclophilin functions required for virus replication. Complex further studies involving multiple CoV systems will be required to delineate these functions.
Nevertheless, we have been able to shed more light on the principal stage of the virus replication cycle that is subject to CspA-dependent inhibition. Our experiments using a SARS-CoV replicon suggest strongly that the processes afforded by the replicative proteins rather than stages of virus entry and egress, are affected. Further experiments need to be done in the future in order to investigate on a mechanistic level the potential breadth of the identified antiviral effect. In particular, the large diversity of CoVs in animal reservoirs generates interest in studying how conserved this particular virus-host interaction might be between non-human CoVs and human cells, and whether this could be exploited as a truly broad-range antiviral target that covers epidemic and reservoir-borne viruses alike.
HEK293, HEK293lp (low passage), Vero E6, CaCo-2, HRT 18, Huh 7, FCWF- and St-cells were grown in Dulbecco’s modified Eagle medium (DMEM) containing 10% FBS, 1% L-glutamine, 1% penicillin/streptomycin and 1% non essential amino acids. C8166 SEAP cells were cultured in RPMI medium with 10% FBS and 1% penicillin/streptomycin. Jurkat cells were propagated in RPMI-medium containing 10% FBS, 1% L-glutamine and 1% penicillin/streptomycin.
Cloning of the SARS ORFeome into destination bait vector pGBKT7-DEST for Y2H screening was described previously [6]. A series of SARS-CoV (Frankfurt isolate) bait clones containing ORF fragments depleted of transmembrane regions were generated in addition (primers and cloning procedures available on request). Automated yeast two-hybrid screens were essentially done as described previously [7], [47], with the following modifications. Human cDNA libraries from human brain and fetal brain (Clontech) as well as a library of individually cloned full–length open reading frames from cDNAs of 5000 different genes were screened to a minimal coverage of 5 million clones per library. To mate yeast strains, exponentially growing cultures at an OD600nm of 1 were combined, pelleted by centrifugation, and resuspended in an equal volume of YPDA (Yeast Extract Peptone Dextrose Adenine) containing 20% PEG 6000. For the generation of a high-confidence dataset, interaction pairs were selected which were isolated at least twice, or where the bait interacted with two highly related preys, and which did not involve promiscuous preys.
For LUMIER assays, proteins were transiently expressed in HEK293 cells as N-terminal fusion proteins with the Staphylococcus aureus protein A tag or Renilla reniformis luciferase. 20 ng of each expression construct were transfected into 10,000 HEK293 cells using 0.05 µl of lipofectamine 2,000 (Invitrogen) in 96 well plates. After 40 hours, medium was removed and cells were lysed on ice in 10 µl of ice-cold lysis buffer (20 mM Tris pH 7.5, 250 mM NaCl, 1% TritonX-100, 10 mM EDTA, 10 mM DTT, Protease Inhibitor Cocktail [Roche # 1 836 170], Phosphatase Inhibitor Cocktail [Roche, # 4 906 837], Benzonase [Novagen #70746, 0,0125 units per µl final concentration]) containing sheep-anti-rabbit IgG-coated magnetic beads (Invitrogen, Dynabeads M280, 2 mg/ml final concentration). Lysates were incubated on ice for 15 minutes. 100 µl of washing buffer (PBS, 1 mM DTT) were added per well, and 10% of the diluted lysate was removed to determine the luciferase activity present in each sample before washing. The rest of the sample was washed 6 times in washing buffer in a Tecan Hydroflex plate washer. Luciferase activity was measured in the lysate as well as in washed beads. Negative controls were transfected with the plasmid expressing the luciferase fusion protein and a vector expressing a dimer of protein A.
For each sample, four values were measured: the luciferase present in 10% of the sample before washing, the luciferase activity present on the beads after washing, and the same values for the negative controls. Normalised interaction signals were calculated as follows: Log(bound)/log(input) – log(bound nc)/log(input nc). Z-scores were calculated by subtracting the mean and dividing by the standard deviation. The mean and standard deviation were calculated from large datasets of protein pairs which were not expected to interact, i.e. from negative reference sets. Normalised signal to noise ratios were calculated as follows: (bound/input)/(bound nc/input nc)[10].
For reporter gene expression HEK293 cells were transiently transfected with 2 µg DNA containing 460 ng of the respective expression plasmids encoding NFAT, calcineurin, SARS-CoV Nsp1 and reporter genes as indicated (six well plates). DNA was transfected using FuGENE HD reagent (Roche Applied Science). For reporter gene expression in Jurkat cells (1×106 cells) 1 µg reporter plasmid and 1 µg expression plasmid encoding SARS-CoV Nsp1 were transiently cotransfected using the Amaxa Cell Line Nucleofector Kit (Lonza). For reporter gene assays in the viral context HEK293lp cells (24-well plates) were transiently transfected with 500 ng reporter plasmid using Lipofectamine LTX (Invitrogen) according to the manufacturer’s instructions. 24 h post transfection cells were infected with SARS-CoV HK at an MOI = 1. 19 h after transfection cells were harvested and Promega’s dual luciferase assays were performed according to the manufacturer’s instructions. All results were normalized to a simultaneously transfected Renilla luciferase (pRL-null, Promega). Cells were stimulated by PMA (40 ng/ml) and Ionomycin (2 µM) in culture medium. The NFAT-pathway was inhibited by Cyclosporin A (50 ng/ml). Virus infected cells were lysed 17 h post infection with Promega passive lysis buffer (PLB) and 1% Igepal (Sigma-Aldrich). All experiments were repeated at least three times.
Cytotoxicity tests of all cell lines were carried out in a 96-well format with the CellTiter-Glo Luminescent Cell Viability Assay (Promega, Madison,USA) according to the manufacturer’s instructions.
To determine which cellular pathways are targeted by SARS-CoV, functional categories enriched among SARS cellular interaction partners were identified using a Gene Ontology (GO) over-representation analysis. For this purpose, p-values were determined with the hypergeometric test implemented in the Ontologizer software [12]. P-Values were corrected for multiple testing using the FDR-method by Benjamini and Hochberg [52] and significant terms were identified at a threshold of 0.05.
We analyzed whether proteins involved in protein complexes were preferentially targeted by SARS-CoV proteins and which complexes were preferentially targeted. For this purpose, protein complexes for humans were extracted from the CORUM database[53]. After removing complexes which were identical to another complex, we obtained a data set of 1184 complexes containing 2079 distinct proteins. P-values for the enrichment of protein complex subunits among SARS targets were calculated with the hypergeometric test assuming a background of ∼ 25,000 proteins (Table S4). We furthermore analyzed 9 complexes which targeted at least four subunits by SARS proteins. Using a hypergeometric test, we determined p-values for the enrichment of SARS target proteins among the subunits of each complex. P-values were corrected for multiple testing using the FDR- method by Benjamini and Hochberg (Table S5).
Interactions between SARS proteins and human proteins were connected to a network of human protein-protein interactions taken from the Human Protein Reference Database (HPRD, Release 7) [54] and the Biological General Repository for Interaction Datasets (BioGRID) database [55]. We then compared the distribution of degree (number of interactions) and betweenness centrality [13] for the viral targets against all other proteins in the human networks with the Kolmogorov-Smirnov test in R [56]. P-Values were again corrected for multiple testing using the FDR-method by Benjamini and Hochberg. Degree and betweenness centrality are alternative measures of network centrality for individual proteins. High degree characterizes so-called hubs which are highly interactive while high betweenness centrality characterizes so-called bottlenecks which are central to many connections between proteins.
|
10.1371/journal.pntd.0004269 | Altered Hypercoagulability Factors in Patients with Chronic Chagas Disease: Potential Biomarkers of Therapeutic Response | Thromboembolic events were described in patients with Chagas disease without cardiomyopathy. We aim to confirm if there is a hypercoagulable state in these patients and to determine if there is an early normalization of hemostasis factors after antiparasitic treatment. Ninety-nine individuals from Chagas disease-endemic areas were classified in two groups: G1, with T.cruzi infection (n = 56); G2, healthy individuals (n = 43). Twenty-four hemostasis factors were measured at baseline. G1 patients treated with benznidazole were followed for 36 months, recording clinical parameters and performance of conventional serology, chemiluminescent enzyme-linked immunosorbent assay (trypomastigote-derived glycosylphosphatidylinositol-anchored mucins), quantitative polymerase chain reaction, and hemostasis tests every 6-month visits. Prothrombin fragment 1+2 (F1+2) and endogenous thrombin potential (ETP) were abnormally expressed in 77% and 50% of infected patients at baseline but returned to and remained at normal levels shortly after treatment in 76% and 96% of cases, respectively. Plasmin-antiplasmin complexes (PAP) were altered before treatment in 32% of G1 patients but normalized in 94% of cases several months after treatment. None of the patients with normal F1+2 values during follow-up had a positive qRT-PCR result, but 3/24 patients (13%) with normal ETP values did. In a percentage of chronic T. cruzi infected patients treated with benznidazole, altered coagulation markers returned into normal levels. F1+2, ETP and PAP could be useful markers for assessing sustained response to benznidazole.
| The manuscript describes the results of a study whose aim was to assess the tendency to coagulate in people suffering from a parasitic infection frequent in Latin America named T. cruzi infection or Chagas disease, by the study of several coagulation factors. According to the state of the art in this topic, specific treatment for Chagas disease is recommended in recent (acute) and late (chronic) stages of the infection. The effectiveness of current available drugs in the chronic stage of infection is still a topic of debate due to inconsistent results across studies and a lack of early measurable parameters of response to specific treatment. Another aim of this study was to determine if the presence of an upregulated procoagulative activity in plasma in people suffering T. cruzi infection could be used as potential marker that indicates therapeutic response in people at chronic stage of the disease. The results of this study suggest that measurements of alterations of procoagulative activity may be useful to indicate specific treatment for T. cruzi chronically infected patients and new data concerning early response to treatment biomarkers.
| Chagas disease (CD) is one of 17 neglected tropical diseases recognized by the World Health Organization. Caused by the protozoan parasite Trypanosoma cruzi, it mainly affects people with poor socioeconomic status and limited health care access in endemic and nonendemic countries. [1, 2]
Thrombosis is considered as a pathological deviation of haemostasis, and it is characterized by intravascular thrombus formation and vessel occlusion. Perturbation of hemostasis is an important factor in the pathogenesis of thromboembolic events, which can be caused by blood flow dysregulation, endothelial injury, and coagulation system alterations.
Recently, is has been described that under certain circumstances thrombosis is a physiological process that constitutes an intrinsic effector mechanism of innate immunity, and the process has been defined as “immunothrombosis”. [3] It is activated after the recognition of pathogens and damaged cells, and inhibits pathogen dissemination and survival. Immunothrombosis can therefore be regarded as a newly identified, crucial element of intravascular immunity, which is a part of the immune system that encompasses a wide range of host strategies to detect and protect against pathogens in the vasculature. Dysregulation of immunothrombosis is likely to constitute a key event in the development of thrombotic disorders. [3]
Infectious disease can cause a hypercoagulable state through the upregulation of tissue factor in monocytes, the generation of procoagulant microparticles, the activation of the coagulation intrinsic pathway, platelet activation, and NETs (Neutrophil Extracellular Traps) release.[3] Different infectious agents may cause different responses but a final degree of hypercoagulability can be a common trait as one of the biological endpoints. Additionally, patients with chronic inflammation may also present platelet adhesion events, which are considered inflammatory processes and can be observed in patients with chronic T. cruzi infection, even in the asymptomatic stages. [4] Infection itself can cause vasculitis, increasing proinflammatory cytokine levels and perpetuating the risk of thrombotic events. [5] In the case of the Chagas’ disease the effect of hemostasis in the bradikinin formation, through the effect of factor XII activation in the Kallikrein-Kinin system, can modify the type 1 immune response and then modulate the antiparasite immunity as suggested in a mice model of subcutaneous infection by T.cruzi. [6]
Thromboembolic events and dilated cardiomyopathy, ventricular aneurysms, and intracavitary thrombosis are associated with CD. [7, 8] Rheological factors can induce intraluminal thrombus formation with the risk of embolism. [9] Alterations of molecular markers of coagulation system activation have been described in T. cruzi infection individuals with or without clinical thrombosis. [9–12] Other factors, such as injury to vessel walls by parasites or changes in blood viscosity due to host immune response, may influence in the development of thromboembolic events in T. cruzi-infected individuals without Chagas cardiomyopathy or other vascular risk factors. [13]Based on studies performed in humans with chronic T. cruzi infection, there are controversial results regarding the existence of a prothrombotic status in T. cruzi-infected patients. [13,14] There is an study in which a of higher prothrombotic status in the CD group was not found, but the control group were individuals without T. cruzi infection and heart failure. [14] In previous studies performed in murine models, several abnormalities of the heart microcirculation of individuals with chronic CD were pointed out, but they did not find evidence of thrombi and neither thromboembolism. [15, 16] Higher levels of the hypercoagulability markers prothrombin fragment 1+2 (F1+2), thrombin-antithrombin complexes (TAT), fibrinogen/fibrin degradation products, plasminogen activator inhibitor type 1 (PAI-1), and D-dimer have been reported in T. cruzi–infected patients compared with healthy individuals. [10, 11] A pilot study performed by our group showed that endogenous thrombin potential (ETP) and F1+2 levels were outside normal ranges in 73% and 80% of T. cruzi–infected patients without advanced heart disease, respectively. [12] We demonstrated a 100% and 73% decrease in these levels six months after treatment with benznidazole. Thus, if they prove to remain stable in time, hypercoagulability factors could be used as biomarkers of therapeutic response in CD. Besides, although whether or not chronic Chagas disease is an independent vascular risk factor remains to be confirmed. [17,18]
While specific treatment is recommended in both acute and chronic stages of infection [19,20], there are only two drugs (i.e., benznidazole and nifurtimox) available for the treatment of CD. The mechanism of action of benznidazole relates to the nitro-reduction of components of the parasite, the binding of metabolites of the nuclear DNA and k-DNA of T. cruzi and the lipids and proteins of the parasite. [21] In adults, benznidazole has a high rate of adverse effects, which can be classified into three groups: (i) hypersensitivity, including dermatitis with cutaneous eruptions (usually appearing between days 7 and 10), myalgias, arthralgias, and lymphadenopathy; (ii) polyneuropathy, paresthesias, and polineuritis usually during the 4th week of treatment); and (iii) bone marrow disorders, such as thrombopenic purpura and agranulocytosis (usually after the second week of treatment). [22]Furthermore, the effectiveness of these drugs in the chronic stage of infection is still a topic of debate due to inconsistent studies’ results [23–25] and a lack of early biomarkers of response to specific T. cruzi treatment with benznidazole. [26]
Following on from our pilot study [12], here we increased the sample size and extended follow-up to further investigate the value of hypercoagulability factors as biomarkers of treatment response in CD. We also added current treatment response parameters measured by conventional serology, serology for lytic anti-α-galactosyl (anti-α-Gal) antibodies against T.cruzi [27–29], and quantitative reverse transcription polymerase chain reaction (qRT-PCR). [30]
The aims of the study were to investigate alterations of hypercoagulability factors in patients chronically infected with T. cruzi and determine whether there is an early and sustainable improvement of the hypercoagulability factors after antiparasitic treatment.
Written informed consent was obtained from participants before being recruited (all of them were adults). Approval for the protocols and for the informed consent was obtained from the Hospital Clínic of Barcelona Ethics Review Committee.
This is a descriptive study of 99 individuals (56 with T.cruzi infection and 43 healthy individuals) from Latin American, where CD is endemic. All the individuals were evaluated at the Centre for International Health at Hospital Clínic in Barcelona, Spain.
Ninety-nine individuals from CD-endemic areas living in Barcelona were invited to participate. Inclusion criteria were an age of over 18 years and provision of signed informed consent. Exclusion criteria were pregnancy, non-Chagasic cardiopathy, late chronic cardiac or digestive forms of CD, other acute or chronic infections, inflammatory or immunological diseases, and chronic systemic diseases (high blood pressure and diabetes).
After signing the informed consent form, participants were asked for clinical and epidemiological data, including area of origin and risk factors for the CD transmission. The information recorded included vascular risk factors, toxic habits, and cardiological and/or vascular events.
Conventional serology of T.cruzi infection was established using two ELISA kits: a commercial kit with recombinant antigens (BioELISA Chagas, Biokit S.A.,Barcelona-Spain) and an in-house kit with whole T.cruzi epimastigote antigen, as described. [12, 31]. Diagnosis was confirmed by a positive result on both tests. [19] Following serological tests results, participants were divided into two groups: those with T.cruzi infection (Group 1 [G1]) and those without (Group 2 [G2]). All the participants underwent human immunodeficiency virus testing, basic blood and biochemical tests (including renal and liver function), and specific evaluation of hemostasis factors.
For the hemostasis studies, blood was collected in citrate-containing tubes (Becton Dickinson), samples were centrifuged, and platelet-poor plasma aliquots were frozen at –80°C until assayed. Prothrombin time, activated partial thromboplastin time, coagulation factor VIII, protein C activity, free and total protein S levels, antithrombin and plasminogen activity, F1+2, plasmin-antiplasmin complexes (PAP), factor VIIa, PAI-1, P-selectin, factor V Leiden and prothrombin gene G20210A mutation, lupus anticoagulant and anticardiolipin antibodies were measured as previously described. [12] D-dimer was measured using an automated turbidimetric test (Siemens Healthcare Diagnostics) and ETP was assessed using a continuous chromogenic thrombin generation assay and ETP Curves software (Siemens). The ETP coagulation test was initiated by using human recombinant tissue factor, phospholipids, and calcium ions. ADAMTS-13 was measured using a commercial chromogenic method (American Diagnostica). Factor XIIa was determined by a direct quantitative commercially available immunoassay (Shield Diagnostics) with a highly specific monoclonal antibody that does not recognize its zymogen factor XII.[32] Plasma tissue factor levels were determined using a commercial kit (American Diagnostica) according to the manufacturer’s protocol. Plasma levels of von Willebrand factor antigen were determined by enzyme-linked immunosorbent assay (ELISA) (Corgenix). Procoagulant activity of microparticles was measured using a functional assay with the addition of factors Xa, Va, and prothrombin after microparticle capture in the solid phase using annexin V (Hyphen Biomed). Soluble CD40L was measured by ELISA (R&D Systems).
qRT-PCR [30] and a chemiluminescent ELISA assay based on a highly purified, trypomastigote-derived glycosylphosphatidylinositol-anchored mucin (tGPI-mucin) antigen for the serological detection of lytic anti-α-Gal antibodies against T.cruzi (AT CL-ELISA) [27–29, 33–36], were performed in G1 at month 0 (baseline), and 6, 12, 18, 24, 30, and 36 months post-treatment. For AT CL-ELISA, a serum sample was considered positive when the titer was ≥1.0 and negative when it was ≤0.9. Inconclusive or equivocal results were determined by a titer between 0.9 and 1.0. [27, 35]All sera were tested in duplicate and the results were expressed as the mean of two simultaneous determinations.
G1 patients were studied using a protocol that included a 12-lead electrocardiogram, chest X-ray, and echocardiogram. They were followed up every 6 months for at least 36 months. At each visit, clinical data were collected and the following tests were performed: ELISA, AT CL-ELISA, qRT-PCR, and hemostasis tests. Other tests were performed according to individual symptoms. Specific treatment with benznidazole (5 mg/kg/day for 60 days) was offered to all T.cruzi–infected patients, and those treated were monitored fortnightly for clinical and analytical assessment. Treatment was considered complete when at least 80% of the total dose was reached.
A hypercoagulable state is defined as the presence, in certain individuals, of thrombotic potentialities that activate the endothelium and the formative elements of the blood (mainly, platelets) that favors plasma kinetics that lead to the formation of thrombin, which disturbs fibrinolytic activity and produces hemorheological changes with turbulence phenomena that predispose to thrombogenesis. [18]
Quantitative variables were presented as medians and interquartile range (IQR) and were compared between groups using the Wilcoxon rank sum test. Qualitative variables were reported using absolute frequencies and percentages and between-group comparisons were made using Fisher’s exact test. Hypercoagulability biomarker variation over time was assessed using a mixed-effect linear regression model with a random intercept structure. Hypercoagulability factors were used as dependent variables and follow-up time as the explanatory variable, with one category for each time point: baseline, month 6 (reference for comparisons), and months 12, 18, 24, and 36. This type of model allows for the inclusion of random effects in addition to the overall error term. Random intercept regression was also used to assess whether antibody levels measured by ELISA and AT CL-ELISA approached the negative threshold during follow-up. The response variable was the distance from this threshold (i.e., the difference between each ELISA or AT CL-ELISA value and the negative cutoff) and the explanatory variable was the follow-up time from month six (reference) to month 36. The regression coefficients express the effect estimate of follow-up on the outcome variable.
The pattern of the relationships between hypercoagulability biomarkers was assessed by multiple correspondence analysis (MCA) using the Burt matrix approach. [37, 38] The MCA represents a method for analyzing multi-way contingency table containing measure of correspondence between row (subjects) and columns (levels of variables). The interpretation is based upon proximities between levels of variables (or points) in a low-dimensional map. The firsts dimensions (usually one or two) account for meaningful amounts of variance and are those retained for the map definition and interpretation. The first dimension accounts for a maximal amount of total variance in the observed variables. Under typical conditions, this means that the first component will be correlated with at least some of the observed variables. The second dimension has two important characteristics: it accounts for a maximal amount of variance in the data set that is not accounted for by the first dimension, thus it is correlated with some of the observed variables that not display strong correlations with dimension 1; and it is uncorrelated with dimension 1. Looking at the map, the proximity between levels of different variables means that these levels tend to appear together in the observations. Since the levels of the same variable cannot occur together, the proximity between levels of the same variable means that the groups of observations associated with these levels are themselves similar. A level far away from the origin (of the dimensions) means that is well-represented in the map, thus that level is meaningful for the interpretation of the dimension(s). All levels that are not useful for the solution are near the origin. Supplementary (passive) variables are those not used for the solution but mapped in the graph in order to help in the interpretation.
The biomarkers were classified into three categories: normalization of values throughout follow-up, non-sustained normalization during follow-up and normal values at baseline. Two additional variables were considered: qRT-PCR results during follow-up (categories: always negative and sometime positive) and level of adherence (categories: 80% and 100%). All the tests were 2-tailed and the confidence level was set at 95%. The analyses were performed using Stata 13 (Stata Corporation, College Station, TX, USA).
Ninety-nine individuals (76 women) were studied. Fifty-six of these (43 women) were T.cruzi–positive (G1) and 43 (33 women) were T.cruzi–negative. The mean ages were 34 (SD, 9) years for the overall group (range 17–56, median 33), 37 (SD, 9) years for G1, and 32 (SD, 7) years for G2. Fifty G1 patients were treated with benznidazole (six were lost to follow-up before starting treatment due to unexpected work-related changes in the migratory process). Forty-five (90%) completed treatment. Eighty-six participants (87%) (51 [91%] in G1 and 35 [81%] in G2) were from Bolivia. None of the participants traveled to their countries or other CD-endemic areas during follow-up. The clinical and demographic data are summarized in Table 1. The epidemiological and baseline clinical data were similar in both groups, making them statistically comparable.
Comparison of the 24 hypercoagulability biomarkers at baseline between (untreated) G1 and G2 individuals showed statistically significant differences for D-dimer (P = .0262); F1+2 (abnormal values in 43/56 G1 patients [77%], P < .0001), PAP (abnormal values in 17/56 G1 patients [30%], P = .0111), P-selectin (abnormal values in 7/56 G1 patients [13%] P = .0177), and ETP (abnormal values in 28/56 G1 patients [50%], P < .0013), and circulating microparticles (P = .0112) (Table 2). D-dimer levels were normal in all the individuals in G1 and G2, and microparticles were within the normal range in a high percentage of patients (86% in G1 and 93% in G2, P = .3402). Our findings showed that a high percentage of patients with chronic T.cruzi infection have a hypercoagulable state regardless the clinical stage of disease, thus confirming the observations of previous studies. [11–13]
Thirty-three (76%) of the 43 patients with abnormal baseline F1+2 values achieved normal levels after a median follow-up of 9 month (IQR, 8). All but one of the 28 patients with abnormal ETP values before treatment showed normal values at 6 months (IQR, 3). These values were maintained throughout follow-up (30 months; IQR, 28) in 15 patients (60%). Fifteen of the 17 patients with abnormal baseline PAP values showed normal values 7 months (IQR, 7) after treatment and nine of these (60%) maintained these values throughout follow-up (28 months; IQR,11). However, PAP values at 12 and 48 months seemed to be higher than those at 6 months, but the confidence interval indicates a lack of precision for both time point effect estimates (Table 3). Thus, once normalized, F1+2 and ETP levels did not increase again significantly after treatment. Fig 1 shows a graphic representation of these results.
F1+2 values are an indirect measure of the amount of thrombin generated in vivo (mainly due to endothelial injury, even in subclinical states) [39], and ETP levels indicate the potential amount of thrombin that can be formed when blood coagulation is activated through the addition of tissue factor. PAP complexes are markers of fibrinolysis. Upon activation, plasmin, which is primarily responsible for a controlled and regulated dissolution of the fibrin polymers into soluble fragments, is immediately inactivated by antiplasmin, forming PAP complexes. [40] Therefore, it is conceivable that the increase formation of PAP complexes stems from excessive formation of fibrin in the blood stream of untreated T. cruzi infected patients. Soluble P-selectin is considered a biomarker of in vivo platelet activation. P-selectin is contained in the α-granules of platelets; following platelet activation, the soluble form is expressed on the platelet surface and then shed by cleavage. P-selectin has been shown to act as a link between thrombosis and inflammation. [41] Additionally, the four biomarkers-F1+2, ETP, PAP complexes, and P-selectin-reflect are highly stable over time.
A hypercoagulable state is a term that pretends to denominate a condition in which there is an increased tendency toward blood clotting. There is not a universally accepted definition for this state based in biomarkers values, but an increase in several of them suggests the possibility of an increase in the person's chances of developing blood clots. The increases in F1+2, PAP and ETP are congruent with this idea: F1+2 and PAP indicate the actual amount of thrombin and plasmin formed, as markers in procoagulant and fibrinolysis pathways, respectively; and ETP indicates the potential amount of thrombin that can be formed considering globally all the activators, inhibitors and substrates of the hemostasis present in the plasma. The increase observed in these biomarkers is good enough to be an argument to point out a hypercoagulable state in patients with Chagas’ disease.
Sixteen (33%) of the 56 G1 patients had a positive qRT-PCR result at baseline, but only four of these had a positive result after treatment (treatment failure rate of 25% in this subgroup). Five of the 34 patients with a negative baseline qRT-PCR result showed a positive result during follow-up. None of the patients with normal F1+2 values during follow-up had a positive qRT-PCR result, but 3(13%) of the 24 patients with normal ETP values during follow-up did. Of the patients with altered levels of F1+2, ETP, or PAP complexes at baseline, a positive qRT-PCR result during follow-up was not significantly associated with changes observed in lytic anti-α-Gal antibodies, F1+2, ETP, and/or PAP levels.
A positive qRT-PCR result after treatment in patients who achieved normalization of F1+2, ETP, and/or PAP could mean that a decrease in parasite load is sufficient to modify the hypercoagulable state or that benznidazole, which acts on the redox system, could modify these biomarkers without eliminating the parasites. This would limit the use of these factors as biomarkers for parasite elimination, although they could be valuable indicators of treatment response and add support to the theory that, by reverting the hypercoagulable state, benznidazole may also prevent clinical thrombotic events.
Conventional ELISA results were positive in all the patients in G1. Although, as expected, antibodies remained positive throughout follow-up, a slight decrease was detected by the commercial and in-house methods during this period. A statistically significant relevant decrease, was only observed with the in-house test from month 18 onwards (P = .0006).
Lytic anti-α-Gal antibodies were positive in 52 (96%) of the 54 patients tested before treatment, and in all patients AT CL-ELISA remained within positive levels to the end of the follow-up (Fig 2). Besides, there was no correlation between lytic anti-α-Gal antibody assay and the hemostasis factors evaluated. In relation to previous studies’ results, early decreases in lytic anti-α-Gal antibodies were expected to be observed. On the contrary, a decrease in levels was evident at month 12 and this was significant since month 18 and forward (P = .0052). [28, 34]
Adherence to treatment was high, with only five patients not achieving 80% of the total dose. All five patients showed abnormal F1+2 values throughout follow-up and 3 (60%) had abnormal ETP and PAP values. One of the five patients had a positive qRT-PCR result during follow-up, and all five maintained the same positive ELISA and AT CL-ELISA results throughout follow-up. A large cohort of adolescents with T cruzi infection treated with benznidazole showed seronegativity in lytic anti-α-Gal antibodies, as measured by AT CL-ELISA, in 58% and 85% of the patients 36 and 72 months after treatment, respectively. [28, 34] The differences between those studies and ours may be due to the nature of the cohorts (adolescents vs. adults) and the stage of the disease. Nevertheless, both studies showed a similar trend towards a reduction in lytic anti-α-Gal antibodies following treatment with benznidazole.
We studied the relationship between normalization of hypercoagulability markers F1+2, PAP, and ETP and qRT-PCR results by multiple correspondence analyses (MCA). Due to the low rate of positive qRT-PCR results, this variable was used as a supplementary variable jointly with treatment adherence. The MCA results (Fig 3) showed an association between complete normalization of PAP and ETP levels and non-sustained and marginally abnormal values in F1+2. These factors had the highest contribution and correlation in the positive part of the second dimension, while normal baseline ETP and PAP values had the highest contribution and correlation in the negative part. F1+2 normalization clearly characterized the positive part of the first dimension, while non-sustained normalization of PAP and ETP values clearly characterized the negative part. In other words, the sustained normalization observed post-treatment in PAP and ETP, could, despite the non-sustained normalization of F1+2 values, reflect response to antiparasitic treatment due to the strong correlation between these three variables.
The projection of qRT-PCR results and adherence to treatment in the solution space provided little additional information. Consistently negative qRT-PCR results throughout follow-up appear to be related to 100% treatment adherence.
In a recent study, the authors found that the serum samples of 37 individuals with chronic Chagas disease showed an upregulation of specific fragments of apolipoprotein A-1 (Apo A1) and one fibronectin fragment, that returned to normal levels in 43% of them three years after a treatment with nifurtimox. [38] Apo A1 and fibronectin fragment were altered in all the 37 patients with T.cruzi infection before treatment, but the number of patients treated with that normalized levels was lower than in our series (60% and 96% of patients who normalized F1+2 and ETP values).
This study has some limitations. Although the sample size was calculated to obtain sufficient statistical power to answer the hypothesis, a larger sample may have detected differences that would be expected to appear earlier (e.g., before 12 months). The lost to follow-up samples also affected the estimates. Even within Spain, it is difficult to follow individuals with high migratory mobility for long periods. In addition, the fact that only 30% of patients had a positive baseline qRT-PCR result was a constraint for assessing the effect of treatment.
In conclusion, patients with chronic T.cruzi infection have a potential hypercoagulable state, regardless of cardiological and/or digestive involvement. The hypercoagulability markers F1+2 and ETP were abnormally expressed in a high percentage of patients with chronic T.cruzi infection before treatment (77% and 50%, respectively) but returned to and remained at normal levels shortly after treatment in 76% and 96% of patients, respectively. Baseline PAP values were altered in just 30% of patients before treatment, but normalized several months after treatment in 88% of these. These three hypercoagulability biomarkers could be useful for assessing short-term response to treatment. However, the fact that normal values were seen in some infected patients, including some with positive post-treatment qRT-PCR results, reduces their usefulness as universal biomarkers. The decrease in hypercoagulability factor levels could be explained by a decrease in parasitemia or by other benznidazole effect.
|
10.1371/journal.pcbi.1006438 | Modeling effects of voltage dependent properties of the cardiac muscarinic receptor on human sinus node function | The cardiac muscarinic receptor (M2R) regulates heart rate, in part, by modulating the acetylcholine (ACh) activated K+ current IK,ACh through dissociation of G-proteins, that in turn activate KACh channels. Recently, M2Rs were noted to exhibit intrinsic voltage sensitivity, i.e. their affinity for ligands varies in a voltage dependent manner. The voltage sensitivity of M2R implies that the affinity for ACh (and thus the ACh effect) varies throughout the time course of a cardiac electrical cycle. The aim of this study was to investigate the contribution of M2R voltage sensitivity to the rate and shape of the human sinus node action potentials in physiological and pathophysiological conditions. We developed a Markovian model of the IK,ACh modulation by voltage and integrated it into a computational model of human sinus node. We performed simulations with the integrated model varying ACh concentration and voltage sensitivity. Low ACh exerted a larger effect on IK,ACh at hyperpolarized versus depolarized membrane voltages. This led to a slowing of the pacemaker rate due to an attenuated slope of phase 4 depolarization with only marginal effect on action potential duration and amplitude. We also simulated the theoretical effects of genetic variants that alter the voltage sensitivity of M2R. Modest negative shifts in voltage sensitivity, predicted to increase the affinity of the receptor for ACh, slowed the rate of phase 4 depolarization and slowed heart rate, while modest positive shifts increased heart rate. These simulations support our hypothesis that altered M2R voltage sensitivity contributes to disease and provide a novel mechanistic foundation to study clinical disorders such as atrial fibrillation and inappropriate sinus tachycardia.
| Heart rate regulation is dependent upon a delicate interplay between parasympathetic and sympathetic nerve activity at the level of the sinus node. Acetylcholine slows the heart rate by activating the M2 muscarinic receptor (M2R) that, in turn, opens the acetylcholine-activated potassium channel (IK,ACh) to slow the firing of the sinus node. Surprisingly, the M2R is sensitive to membrane potential and undergoes conformational changes throughout the cardiac action potential that alter the affinity for acetylcholine, with secondary consequences for IK,ACh activity. Here, we investigated the contribution of M2R voltage sensitivity to the rate and shape of the human sinus node action potential in physiological and pathophysiological conditions, using a Markovian model of the IK,ACh channel integrated into a computational model of human sinus node. The computational model allowed us to assess the effects of potential genetic variants that alter specific properties of voltage sensitivity. Our results indicate that alterations in the M2R voltage sensitivity play a significant role in the physiology and pathophysiology of the human sinus node and atria. Our computational model is relevant for future studies aimed at the design and development of anti-arrhythmic agents that specifically target the unique voltage-sensitive properties of M2R.
| The cardiac muscarinic receptor (M2R) plays a crucial role in regulating heart rate variability and vulnerability to atrial arrhythmia by modulating the acetylcholine (ACh) activated K+ current IK,ACh. Cardiac KACh channels are heteromultimers composed of two G-protein-coupled inward rectifier K+ channel subunits, Kir 3.1 and Kir 3.4 [1]. ACh activation of M2R triggers dissociation of the G beta-gamma subunits (Gβɣ) that in turn directly activate Kir 3.x subunits to conduct IK,ACh. Unexpectedly, M2Rs were discovered to possess an intrinsic ability to sense transmembrane voltage [2] and the affinity of the receptor for ligands was noted to vary in response to changes in membrane voltage [3]. In particular, the affinity of the receptor for ACh is increased at hyperpolarized membrane potentials and decreased at depolarized potentials. The changes in affinity exert a downstream effect on the KACh channel such that the channel is more active (more current) at hyperpolarized potentials and less active (less current) at depolarized potentials. The observation that M2Rs are intrinsically voltage sensitive has profound implications for cellular signaling in excitable tissues, such as heart. For example, voltage sensitive behavior provides a mechanistic explanation for a decades-old enigmatic process called IK,ACh “relaxation” gating. Relaxation gating refers to a time-dependent change in current magnitude following a depolarizing or hyperpolarizing voltage step [4] and has important consequences for shaping the cardiac action potentials (AP), especially in the sinus node. We recently proposed that relaxation gating represents a voltage dependent change in ACh affinity induced by voltage dependent conformational changes within M2R [5]. Our experimental data provide a mechanistic basis to explain the participation of IK,ACh in the modest chronotropic effects induced by resting vagal tone. As a result of conformational changes in the M2R, the affinity for ACh varies throughout the cardiac electrical cycle such that low (subsaturating) ACh concentrations preferentially activate IK,ACh during diastolic membrane voltages thereby slowing the spontaneous firing rate without appreciably altering AP duration (APD).
Alterations in the voltage sensitivity of M2R could theoretically contribute to cardiovascular diseases that clinically present with apparent changes in vagal tone. For example, genetic variants in M2R that shift the receptor occupancy into the hyperpolarized state would be expected to increase the affinity of the receptor for ACh and thus activate more KACh channels at a given ACh concentration (or degree of vagal tone). Accordingly, genetic variants in M2R that shift the receptor occupancy into the hyperpolarized state might explain the clinical phenotype of vagally-mediated atrial fibrillation (AF), patients who present with bradycardia in the setting of physiological (basal) ACh concentrations. Alternatively, genetic variants in M2R that shift the receptor occupancy into the depolarized state would be expected to decrease the affinity of the receptor for ACh and thus fewer KACh channels activate at a given ACh concentration (or given degree of vagal tone). This would decrease the effects of vagal modulation of heart rate, thereby increasing basal heart rate, as observed in the syndrome of inappropriate sinus tachycardia (IST).
To provide insights into the contribution of M2R voltage sensitivity to cardiac electrophysiology in physiological and pathophysiological conditions, we extended our previous Markovian model of M2R [5] to incorporate Gβɣ-mediated activation of the KACh channel and integrated the revised Markovian model into a human model of the sinus node (SN) cell model [6]. Based on experimental data from isolated human SN cells [7, 8], Fabbri and colleagues recently published a comprehensive model of the human SN pacemaker cell that faithfully recapitulated the effects of autonomic modulation as well as mutations associated with SN dysfunction [6]. In the Fabbri model and its parent model [9], IK,ACh is described by a voltage- and [ACh]-dependent gate, but the intrinsic voltage sensitivity of M2R is not incorporated. Here, we introduce a model reproducing the effects of M2R voltage sensitivity on human SN cell function under physiological and pathophysiological conditions. These simulations support our hypothesis that altered M2R voltage sensitivity contributes to disease and provide a novel mechanistic foundation to study clinical disorders such as AF and IST.
For decades, the contribution of IK,ACh to the modest chronotropic effects of ‘physiological’ or low-dose ACh has been debated [10–12]. Based on the M2R voltage-dependent properties, we predict that subsaturating ACh concentrations exert a larger effect during diastolic (hyperpolarized) membrane voltages, compared to the voltages during the cardiac AP (depolarized) [3, 5]; thus preferentially slowing the pacemaker rate with minimal effect on APD. To test this hypothesis, we simulated the effects of varying low, sub-saturating concentrations (e.g., 20–100 nM) of ACh on sinus node AP properties (Fig 1). The most striking effect of increasing ACh concentration was reduced slope of phase 4 depolarization and the corresponding increase in basic cycle length (Fig 1A, Table 1, Vshift = 0), with minimal shortening of APD90 (Table 2, Vshift = 0). Thus, the basic cycle length (BCL) increased from 827 ms in the absence of ACh, to 1585 ms in the presence of 0.1 μM ACh. The AP amplitude decreased steadily by up to 4.7 mV at the highest tested concentration of 0.1 μM ACh (inset of Fig 1A). Simulated open probability O and IK,ACh for different ACh concentrations are shown in Fig 1B and 1C. Taken together, these simulations indicate that subsaturating concentrations of ACh slow spontaneous excitation of the SN cell by inhibiting the rise of phase 4 depolarization, without appreciably shortening APD90 or reducing the amplitude of the AP.
In our previous experimental studies using isolated feline left atrial myocytes, the voltage dependence of M2R was explored by measuring the ACh concentration-IK,ACh response relationship at hyperpolarized (-100 mV) and depolarized membrane voltages (+50 mV) [3, 5]. These experiments indicated that the affinity of the receptor for ACh was greater at hyperpolarized membrane voltages, compared to depolarized voltages. We reasoned that, similar to voltage-gated ion channels, putative disease-associated mutations in M2R might alter the voltage sensitivity of the M2R, with unique consequences for sinus node AP properties and heart rate responses. We modified rate parameters (Eqs 5 and 6) to shift the receptor occupancy towards a hyperpolarized state (higher affinity) or a depolarized state (lower affinity) (S1A Fig). Thus, we simulated the effects of shifting the M2R voltage sensitivity (Fig 2, S3C Fig). S1B Fig highlights the effects of voltage shifts on the state O. Negative shifts (e.g., Vshift of -30 mV and -150 mV), which caused a more hyperpolarized state of the receptor, increased the occupancy of the U1 and B1 states, as well as the state O. Negative voltage shifts resulted in a slight leftward shift in the concentration-response curve when the cell was held at +50 mV (Fig 3E). Likewise, positive shifts in M2R voltage sensitive parameters caused a slight rightward shift in the concentration-response curve for a holding potential of Vh = -100 mV (Fig 3A). To avoid local minima during the parameter fitting, the model was forced to favor the open state (state O equal to 1), at the maximum concentration of 10 μM ACh and to favor the closed state (O equal to 0), with no ACh present. Further, U1 was forced to be as high as possible at a holding potential of -100mV and U2 at +50 mV, in the absence of ACh. Thus, negative and positive voltage shifts did not move the steady-state concentration-response relationships outside of these ranges. Notwithstanding, the voltages experienced by the single cell model vary between -60 and +30 mV (Fig 1A), within the minimum and maximum ranges defined by the parameter fitting.
We previously described the kinetics of “relaxation” gating of IK,ACh in terms of activation and deactivation of KACh channels in the setting of subsaturating ACh concentrations [5]. Activation kinetics were measured by first stepping to a depolarized voltage (+60 mV) to close a large portion of KACh channels at a physiological voltage, followed by stepping through a range of voltages to measure activation of IK,ACh. Deactivation kinetics were assessed by a pre-pulse to a hyperpolarized voltage (-100 mV) to open channels, followed by variable test voltage steps to measure the rate of KACh channel closure. Accordingly, simulated IK,ACh evoked by 0.1 μM ACh using the activation and deactivation voltage protocols are presented in S1 Fig and the kinetic parameters are described in Table 3. The simulations recapitulate the experimental features of IK,ACh relaxation gating [5], as shown in S1 Fig. The effects of voltage shifts in M2R voltage sensitive parameters on IK,ACh relaxation gating parameters are shown in Fig 3.
Next, we simulated the effects of negative and positive voltage shifts in M2R voltage-sensitive parameters on sinus node APs, together with the corresponding effects on state O and IK,ACh (Fig 4). Negative shifts (e.g., Vshift of -10 mV and -30 mV), which would be predicted to increase ACh affinity, reduced the slope of phase 4 depolarization in a concentration-dependent manner, with minimal effects on AP amplitude or APD90 (Fig 4A versus 4D, Tables 1 and 2). The reduction in the slope of phase 4 depolarization induced by a negative Vshift was due to an increase in the open probability of KACh channels relative to the control condition, thereby increasing the magnitude of IK,ACh. These results indicate that shifts in the M2R voltage sensitivity impact the rate of spontaneous depolarization of sinus node APs.
To further characterize the physiological consequences of positive and negative voltage shifts, we studied the effects of variable M2R voltage sensitivity on the ACh concentration-heart rate response relationship (Fig 5A). Our model recapitulates experimental data [12] indicating that ACh concentrations ranging from 0.01 to 0.1 μM induce slowing of the spontaneous activity by 10% and 45%, respectively. Hyperpolarizing shifts in the M2R voltage dependent parameters shifted the relationship toward progressive spontaneous activity slowing. By contrast, depolarizing shifts in these parameters antagonized the spontaneous activity slowing induced by ACh. Taken together, these results suggest that shifts in M2R voltage sensitive parameters exert significant physiological effects on SN firing rate and AP parameters.
Next, we quantified the effects of the individual ACh-influenced cardiac currents on heart rate slowing. Fig 5B illustrates the relative contributions of ACh-sensitive currents, including our new model of IK,ACh, together with formulations of the L-type Ca2+ current (ICa,L) and the hyperpolarized-activated ‘funny current’ (If) from Fabbri et al. [6]. Inspection of the individual contributions of the ACh-sensitive currents reveals that our model of IK,ACh (incorporating the voltage-sensitivity of M2R) accounts for roughly 50% of the slowing at ~30nM and more for higher concentrations. These results highlight the important contribution of M2R voltage sensitivity to heart rate slowing induced by ACh.
The observation that M2Rs are intrinsically voltage sensitive has profound implications for cellular signaling in excitable tissues, such as heart. This could have important consequences for cardiovascular drug development in that the affinity for ligands may vary in a voltage- and ligand-specific manner. The properties of M2R voltage dependence provide a novel molecular basis to explain the previously perplexing “relaxation” gating of IK,ACh, originally described in the 1970s [4]. Precisely how the receptor’s voltage dependent properties influence the firing rate and shape of sinus node APs under physiological conditions, or in the presence of a M2R genetic defect, remains unknown. In order to address this question, we enhanced our previous Markovian model of M2R [5] by a 2-state Markovian model to incorporate G-protein activation of the KACh channel. The parameters of this revised model were adapted to experimental data from isolated feline left atrial myocytes, see Methods. We integrated the revised model into a human SN single cell model and used computational simulations to gain insights into effects of voltage sensitivity of M2Rs.
While there is no controversy as to the cardiac effects of strong vagal stimulation, the participation of IK,ACh in mediating the purely chronotropic effects of low (or “physiologic”) ACh concentrations has been debated over the past thirty years. Low concentrations of ACh (e.g., below 100 nM) and weak vagal stimulation reduce spontaneous pacemaker rate without altering APD or increasing the maximal diastolic potential [10]. DiFrancesco et al. proposed that weak vagal stimulation primarily inhibited If with little to no contribution from IK,ACh. Indeed, the EC50 for ACh modulation of If is an order of magnitude lower than that required for IK,ACh activation [11]. However, subsequent studies have corroborated an important role for IK,ACh in mediating the chronotropic effects of weak vagal stimulation and externally applied low ACh concentrations [12]. Perhaps the strongest evidence for IK,ACh contribution to basal chronotropy comes from the Kir3.4 knock-out mouse which displays a specific deficit in heart rate variability at rest [13, 14]. These studies confirm that low ACh concentrations (basal vagal tone) activate IK,ACh to slow heart rate. Our experimental and simulated data provide a mechanistic basis to explain the participation of IK,ACh in the modest chronotropic effects induced by resting vagal tone. Our experimental data predict that the affinity of M2R for ACh varies throughout the AP. The simulations performed here corroborate that subsaturating ACh concentrations preferentially open the KACh channel during diastolic membrane voltages and thereby slow the spontaneous firing rate. As seen in Fig 5B, IK,ACh accounts for roughly half of the heart rate slowing at concentrations around 30 nM, and more for higher concentrations.
In silico modeling is valuable for understanding fundamental features of physiology and pathophysiology [15–17]. This is especially relevant for modeling of IK,ACh in disease states. For example, numerous disease-causing mutations in ion channel genes alter the voltage sensitivity of the channel [18, 19]. In light of these observations, we simulated effects of hypothetical mutation-induced positive and negative shifts in M2R voltage sensitivity to ACh on sinus node AP properties and spontaneous firing. Modest voltage shifts (Vshift of ±10 mV) exerted significant effects on the slope of phase 4 depolarization and thus on spontaneous activity responses to subsaturating ACh concentrations, such as those elicited by vagal resting tone. Moderate voltage shifts (Vshift of ± 30 mV) caused more profound changes in spontaneous activity. These simulations provide a proof-of-principle for a theoretical contribution of altered M2R voltage sensitivity to cardiovascular disease states associated with changes in vagal tone. For example, parasympathetic induction and maintenance of atrial arrhythmias is a well-described phenomenon, first reported in 1930 [20]. Moreover, increased parasympathetic tone is an initiating factor in a subset of AF patients [21]. Our simulations suggest that genetic variants in M2R that shift the receptor occupancy towards hyperpolarized states (U1/B1) associated with increased ACh affinity of the receptor increase IKACh at a given ACh concentration (or degree of vagal tone). These genetic variants might explain, in part, the clinical phenotype of vagally-mediated AF patients who present with bradycardia in the setting of physiological (basal) ACh concentrations. Also, our simulations suggest that genetic variants in M2R that shift receptor occupancy towards depolarized states (U2/B2) associated with decreased ACh affinity of the receptor reduce IKACh at a given ACh concentration (or degree of vagal tone). Thus, positive shifts in M2R voltage sensitivity would decrease the effects of vagal modulation of heart rate, thereby increasing basal heart rate, as observed in the syndrome of IST. Indeed, a recent study confirmed decreased parasympathetic tone in IST patients [22]. While speculative, the hypothesis that altered M2R voltage sensitivity is relevant for cardiac diseases provides a novel mechanistic foundation to study disorders such as AF and IST.
There are several limitations inherent in the application of the model to describe the kinetics and behavior of IK,ACh. First, the receptor-channel model was fit and optimized to recreate the kinetics of IK,ACh occurring at a concentration of 0.1 μM ACh. While this is within the range of measured concentrations of ACh, measurements at lower concentrations could enable a more accurate reconstruction of the behavior of IK,ACh and a direct comparison to the effect of If at such concentrations. Furthermore, we acknowledge that the description of the process of dissociation of Gβɣ from the M2R to the activation of the KACh channel is highly simplified. We used a simple channel opening description with a 2-state Markovian model neglecting different binding properties of different Kir subunits or cooperativity mechanisms in binding of Gβɣ. This simplification was necessary as parameters for more complex models are not identifiable with the existent experimental data. Also, the model is not able to reproduce the distinct characteristics of the deactivation protocol time constants. The experimental data indicate that the deactivation time constant is nearly voltage-independent (Fig 3H), the model predicts an increase with membrane depolarization. Nonetheless, because the other simulated features have a very high similarity to the measured values (Fig 3B, 3C, 3F and 3G) and the time constants are in the range of less than half the length of an action potential, we believe that any error introduced does not significantly influence our findings. Additionally, dissociation or binding of Gβɣ in our model does not account for changes in the process due to other influences or any binding to sites other than the KACh channel.
The SN cell model recapitulates the experimental findings that ACh inhibits If and ICa,L, which also contribute to slowing of spontaneous pacing rate [23]. We argue that unlike IK,ACh, the M2R voltage dependent effects do not influence If and ICa,L on the time scale of the AP. If and ICa,L inhibition are mediated by inhibition of the cAMP-dependent protein kinase A cascade that functions on a much slower time scale than the APD. By contrast, our simulations demonstrate that voltage dependent conformational changes in M2R influence IK,ACh throughout the cardiac AP, modulating both firing rate and APD. Finally, because our model does not fully integrate all the components of the autonomic nervous system, it is possible that the effects of the putative genetic mutations might be mitigated by compensatory changes in the autonomous nervous system’s response to changes in heart rate.
The recent observation that M2Rs are intrinsically voltage sensitive has important implications for understanding the physiology and pathophysiology of parasympathetic regulation of heart rate and APD. By optimizing and integrating a new Markovian model into a human SN model, we show that low ACh concentrations preferentially slow beating rate, without shortening APD, and thereby provide additional support that IK,ACh participates in the purely chronotropic effects of basal vagal tone. Moreover, we explore the effects of altered M2R voltage sensitivity and provide a proof-of-principle foundation that altered sensitivity could result in clinical manifestations of disease states such as vagally-mediated atrial fibrillation and syndrome of inappropriate sinus tachycardia. Given the importance of parasympathetic regulation of atrial vulnerability, M2Rs represent an important therapeutic target to control or prevent atrial arrhythmias.
We developed a Markovian model to reconstruct the behavior of KACh channels at different ACh concentrations and varying transmembrane voltages (Fig 2). The model comprises 3 sub-models: (1) A Markovian model describing the kinetics of the M2R depending on different concentrations of ACh at different voltages (M2R model), (2) a Markovian model describing the activation of the KACh channel based on dispersion of Gβɣ protein from the receptor to the channel (KACh channel model), and (3) a model of potassium current through KACh channels.
The parameters of the model were determined by iterative stochastic optimization as previously described [5]. Model parameterization was implemented in Matlab R2017a (The Mathworks Inc., Natick, MA) and the Matlab Parallel Computing Toolbox. An error function based on the root mean squared differences of the measured versus simulated features of the activation protocol, deactivation protocol, and concentration response curve from [5] was minimized.
Measured features were based on whole-cell voltage-clamp experiments [5]. IK,ACh for the activation and deactivation clamp protocols was recorded in the presence of 0.1 μM ACh. The 3 best measured currents, see S1 Fig, from [5] of each protocol were then fitted to a mono-exponential equation, averaged and then normalized to Coff at -10 mV:
I(t)=C+Aexp(-tτ)
(10)
with the constants A and C, and the time-constant τ of activation (on) and deactivation (off). We choose this strategy, as extracting mono-exponential features and then averaging them introduces less error in the overall reconstructed behavior than averaging the signals themselves. The same fitting approach was used during the model parameterization.
Concentration response curves were measured as the resulting IK,ACh at either a holding potential of -100 mV or +50 mV at different ACh concentrations (Fig 5A and 5E). The currents were then normalized to the current measured at maximal ACh concentration for each holding potential.
A total of 8 simulated features fs,i, consisting of six features for the voltage clamp protocols (i.e. Con, Coff, Aon, Aoff, τon, and τoff) and two for the concentration response curves (i.e. IKACh,norm,50mV and IKACh,norm,-100mV), were compared to their corresponding measured features fm,i (Fig 5B–5D and 5F–5H):
E2=∑i=18(||fm,i−fs,i||2||fs,i||2)2+(1−max(O))+((1−max50mV(U2))2+(1−max−100mV(U1))2)0μM+((1−max50mV(B2))2+(1−max−100mV(B1))2+(1−(B1+B2)2))10μM
(11)
Other components of the cost function were used to constrain the behavior of the model at specific ACh concentration and Vm, and to ensure high open probability of the channel. Without bound ACh and Vm of +50 mV, the state U2 was forced to be maximal. Respectively, U1 was forced to be maximal without bound ACh and a transmembrane voltage of -100 mV. Equivalently, with bound ACh, state B2 and B1 were forced to be maximal at -100 mV and 50 mV, respectively. Further, the sum of B1 and B2 was forced to be maximal with bound ACh.
For simulations of single cell electrophysiology, the new receptor-channel model was integrated in the publically available model of human SN cells [6]. The formulations for the effect of ACh on ICaL and If were left unaltered throughout the experiments. The model was first exported to Matlab using OpenCOR (www.opencor.ws) and then modified accordingly. Numerical integration was performed by using the integrated ode15s formulation provided by Matlab. We measured BCL to characterize the rate of spontaneous activation of the simulated SN cell at varying concentration of ACh and Vshift. The maximum conductivity gK,ACh,max was set 0.0022 μS to reproduce previously published heart rate slowing in the presence of 0−0.1 μM ACh [12]. Furthermore, we measured the APD at 90% repolarization (APD90) to characterize the cardiac AP.
The stochastic parameterization yielded the model parameters (Table 3). In comparison to the parameterization of our previous model [5], the fit error of the features from the activation and deactivation protocols was reduced from 1.5 to 0.68 despite the additional error terms. Respectively the squared fit error for each feature is; Con: 0.06, Coff: 0.025, Aon: 0.11, Aoff: 0.014, ton: 0.04, toff: 0.18, IKACh,norm,50mV: 0.03, IKACh,norm,-100mV: 0.003. The total error of the features is equal to the square root of the sum of the respective squared fit errors. The corresponding modeled and measured current traces of the voltage protocols are shown in S2 Fig.
|
10.1371/journal.pntd.0000431 | A Comparative Study of the Spatial Distribution of Schistosomiasis in Mali in 1984–1989 and 2004–2006 | We investigated changes in the spatial distribution of schistosomiasis in Mali following a decade of donor-funded control and a further 12 years without control.
National pre-intervention cross-sectional schistosomiasis surveys were conducted in Mali in 1984–1989 (in communities) and again in 2004–2006 (in schools). Bayesian geostatistical models were built separately for each time period and on the datasets combined across time periods. In the former, data from one period were used to predict prevalence of schistosome infections for the other period, and in the latter, the models were used to determine whether spatial autocorrelation and covariate effects were consistent across periods. Schistosoma haematobium prevalence was 25.7% in 1984–1989 and 38.3% in 2004–2006; S. mansoni prevalence was 7.4% in 1984–1989 and 6.7% in 2004–2006 (note the models showed no significant difference in mean prevalence of either infection between time periods). Prevalence of both infections showed a focal spatial pattern and negative associations with distance from perennial waterbodies, which was consistent across time periods. Spatial models developed using 1984–1989 data were able to predict the distributions of both schistosome species in 2004–2006 (area under the receiver operating characteristic curve was typically >0.7) and vice versa.
A decade after the apparently successful conclusion of a donor-funded schistosomiasis control programme from 1982–1992, national prevalence of schistosomiasis had rebounded to pre-intervention levels. Clusters of schistosome infections occurred in generally the same areas accross time periods, although the precise locations varied. To achieve long-term control, it is essential to plan for sustainability of ongoing interventions, including stengthening endemic country health systems.
| Geostatistical maps are increasingly being used to plan neglected tropical disease control programmes. We investigated the spatial distribution of schistosomiasis in Mali prior to implementation of national donor-funded mass chemotherapy programmes using data from 1984–1989 and 2004–2006. The 2004–2006 dataset was collected after 10 years of schistosomiasis control followed by 12 years of no control. We found that national prevalence of Schistosoma haematobium and S. mansoni was not significantly different in 2004–2006 compared to 1984–1989 and that the spatial distribution of both infections was similar in both time periods, to the extent that models built on data from one time period could accurately predict the spatial distribution of prevalence of infection in the other time period. This has two main implications: that historic data can be used, in the first instance, to plan contemporary control programmes due to the stability of the spatial distribution of schistosomiasis; and that a decade of donor-funded mass distribution of praziquantel has had no discernable impact on the burden of schistosomiasis in subsequent generations of Malians, probably due to rapid reinfection.
| Mali was one of the first countries in sub-Saharan Africa to initiate a national schistosomiasis control programme. Control efforts started regionally in 1978 in Dogon Country (region of Mopti) after the construction of small dams for growing vegetables, and became a national programme in 1982. During the first 10 years, the programme was run by the Malian Ministry of Health in partnership with the World Health Organization and the German Technical Cooperation (Deutsche Gesellschaft für Technische Zusammenarbeit, GTZ) [1]. Parasitological surveys followed by mass treatment of the entire population in target areas were conducted by a central team from Bamako. Additionally, in selected areas, identification of infected individuals and case treatment was implemented. The control programme was intensively focused on two major endemic areas: Office du Niger (irrigation area) and in the area around Bandiagara in the Plateau Dogon (small dams area). Initial evaluation (1–3 years after intervention) showed reductions in both prevalence of infection and prevalence of heavy-intensity infections (>50 eggs/10 ml urine for Schistosoma haematobium and >100 eggs/gram stool for S. mansoni). For S. haematobium, prevalence of infection was reduced from 58.9 to 26.8% and that of heavy infections from 18.4 to 3.8%, whereas for S. mansoni, prevalence of infection was only reduced from 49.0 to 48.1% and that of heavy infections from 10.6 to 8.9% [2]. Estimated impact of the intervention varied by intervention approach, ecological zone and time to follow up (1–3 years).
GTZ support for the programme ceased in 1992, with the government taking over financial responsibility. However, lack of resources led to control activities being considerably reduced and the implications of this for infection levels were not assessed in the immediate post treatment period. From 1998, a new, decentralised control programme was approved by the Ministry of Health but, due to lacking continuous financial support from the government, many planned activities were not implemented. In 2004, a new initiative to recommence national control activities was established with support from the Schistosomiasis Control Initiative (SCI; http://www.schisto.org). Again the main intervention strategy was mass treatment with praziquantel, with a particular focus on treating school-age children [3].
The potential of using risk mapping to describe the spatial patterns of infections is now well-established, and has been demonstrated for a range of diseases including malaria [4],[5], schistosomiasis [6], Loa loa filariasis [7] and lymphatic filariasis [8]. The combination of geographical information systems (GIS), remote sensing and geostatistics has led to an increase in the understanding of the spatial epidemiology of infectious diseases, the prediction of occurrence, and the targeting of large-scale control programmes. For example, Bayesian geostatistical modelling is being used increasingly to predict spatial patterns of human schistosomiasis in Africa [9],[10],[11],[12],[13].
Much of this work to date has used data from a single geographical area at a single point in time to develop predictions for similar locations. Preliminary work has investigated the spatial extent to which risk models can be reliably extrapolated [14] but it remains unclear the extent to which models based on data from one area can be extrapolated temporally. This is particularly important in determining whether control programmes can be spatially targeted on the basis of historic data, or whether it is necessary to conduct new surveys (which are expensive and time consuming) to define the spatial distribution of disease. This issue is especially relevant in the context of the dramatic up-scaling of disease control interventions and the need for survey data to target suites of alternative interventions.
In this paper, we use unique data on schistosome infections, available from two nationwide surveys conducted in Mali, the first undertaken during the 1980s prior to the implementation of the GTZ-supported national control programme and the second between 2004–2006, 12 years after this programme had ceased and prior to implementation of the SCI-supported programme. We aim to determine whether the overall prevalence and spatial distribution of schistosomiasis in Mali is different in 2004–2006 compared to the 1980s and to determine whether the spatial distribution, including covariate relationships with environmental variables and parameters that describe the spatial dependence structure (i.e. clustering), have changed in Mali over the last two decades.
A nationwide survey was carried out between May 1984 and May 1989 prior to implementation of the GTZ-supported programme (see Traoré et al. [15] for further details). In brief, villages were selected using a three-stage sampling approach: two to three districts were randomly selected in each province, then three to five arrondissements (sub-districts) were randomly selected in each district, and five villages were randomly selected in each arrondissement. In each village, individuals were randomly selected to provide urine (200 individuals) and stool samples (150 individuals). For each individual, a single urine slide (for diagnosis of S. haematobium infection by filtration method), and two Kato-Katz slides prepared from a single faecal sample (for diagnosis of S. mansoni) were examined microscopically using standard methods. While egg counts were done, only data on the number tested and proportion positive (i.e. with one or more eggs) in a given location were available for the current study. Longitude and latitude co-ordinates of each village were identified during the current study from a national village GIS database (http://www.who.int/health_mapping/tools/healthmapper/en/); of the 323 villages surveyed we were able to geo-reference 300 villages, from which data were available on 52,104 individuals.
A more recent nationwide survey was conducted in 194 schools (including 15,051 school-aged children) between December 2004 and May 2006. Ethical approval for these surveys was obtained from St. Mary's Hospital Research Ethics Committee UK and the National Public Health Research Institute's (INRSP) scientific committee in Mali. All data collection activities were carefully explained to, and oral consent was obtained from traditional authorities in the village (the village head and the elders), the schoolmaster, the representative of the pupils' parents and the local health authorities. Child participants were given an explanation of the data collection activities and were free not to participate if they so chose. Written consent was not obtained and oral consent was not specifically documented because the survey was considered by the UK and Malian ethical committees as part of the monitoring and evaluation of routine health activities carried out by the Malian Ministry of Health's national schistosomiasis control programme.
Survey protocols (available on request) instructed survey teams to select 30 boys and 30 girls per school using systematic random sampling. Schools were selected to maximise geographical coverage of the study area; all parts of Mali excluding the northern desert and far eastern regions, where transmission is known not to occur [16], were included in the survey. This was done in a GIS (ArcView 9.2, ESRI, Redlands, CA) by overlaying a 1 decimal degree squared grid over the country. The locations of communities in Mali were obtained from the aforementioned national village database. Communities were selected using simple random selection from each grid cell and, if more than one school was present in a town or village, a school was sampled on arrival using simple random selection. The selected children were assembled and asked to provide a urine and stool sample. For each child, a single urine slide and two Kato-Katz slides prepared from a single faecal sample were examined microscopically as described above. Numbers of eggs of S. haematobium and S. mansoni in each child's sample were recorded on paper forms, in addition to the geographic location of the school (determined using a hand-held global positioning system). All school and individual data were transferred to a Microsoft Access database.
For the current study, numbers tested and positive (defined as one or more eggs for each species of schistosome) were calculated for each survey location. School or community-level raw prevalence was then plotted in the GIS. Electronic data for land surface temperature (LST) and normalised difference vegetation index (NDVI) were obtained from the National Oceanographic and Atmospheric Administration's (NOAA) Advanced Very High Radiometer (AVHRR; see Hay et al. [17] for details on these datasets) and the location of large perennial waterbodies was obtained from the Food and Agriculture Organization of the United Nations (FAO-GIS). Values for LST, NDVI and distance to the nearest perennial water body (DPWB) were calculated in the GIS for each survey location.
Multivariable logistic regression models were developed for each species of schistosome and each of the two survey periods in a frequentist statistical software package (Stata version 10.1, Stata corporation, College Station, TX). Prelimary results were similar for each species of schistosome and each study period. A quadratic association between LST and prevalence was assessed and was found to be significant and DPWB was also significantly and negatively associated with prevalence. NDVI was not found to be significantly associated with prevalence in the preliminary multivariable models and was excluded from further analysis. Therefore, it was decided to enter LST (in quadratic form) and DPWB as covariates into the final spatial models. Bayesian geostatistical models, developed in WinBUGS 1.4 (Medical Research Council, Cambridge, UK and Imperial College London, UK), were identically structured for each species of schistosome and each study period. Statistical notation is presented in Text S1.
Three chains of the models were run consecutively. A burn-in of 1,000 iterations was allowed, followed by 10,000 iterations where values for the intercept and coefficients were stored. Diagnostic tests for convergence of the stored variables were undertaken, including visual examination of history and density plots of the three chains. Convergence was successfully achieved after 10,000 iterations in each model and the posterior distributions of model parameters were combined across the three chains and summarized using descriptive statistics. Geostatistical prediction across Mali was done in WinBUGS using the spatial.unipred command [18].
To compare predictions accross time periods, the 1984–1989 model was used to predict infection prevalence at the 2004–2006 survey locations and vice versa, for both S. haematobium and S. mansoni. The predicted prevalence was compared to the observed prevalence, dichotomised at 50, 20, 10 and 0% (to assess predictive performance relative to different observed prevalence thresholds, including the World Health Organisation-recommended thresholds for annual and biannual mass chemotherapy of 50% and 10% respectively). The diagnostic test evaluation statistic, area under the curve (AUC) of the receiver operating characteristic, was used for the comparison. An AUC value of >0.7 was taken to indicate acceptable predictive performance [19].
A stationary model is one where the parameters that define the spatial dependence structure are the same for the two time periods and a non-stationary model is one where the parameters are different (note we refer to stationarity across time periods, not different parts of the study area). Models were developed using the combined datasets, including with different intercepts for each time period and: 1) different coefficients, spatial dependence parameters and random effects (i.e. assuming separate sub-models for each time period); 2) the same coefficients but different spatial dependence parameters and random effects (i.e. allowing the sub-models to have common covariate effects); 3) the same coefficients and spatial dependence parameters but different random effects (i.e. allowing common covariate effects and stationary spatial dependence structures, but separate predicted risk surfaces); and 4) the same coefficients, spatial dependence parameters and random effects (i.e. a single model giving an overall predicted risk surface across the two time periods). Models 1 and 2 were non-stationary models and models 3 and 4 were stationary models. Statistical notation is presented in Text S2.
The best-fitting model (of 1–4) was selected using the deviance information criterion (DIC). An additional comparison of the spatial distribution of schistosomiasis accross time periods was done by subtracting predicted prevalence from the best-fitting S. haematobium and S. mansoni models in 2004–2006 from predicted prevalence in 1984–1989.
The national prevalence of infection with S. haematobium in 1984–1989 was 25.7% (range, 0.0–93.0%; 95% CI 25.3, 26.0%) and in 2004–2006 was 38.3% (range, 0.0–99.0%; 95% CI 37.5, 39.1%), whereas for S. mansoni, prevalence in 1984–1989 was 7.4% (range, 0.0–77.8%; 95% CI 7.1, 7.6%) and in 2004–2006 was 6.7% (range, 0.0–94.9%; 95% CI 6.3, 7.1%; note, CIs are binomial exact CIs which do not account for the clustered survey design or spatial autocorrelation – see the section on comparative models for significance testing of prevalence in 1984–1989 versus 2004–2006). Maps of community (1984–1989) and school (2004–2006) level prevalence (Figures 1 and 2) show that the data from 1984–1989 had a less uniform geographical distribution than the data from 2004–2006. High prevalence of infection with S. haematobium was widespread in Mali in both survey periods, whereas for S. mansoni, both surveys indicated small clusters of high infection prevalence in central Mali (Macina and Niono districts in the Office du Niger irrigation area) and southwestern areas (e.g. Kati district on the Niger River and Kita and Bafoulabé districts on the Senegal River), but zero or very low prevalence of infection throughout the rest of the country.
The Bayesian geostatistical models for each time period are presented in Table 1. Note that the odds ratios are on the same scale for each variable, which were standardised to have a mean of zero and standard deviation of one. DPWB was significantly and negatively associated with each outcome, with very similar odds ratios for all four models. The quadratic term for LST was not significant in any of the models, where significance is defined by a 95% posterior interval that excludes one (note, outputs of Bayesian models are distributions termed posterior distributions that describe the probability associated with each of a range of plausible values for the variable being estimated). Phi (), which indicates the rate of decay of spatial correlation (with a bigger indicative of smaller clusters) varied from 1.68 to 9.02 for S. haematobium and S. mansoni in 2004–2006. S. haematobium clusters were, therefore, generally larger than S. mansoni clusters. For both types of infection, the sill was lower in 1984–1989 than in 2004–2006, indicating a stronger tendency towards spatial clustering in the latter time period.
Models developed on 1984–1989 and 2004–2006 data were generally able to discriminate infection prevalence for the other dataset to an acceptable level (Table 2). For S. haematobium, models tended to perform better when discriminating at lower prevalence thresholds (present versus absent, <10% versus ≥10%), while for S. mansoni, models tended to perform better at high prevalence thresholds (<50% versus ≥50%). The only comparison that gave an AUC <0.7, the acceptability criterion, was for prediction of S. mansoni presence (prevalence >0%) in 1984–1989.
The deviance information criterion for models 1–4, for S. haematobium and S. mansoni, are presented in Table 3. For S. haematobium, the model with the lowest DIC (indicating the model with the best compromise between model fit and parsimony) was model 2 (Table 4), with common covariate effects but a non-stationary spatial dependence structure across time periods. For S. mansoni, the model with the lowest DIC was model 3 (Table 5), with common covariate effects and a stationary spatial dependence structure across time periods. As for the period-specific models, prevalence of both infections was negatively associated with increasing DPWB and was not significantly associated with LST. In the non-stationary model for S. haematobium (Table 4), the sill was lower for 1984–1989 than for 2004–2006, again indicating greater clustering in the latter time period, and the rates of decay of spatial correlation, phi, were similar for the two time periods. The overlapping 95% posterior interval limits for the 1984–1989 and 2004–2006 intercepts in both the S. haematobium and S. mansoni models suggest that overall (mean) prevalence was not significantly different across time periods for either species of schistosome.
Spatial predictions (showing the mean of the posterior distributions for predicted prevalence) based on the best model for each type of schistosome infection are presented in Figures 3 and 4. In 2004–2006, S. haematobium occurred in large clusters in a mid-latitudinal band from western to central Mali and low predicted prevalence was apparent in both southern and northern latitudinal bands (Figure 3B). In 1984–1989 (Figure 3A), the pattern was similar but more fragmented. The prediction maps for S. mansoni (Figure 4) were remarkably similar to each other, with infection limited to small high-prevalence clusters in central and southwestern regions, althought the clusters occurred in slightly different locations.
Comparative maps show predicted prevalence in 1984–1989 subtracted from predicted prevalence in 2004–2006, using the best-fitting models (Figure 5). Most areas of both maps had an estimated difference of <10% in predicted prevalence between the two periods. However, there were some areas on both maps that had an estimated difference of >20% in predicted prevalence; for S. haematobium, higher predicted prevalence in 2004–2006 mainly occurred in central and western regions and lower predicted prevalence was mainly along the Niger river and in southwestern regions; for S. mansoni, differences coincided with the locations of the small high-prevalence foci in central and southwestern regions because the precise location of these clusters varied somewhat between the study periods.
Despite differences in survey design and study population between the time periods, this study demonstrated remarkable similarities in the spatial distribution of prevalence of infection with S. haematobium and S. mansoni in Mali between 1984–1989 and 2004–2006. While clusters of infection occurred in generally the same area of the country, the precise location did vary slightly between the two time periods. Nonetheless, our analysis of predictive performance of models across time periods suggests it may be possible, in the first instance, to use historical data to predict contemporary distributions at national scales (assuming a stable climate and an absence of new, large water resource development projects, both of which should be investigated). It is perhaps not surprising that the statistical associations between prevalence and DPWB did not vary between the study periods as the essential biology of schistosome infections is unlikely to have changed, but it is interesting that the spatial dependence structure was different (i.e. non-stationary) for S. haematobium between the time periods. Possible reasons for non-stationary spatial variation of S. haematobium can be broadly categorised into those related to the different sampling strategies used, and those related to changing epidemiology between the two study periods.
Regarding the sampling strategies, the data were based on different sample locations, collected for different purposes and from different populations. The data from 1984–1989 were collected from the general population including adults, whilst the 2004–2006 data were from school-aged children. Age-stratified prevalence and intensity of S. haematobium infections in Mali have been reported [15] but individual or location-specific, age-stratified prevalence data were not available in the current study, which can be seen as its major limitation. However, previous analyses (including an analysis of the same 1984–1989 dataset used in this report) have shown that, while prevalence in school-aged children is generally higher than in the adult population, there is a consistent relationship between the prevalence in the two populations such that prevalence in one can be used to predict prevalence in the other [15],[20]. The overall prevalence of S. haematobium in 1984–1989, 25.7%, corresponds to an age-adjusted prevalence of approximately 36% in children aged 7–14 years [15], which is very similar to the prevalence in school-aged children (38.3%) in 2004–2006.
The 1984–1989 surveys had a less uniform geographical distribution than the 2004–2006 surveys, which is not surprising given that the 1984–1989 surveys were not explicitly designed with subsequent spatial analysis in mind, whereas uniform geographical coverage was an aim of the survey design for the 2004–2006 study to facilitate spatial analysis. Investigation of the impact of different sampling strategies on observed spatial correlation is an area of future research.
Factors potentially related to changing epidemiology include desertification, urban growth and rural-urban migration [21],[22], changing demographic and socioeconomic characteristics of the population, long-term impacts of interventions on transmission and implementation of water resource development projects such as irrigation schemes, large dams and reservoirs [23],[24],[25]. These factors can influence not only stationarity of spatial variation but any differences observed in the location of spatial disease clusters. The earlier, GTZ-supported control programme focussed on specific, perceived high-risk areas of the country, with treatment coverage highest in Bandiagara, Office du Niger, Baguinéda and Sélingué. It might be suggested that spatial variation in changes in prevalence (Figure 5) could relate to uneven geographical coverage of the intervention, but the main intervention areas do not correspond consistently to those where prevalence was lower in 2004–06 than 1984–89.
In addition to the limitation of different survey designs between periods, we were not able to compare spatial variation in intensity of infection between time periods because location-specific mean egg counts were not available from the 1984–1989 surveys. Maps of intensity would be useful for determining any changes in transmission across the periods. Examination of a single urine slide or single stool sample as a diagnostic approach results in sub-optimal sensitivity and this will also have affected the accuracy of our maps. We also did not incorporate anisotropy (where the spatial correlation structure varies by direction) or non-stationary spatial variation between different parts of the country, within each time period; these are future potential refinements of the models. We should also point out that the model predictions are distributions and here we have only presented the posterior mean. Examination of the full posterior distribution of predicted prevalence enables assessment of uncertainties arising from sampling and measurement error (including in the model covariates). We have recently described how an understading of these uncertainties can assist decion making in schistosomiasis control programme planning [9].
Our results show that, while there were differences in the raw data, the overall prevalence of neither S. haematobium nor S. mansoni was significantly different between the time periods, despite ten years of donor-funded schistosomiasis control throughout the 1980s and early 1990s. The most likely explanation is that, in the absence of ongoing exposure reduction measures, re-infection with schistosomes following chemotherapy inevitably occurred. In endemic settings this is often apparent within 24 months [26],[27]. Rates of infection and re-infection are generally similar among different age groups, although older people typically reacquire schistosome infection at slower rates than younger people [28]. Problems of re-infection were acknowledged by the managers of the 1980s control programme and this was reflected in the goal to reduce morbidity associated with infection in the treated communities (which was successfully demonstrated in some areas [29]) rather than transmission. The result was a predictable failure of the national programme to have a lasting impact on the burden of schistosomiasis in subsequent generations of Malians.
One of the most important conclusions arising from the current work is that it is essential to develop a sustainability strategy to ensure ongoing benefits from the current national control programme. Recognising this fact, SCI has developed a sustainability plan which is outlined in Fenwick et al. [30]. Briefly, sustainability is based on initially using annual mass chemotherapy in areas with prevalence ≥50%, or biannual mass chemotherapy where prevalence is ≥10% and <50%, to rapidly reduce prevalence and intensity of infection. Then, when prevalence reaches <10% (after up to four rounds of treatment, depending on levels of transmission), the Malian government plans to make treatments available in health facilities, carry out regular surveys and target treatment in schools if the prevalence rises above 10%. Sustainability also depends on developing the Malian health system and integrating schistosomiasis control with routine health care delivery [31]. Improved water sanitation and health education could be promoted for sustainable control [32], snail control could be revisited and schistosomiasis vaccines might also have a future role [33].
The maps presented here can be used to target what are likely to be more limited national resources in the longer term to the highest-risk areas, where they will have the greatest impact on infection, morbidity, and (hopefully) transmission. The current move towards integration of control of neglected tropical diseases means that the government may have the opportunity to implement a cost effective control programme encompassing schistosomiasis, soil transmitted helminth infections, lymphatic filariasis, river blindness and trachoma. It is clear that a commitment from the Malian government and international donors for substantial resources is required long into the future, or alternative strategies need to be found, if control of schistosomiasis transmission in Mali is to be achieved.
|
10.1371/journal.ppat.1000921 | Effective, Broad Spectrum Control of Virulent Bacterial Infections Using Cationic DNA Liposome Complexes Combined with Bacterial Antigens | Protection against virulent pathogens that cause acute, fatal disease is often hampered by development of microbial resistance to traditional chemotherapeutics. Further, most successful pathogens possess an array of immune evasion strategies to avoid detection and elimination by the host. Development of novel, immunomodulatory prophylaxes that target the host immune system, rather than the invading microbe, could serve as effective alternatives to traditional chemotherapies. Here we describe the development and mechanism of a novel pan-anti-bacterial prophylaxis. Using cationic liposome non-coding DNA complexes (CLDC) mixed with crude F. tularensis membrane protein fractions (MPF), we demonstrate control of virulent F. tularensis infection in vitro and in vivo. CLDC+MPF inhibited bacterial replication in primary human and murine macrophages in vitro. Control of infection in macrophages was mediated by both reactive nitrogen species (RNS) and reactive oxygen species (ROS) in mouse cells, and ROS in human cells. Importantly, mice treated with CLDC+MPF 3 days prior to challenge survived lethal intranasal infection with virulent F. tularensis. Similarly to in vitro observations, in vivo protection was dependent on the presence of RNS and ROS. Lastly, CLDC+MPF was also effective at controlling infections with Yersinia pestis, Burkholderia pseudomallei and Brucella abortus. Thus, CLDC+MPF represents a novel prophylaxis to protect against multiple, highly virulent pathogens.
| Conventional treatment of bacterial infections typically includes administration of antibiotics. However, many pathogens have developed spontaneous resistance to commonly used antibiotics. Development of new compounds that stimulate the host immune system to directly kill bacteria by mechanisms different from those utilized by antibiotics may serve as effective alternatives to antibiotic therapy. In this report, we describe a novel compound capable of controlling infections mediated by different, unrelated bacteria via the induction of host derived reactive oxygen and reactive nitrogen species. This compound is comprised of cationic liposome DNA complexes (CLDC) and crude membrane preparations (MPF) obtained from attenuated Francisella tularensis Live Vaccine Strain (LVS). Pretreatment of primary mouse or human cells limited replication of virulent F. tularensis, Burkholderia pseudomallei, Yersinia pestis and Brucella abortus in vitro. CLDC+MPF was also effective for controlling lethal pulmonary infections with virulent F. tularensis. Thus, CLDC+MPF represents a novel antimicrobial for treatment of lethal, acute, bacterial infections.
| Historically, control of bacterial infections has been dependent on administration of antibiotics. However, with many acute, lethal, bacterial infections, e.g. those mediated by Francisella tularensis, Yersinia pestis and Staphylococcus aureus, diagnosis and timely administration of appropriate antibiotics represents a significant hurdle in successful treatment of disease mediated by these pathogens. Further, mass administration of prophylactic antibiotics in an outbreak situation may result in the generation of antibiotic resistant strains, rendering this treatment ineffective for both ongoing infections and future outbreaks. Thus, there is a need for novel, broad spectrum, prophylaxis against highly pathogenic bacterial infections.
F. tularensis is a Gram negative, facultative intracellular bacterium and the causative agent of Tularemia. F. tularensis is extremely infectious, capable of causing acute, lethal, disease following inhalation of as few as 10–15 bacteria [1], [2]. Currently, there is no vaccine approved for use against F. tularensis. Although antibiotic therapy can successfully treat pneumonic Tularemia, therapy must be initiated within the first few days following the onset of symptoms when individuals are often unaware of the severity of their infection [3]. Furthermore, treatment with antibiotics can fail to adequately clear F. tularensis, resulting in recrudescence of infection once antibiotic therapy ends [4], [5], [6].
A number of studies have described development of novel anti-microbials that target the host immune response rather than the invading pathogen [7], [8], [9]. These immunotherapeutics target host pathways which either directly activate effector cells or relieve pathogen induced suppression of host killing mechanisms, resulting in control and elimination of a wide variety of microorganisms. In the case of intracellular pathogens such as F. tularensis, targeting host effector mechanisms is appealing since some antibiotics preferred for treatment of Tularemia, e.g. gentamicin, poorly permeate the host cell and therefore fail to reach the targeted organism. Activation of host effector cells capable of killing intracellular pathogens with novel immunotherapeutics or prophylaxes represents a viable alternative, or supplement, to exiting chemotherapy.
In this report we describe a novel anti-microbial comprised of cationic liposome DNA complexes (CLDC) and crude membrane protein fraction (MPF) derived from attenuated F. tularensis strain LVS. CLDC+MPF effectively controlled in vivo and in vitro infections of virulent F. tularensis strain SchuS4 in mouse and human cells, respectively. The combined delivery of CLDC and MPF was critical for mediating this protection, since treatment with CLDC or MPF alone failed to attenuate F. tularensis replication and pathogenicity. The dramatic control of F. tularensis infection mediated by CLDC+MPF was dependent on stimulation of both reactive oxygen and nitrogen species (ROS and RNS, respectively) in vivo and in vitro. Finally, we demonstrate that CLDC+MPF was also an effective antimicrobial against three other important bacteria, Burkholderia pseudomallei, Yersinia pestis and Brucella abortus. Thus, data presented herein represents an important step toward development of novel, efficacious, broad spectrum, antimicrobial therapy directed against highly pathogenic microbes.
To establish the anti-microbial potential of CLDC+MPF we first examined the effect of this compound on the infection and replication of SchuS4 in mouse and human macrophages. Cells were treated with either 5% dextrose water (D5W;untreated), CLDC, MPF or CLDC+MPF approximately 18 h prior to SchuS4 infection. At the indicated time points, viable intracellular bacteria were enumerated. Macrophages treated with MPF or CLDC alone failed to control SchuS4 replication (Table 1 and 2). Additionally, treatment of mouse macrophages with MPF or CLDC alone exacerbated SchuS4 replication in these cells (Table 1). In contrast to cells treated with the individual components or untreated controls, cells pre-treated with CLDC+MPF had significantly fewer intracellular bacteria (Table 1 and 2). It was possible that treatment with CLDC+MPF induced cell death which in turn resulted in smaller numbers of intracellular SchuS4. However, staining by trypan blue revealed that, similar to untreated controls, greater than 90% of CLDC+MPF treated macrophages were viable at the time of infection (Figure S1). Thus, generalized cell death could not account for the reduction in bacterial loads following CLDC+MPF treatment. Together this data shows that while CLDC+MPF did not limit uptake of SchuS4, it did control replication of the intracellular bacterium.
Intracellular replication of SchuS4 in macrophages is dependent on their ability to escape the phagolysosome within the first hour of infection [10]. Thus, although CLDC+MPF did not appear to alter the initial uptake of SchuS4 by macrophages, it was possible that this mixture inhibited the ability of SchuS4 to escape the phagolysosome subsequently resulting in killing of the bacterium. Phagosomal escape by SchuS4 can be measured by microscopy by the loss of LAMP-1 colocalization with the bacterium following macrophage phagocytosis [10]. To assess the effect CLDC, MPF and CLDC+MPF had on SchuS4 phagosomal escape we next examined co-localization of SchuS4 with LAMP-1 in treated mouse and human macrophages 24 h after infection by microscopy. As described above, CLDC+MPF did not significantly alter the number of bacteria phagocytosed by either mouse or human macrophages compared to untreated controls (Figure 1B and 2B, respectively). Further, CLDC+MPF treatment did not inhibit the ability of SchuS4 to escape into the cytosol of infected cells as evident by the absence of SchuS4 co-localization with LAMP-1 4 hours after infection in all treatment groups (Figure 1A and 2A). In contrast, CLDC+MPF significantly reduced the number of infected human and mouse macrophages 24 h after infection (Figures 1A, C and 2A, C, respectively). We also examined if CLDC+MPF mediated killing of SchuS4 by electron microscopy. As expected, in untreated cells SchuS4 was present as intact bacteria in the cytosol (Figure S2). In contrast, bacteria present in CLDC+MPF treated macrophages were largely degraded with few to no intact bacteria present in the cytosol (Figure S2).
Following entry and escape into the host cytosol, SchuS4 undergoes a lag period of approximately 4 hours prior to initiation of replication. Replication in the primary infected cell then commences and continues over a 12–18 hour time period [11]. Thus, we next determined at what point after infection CLDC+MPF could restrict intracellular SchuS4 growth in macrophages, Cultures treated with CLDC+MPF 4 hours after infection had significantly fewer infected cells compared to untreated controls (p<0.05) (Figure S3). In contrast, macrophages treated with CLDC+MPF 12 hours after SchuS4 infection failed to control bacterial replication compared to untreated controls (Figure S3). Together, these data suggested that in order for CLDC+MPF to exert its protective effects macrophages must be stimulated at time points prior to intracellular replication of SchuS4.
Stimulation of ROS and RNS in mammalian cells represents an important mechanism by which the host controls and eliminates bacterial growth, especially in the intracellular environment (as reviewed, [12]). Furthermore, ROS and RNS have been implicated as important mediators for control of attenuated strains of F. tularensis [13], [14], [15], [16]. Thus, we hypothesized that CLDC+MPF may be mediating control of SchuS4 infection in macrophages via induction of ROS and/or RNS. We first determined if MPF, CLDC, or CLDC+MPF induced expression of genes associated with oxidative stress. Mouse macrophages treated with CLDC+MPF had higher expression levels of genes associated with RNS (nitric oxide synthetase 2) and ROS (superoxide dismutase 2, NADPH oxidase 1) compared to untreated controls (Figure 3). Induction of RNS and ROS related genes was highest when cells were treated with CLDC+MPF rather than CLDC or MPF alone (Figure 3). Similarly, treatment of human cells with CLDC+MPF resulted in increased expression of genes associated with generation of ROS, e.g. phox p47, superoxide dismutase 2, NADPH oxidase, and GTP cyclohydrolase, as well as RNS, e.g. and nitric oxide synthase 2A compared to untreated cells and cells treated with MPF alone (Figure 3). Interestingly, although CLDC alone failed to control SchuS4 infection in human macrophages, cells treated with this compound in the absence of MPF had greater gene transcription for three genes involved in generation of ROS, i.e. superoxide dismutase 2, phox47 and GTP cyclohydrolase compared to cells treated with CLDC+MPF (Figure 3). We also compared induction of ROS and RNS related genes in cells treated with a known inducer of RNS, IFN-γ to CLDC+MPF. CLDC+MPF elicited higher gene transcription of nitric oxide synthetase (nos2), NADPH oxidase and superoxide dismutase (sod2) in mouse cells and GTP cyclohydrolase, nitric oxide synthetase 3 (nos3), NADPH oxidase, phox 47 and sod2 in human cells than IFN-γ (Figure S4).
Induction of both RNS and ROS is often dependent on the presence of IFN-γ, TNF-α, IFN-β and other pro-inflammatory cytokines. In correlation with the expression levels of RNS and ROS associated genes, mouse macrophages treated with CLDC+MPF secreted significantly higher concentrations of TNF-α compared to cells treated with CLDC or MPF alone (Figure 4). Addition of MPF to CLDC also increased secretion of IL-6 and IFN-β from mouse cells compared to cells treated with CLDC alone, however these differences were not significant (Figure 4).
We also monitored production of cytokines from human cells treated with MPF, CLDC or CLDC+MPF. Although no IL-12p40 or IFN-β was detected in any cell culture supernatant, CLDC+MPF induced secretion of significantly higher concentrations of IL-6 from human macrophages compared to cells treated with CLDC alone (Figure 4). Human macrophages also produced significantly higher concentrations of TNF-α in response to CLDC and CLDC+MPF compared to cells treated with D5W or MPF (Figure 4). Thus, both CLDC and CLDC+MPF elicited production of cytokines associated with induction of RNS and ROS, and in some cases, the combination of CLDC+MPF resulted in higher concentrations of these cytokines.
To determine if the induction ROS, RNS or both ROS and RNS by treatment of cells with CLDC+MPF contributed to control of SchuS4 infection, we first examined the ability of CLDC+MPF to limit SchuS4 replication 24 h after infection in macrophages obtained from mice deficient for both RNS and ROS. There was not a difference in the percentage of infected cells among untreated wild type and nos2/gp91−/− macrophages, suggesting that nos2/gp91−/− cells were not more susceptible to SchuS4 infection compared to wild type cells (Figure 5A). As previously observed, wild type macrophages treated with CLDC+MPF had significantly fewer infected macrophages compared to untreated cells (p<0.01) (Figure 5A). In contrast, CLDC+MPF failed to control SchuS4 infection in macrophages from nos2/gp91−/− mice (Figure 5A). Furthermore, pre-treatment of nos2/gp91−/− macrophages with CLDC+MPF resulted in significantly more infected macrophages 24 h after the onset of the experiment (p<0.01) (Figure 5A). Thus, CLDC+MPF mediated control of SchuS4 infection was at least partially dependent on ROS and RNS in mouse macrophages.
To determine the contribution of RNS and ROS in CLDC+MPF mediated control of SchuS4 infection in wild type mouse macrophages, we conducted additional experiments using compounds that specifically interfere with either the generation of RNS (L-NMMA) or ROS (NAC). IFN-γ has been shown to mediate killing of intracellular bacteria, including Francisella, following stimulation of both ROS and RNS (23). Thus, macrophages pretreated with IFN-γ followed by exposure to L-NMMA or NAC served as positive controls for inhibition of ROS and RNS species. As expected, pretreatment of mouse and human macrophages with IFN-γ significantly reduced the number of cells infected with SchuS4 compared to untreated controls (p<0.01) (Figure 5B–E). The role of both RNS and ROS in IFN-γ mediated control of SchuS4 in mouse macrophages was confirmed by significant increases in SchuS4 infected cells following addition of either L-NMMA or NAC to IFN-γ treated cells (p<0.05) (Figure 5B–C). Similarly, addition of L-NMMA or NAC to CLDC+MPF pretreated mouse macrophages reversed the protective effect observed with this prophylaxis (Figure 5B–C). Furthermore, addition of either L-NMMA or NAC to CLDC+MPF treated cells also appeared to increase the number of infected cells over untreated controls (Figure 5B–C). Thus, in mouse macrophages CLDC+MPF mediated killing of SchuS4 was dependent on the generation of both RNS and ROS.
In contrast to mouse cells, inhibition of RNS had little effect on the ability of CLDC+MPF to control SchuS4 infection in primary human macrophages (Figure 5D). Rather, inhibition of ROS following addition of NAC significantly increased the number of infected cells among CLDC+MPF treated samples (p<0.05) (Figure 5E). This suggested that in human macrophages generation of ROS, rather than RNS, following treatment with CLDC+MPF was the primary mechanism for control of SchuS4 in human macrophages.
Macrophages are one of the primary cells targeted by SchuS4 for replication in vivo [17]. The dramatic effect CLDC+MPF had on control of intracellular growth of SchuS4 in macrophages in vitro (Figures 1–5) suggested that this compound may be an effective anti-microbial in vivo. Thus, we next assessed the ability of CLDC, MPF and CLDC+MPF to protect mice from pulmonary challenge with SchuS4. Pretreatment of mice with CLDC alone failed to protect animals from SchuS4 related mortality, regardless of the route of time at which the CLDC was administered. In fact, administration of CLDC within 48 hours prior to infection by any route exacerbated disease, as indicated by an increase in the mean time to death of treated animals compared to untreated controls (Table 3). In contrast, administration of CLDC intravenously, intranasally or intraperitoneally 72 hours prior to infection modestly increased the mean time to death (∼0.2 days) (Table 3). Thus, as observed among in vitro stimulated macrophages, CLDC did alone did not protect animals for death following SchuS4 infection.
Administration of CLDC three days prior to challenge resulted in a minor increase in mean time to death. Furthermore, treatment of animals prior to that time point resulted in exacerbated disease. Thus, we chose to examine the protective efficacy of CLDC+MPF in animals treated three days prior to infection. The results of these studies are depicted in Table 4. Administration of CLDC+MPF intranasally failed to protect or improve survival of SchuS4 infections. Twenty percent of animals treated with CLDC+MPF subcutaneously or intraperitoneally survived SchuS4 infection. Additionally, animals treated via these routes that succumbed to infection survived longer than untreated controls. Animals treated with CLDC+MPF intravenously had the greatest survival rate, with approximately 50% of these animals surviving SchuS4 infection (Table 4 and Figure 6A). Intravenous injection of either CLDC or MPF failed to protect animals from succumbing to infection (Table 4). Furthermore, intravenous injection of CLDC+MPF 1,2 or 7 days prior to infection failed to protect animals against SchuS4 (data not shown). Thus, in vivo protection against SchuS4 required CLDC and MPF delivered three days prior to infection.
Previous studies have shown that LVS LPS can protect mice from lethal LVS infections [18], [19], [20], [21], [22]. MPF contains LVS LPS. Thus, we postulated that LVS LPS present in MPF may represent an important immunogen for conferring the protection observed in CLDC+MPF treated mice. To test this hypothesis, mice were injected intravenously with LPS purified from LVS alone or in combination with CLDC. Surprisingly, neither LVS LPS alone nor LVS LPS in CLDC protected mice from lethal SchuS4 infection (Table 4). However, LVS LPS in CLDC did increase the mean time to death by approximately 2 days compared to untreated controls (Table 4). This suggested that LVS LPS combined with CLDC was able to stimulate the host immune response for minimal control of SchuS4 infection. Although LPS from LVS and SchuS4 do not elicit strong production of pro-inflammatory cytokines, it was possible that there might be other differences in the immunostimulating potential of these two LPS molecules that would only be revealed in vivo [23], [24], [25]. Thus, we also compared protective efficacy of SchuS4 LPS to protect against SchuS4 infection. Similar to LVS LPS, mice treated with SchuS4 LPS were not protected from death (Table 4). However, inclusion of CLDC did modestly increase the mean time to death in SchuS4+CLDC treated animals (Table 4). Together, these data suggest that Francisella LPS was not the major component of MPF mediating protection against SchuS4 infection in CLDC+MPF treated animals.
Given the importance of both ROS and RNS in CLDC+MPF mediated control of SchuS4 infection in mouse macrophages in vitro, we also assessed the role of these antimicrobial host components in vivo. Pretreatment of mice with CLDC+MPF protected significantly more wild type animals from SchuS4 compared to untreated controls (p = 0.0027) (Figure 6B). In contrast to the protection observed in wild type animals, CLDC+MPF did not significantly increase the number of surviving nos2/gp91−/− compared to untreated controls (p = 0.3173) (Figure 6B). Furthermore, untreated nos2/gp91−/− mice were not more susceptible to SchuS4 infection compared to untreated wild type mice. This suggested that the lack of protection observed in CLDC+MPF treated nos2/gp91−/− was not due to inherent lack of resistance to F. tularensis in these animals. Rather, our data suggests that in the absence of intact pathways for generation of ROS and RNS CLDC+MPF activates cells that render them more susceptible to infection. These results further underscore the importance of induction of ROS and RNS in F. tularensis infections. Together this data confirmed that protection mediated by CLDC+MPF was dependent on stimulation of pathways associated with oxidative stress in vivo.
Previous experiments have shown that CLDC alone is an effective prophylaxis against B. pseudomallei and attenuated strains of Francisella [26], [27]. However, CLDC alone failed to protect against virulent Francisella and Y. pestis ([27] and CM Bosio, unpublished data). As shown above, compared to CLDC alone, CLDC+MPF provided superior protection against virulent Francisella (Table 1 and 2; Figures 1 and 2). Thus, it was possible that the combination of CLDC and antigen may also effectively control infections mediated by other, unrelated pathogens. Furthermore, the ability of CLDC or CLDC+MPF to control bacterial infections in human cells has not been examined. To assess the anti-bacterial capabilities of CLDC+MPF, we pretreated human macrophages with CLDC+MPF, infected them with B. pseudomallei, Y. pestis or B. abortus, and evaluated bacterial replication over time. As previously observed with CLDC alone in mouse cells, CLDC+MPF controlled both the number of B. pseudomallei infected human macrophages and replication of intracellular bacteria within 6 h of infection (Figure 7). Surprisingly, CLDC+MPF also reduced the number of cells infected with Y. pestis at 2 and 6 h after infection and B. abortus 24 h after infection (Figure 7). Furthermore, treatment of cells with CLDC+MPF inhibited the intracellular replication of both Y. pestis and B. abortus (Figure 7). Together this data suggests that, unlike CLDC alone, stimulation of cells with CLDC+MPF can aid in the control of several different, unrelated, virulent bacteria.
The discovery of antibiotics as broad spectrum chemotherapeutics for bacteria greatly enhanced our ability to fight off and control bacterial diseases. However, commiserate with the general use of antibiotics, bacteria have responded by developing resistance to these important and ubiquitous compounds. One strategy employed to enhance resistance against microbial pathogens is to directly stimulate the host immune response and allow natural, host mediated, killing mechanisms to control microbial infections. These novel immunotherapeutics could be used independently or in context of antibiotic therapy to aid in clearance of bacteria. In turn this would allow for a decrease the amount of time antibiotics should be administered, an increase in the time before antibiotics must be administered, and/or lower dosages of antibiotics required for complete clearance of the bacterium.
Here we describe a novel, broad spectrum antimicrobial immunoprophylaxis consisting of cationic DNA liposome complexes (CLDC) and crude membrane preparations (MPF) derived from F. tularensis that effectively limited replication of virulent F. tularensis, B. pseudomallei, Y. pestis and B. abortus in human and mouse macrophages in vitro. Importantly, administration of CLDC+MPF prior to pulmonary infection with F. tularensis also contributed to survival in mice. The mechanism of protection mediated by CLDC+MPF was, in part, dependent on the induction of reactive oxygen and nitrogen species in vivo and in vitro.
Previous reports have shown that either CpG oligodeoxynucleotides (ODN) or CLDC alone can protect against lethal infections with attenuated strains of F. tularensis, e.g. LVS [27], [28], [29]. However, neither of these therapeutics have been able to protect animals from death following SchuS4 infection [27], [29]. We confirmed and extended these results by demonstrating that regardless of the route of time CLDC was administered prior to challenge, this reagent could not decrease the number of mortalities among SchuS4 infected mice. In fact, in our hands injection of CLDC 1 or 2 days prior to SchuS4 challenge exacerbated disease as indicated as a decrease in the mean time to death (Table 3). Our results do slightly differ from those reported by Troyer et al, but are closer in agreement with the report by Rozak et al in which administration of CpG ODN less than 24 hours prior to infection exacerbated disease [27], [29]. In the experiments reported by Troyer et al., addition of DNA to cationic liposomes was performed under standard laboratory conditions with no obvious means to monitor quality control. Further, there was no indication of endotoxin levels present in the preparations of DNA. Addition of other TLR agonists to cationic liposomes enhances their immunogenicity [30]. Thus, contaminating endotoxin could have increased and/or changed the inflammatory response in the Troyer study resulting in a different mean time to death. The CLDC used in the study presented herein were produced under strict GMP laboratory conditions and underwent a battery of quality control assays prior to use to insure consistency from lot to lot. Thus, it is possible that minor variations in CLDC preparations used in the Troyer study could account for the 1.4 day extension in mean time to death among SchuS4 infected mice.
It is not clear why CLDC exacerbated infection in vitro and in vivo. One possibility is that activation of macrophages with CLDC increases their phagocytic capability without eliciting effective killing mechanisms. Indeed, immediately after infection CLDC treated cells had significantly more intracellular bacteria compared to untreated cells (p<0.05) (Table 2). MPF alone also activates macrophages and increased uptake of SchuS4. Thus, it is possible that the exacerbation of infection observed in macrophages treated with either CLDC or MPF alone may be a direct result of increased phagocytosis in the absence of effective killing.
In contrast to exacerbation of infection in mice treated with CLDC alone, we observed a small increase in the mean time to death among animals treated with CLDC alone 3 days prior to infection (Table 3). Similarly, CpG has been noted to increase the mean time to death of SchuS4 infected mice by 1 day if delivered 2 days prior to infection [29]. This suggested CLDC and CpG alone could contribute toward controlling SchuS4 infection when delivered at the appropriate time before infection. CLDC does contain CpG ODN sequences. However, these sequences are not required for CLDC to exert protective effects against infections in vivo [7]. The induction of inflammatory responses by CLDC which do not contain CpG motifs may be attributed to cellular recognition of bacterial DNA. Although TLR9 represents an important receptor for recognition of CpG motifs present in bacterial DNA, it is not the only host receptor capable of detecting prokaryotic DNA. For example, DAI is a cytosolic receptor capable of detecting bacterial DNA that does not contain CpG motifs and can trigger immune responses in mammalian cells in a TLR9 independent manner [31]. Thus, the immunogenicity of CLDC cannot be completely attributed to the presence of CpG.
One of the major components of MPF is LPS. LPS from LVS lacks properties typically associated with endotoxin, e.g. stimulation of pro-inflammatory cytokines (25). However, injection of LVS LPS can protect animals from lethal LVS infections. Thus, we tested LPS from both LVS and SchuS4 for protective efficacy against pulmonary tularemia. Surprisingly, neither LPS preparation was able to protect animals from death following intranasal SchuS4 infection (Table 4). Administration of Francisella LPS in CLDC did modestly increase the mean time to death in SchuS4 infected animals, but these animals eventually succumbed to infection (Table 4). This suggested that Francisella LPS alone is not the bacterial antigen contributing to the protective effects of CLDC+MPF. Another bacterial ligand that may contribute to the protective efficacy of CLDC+MPF is peptidoglycan or one of its precursors, i.e. muramyl dipeptide (MDP) or tracheal cytotoxin (TCT). Both MDP and TCT can contribute toward the induction of nitric oxide [32], [33], [34], [35], [36]. Interestingly, MDP typically requires cells to be primed with IFN-γ or other immunostimulants in order to induce nitric oxide [35], [36]. We have not quantitated MDP or TCT present in our MPF preparations. However, it is tempting to speculate that either or both of these compounds may contribute to the protective efficacy of CLDC+MPF.
In both the in vitro and in vivo models pre-stimulation of macrophages before replication of SchuS4 began was required for killing of bacteria. In vitro, cells stimulated with CLDC+MPF 12 h after infection failed to significantly control Francisella replication (Figure S3). Similarly, administration of CLDC+MPF less than three days prior to pulmonary challenge failed to protect animals from lethal disease (CM Bosio, unpublished data). There are several explanations for this pre-stimulation requirement. First, the kinetics of SchuS4 intracellular replication on the level of individual cells following in vivo infection has not been defined. It is possible that following infection of macrophages in vivo SchuS4 does not undergo a lag phase prior to replication such as that observed among in vitro infected macrophages. Second, SchuS4 targets multiple cell types in vivo. In addition to macrophages, this bacterium also infects dendritic cells and epithelial cells at the outset of infection [37], [38]. It is not known if CLDC+MPF activates dendritic cells and epithelial cells in the same manner we have observed in macrophages. Longer stimulation of these cells may be required for adequate priming of killing mechanisms in these and/or neighboring cells.
The third possibility may be a requirement for activation of other host effector cells or molecules. It has been suggested that NK cells contribute to eradication of F. tularensis following pulmonary infections [39]. Interestingly, intravenous injection of CLDC results in accumulation of NK cells in the lungs which peaks three days after injection [40]. Data from our laboratory suggests that SchuS4 specific NK cells present in lungs of vaccinated mice are capable of controlling SchuS4 replication in this tissue (CM Bosio and RV Anderson, unpublished data). Thus, injection of CLDC+MPF three days prior to infection may allow the accumulation of SchuS4 specific NK cells capable of restricting bacterial replication in the lungs.
The requirement for pre-stimulation may also lie in the amount of time necessary for ROS and RNS to be activated. F. tularensis encodes genes that specifically interfere with production of reactive oxygen and nitrogen species [41]. In the absence of pre-activation, RNS and ROS generation in host macrophages is efficiently impeded by virulent F. tularensis [41]. Similarly, interference with induction of RNS and ROS as a mechanism to evade killing has also been reported for Y. pestis and B. pseudomallei [42], [43]. Thus, generation of adequate RNS and ROS by hosts cells prior to infection or replication of bacteria if host cells would be required for optimal control of infections with these virulent bacteria.
An additional explanation for the necessity of pre-stimulation of host cells lies in the mechanism by which RNS and ROS are generated. Reactive oxygen and nitrogen species are most effectively produced in response to several pro-inflammatory cytokines. For example, although neither TNF-α nor IFN-β can act alone to induce release of nitric oxide or hydrogen peroxide, these cytokines act synergistically with bacterial antigens to augment production of these two antimicrobial compounds [44]. Our data demonstrate that CLDC, and in some cases CLDC+MPF, induced significantly more TNF-α and IFN-β compared to untreated cells or cells exposed to MPF alone (Figure 4). Thus, it is possible that optimal stimulation of RNS and ROS was dependent on the generation of these, and perhaps other, proinflammatory cytokines which would require additional time for generation of ROS and RNS prior to infection.
Another important observation made in the studies presented herein was that the putative roles for ROS and RNS for control of F. tularensis were dramatically different in mouse and human macrophages. As described above, CLDC+MPF mediated killing in mouse macrophages was dependent on the generation of ROS and RNS (Figure 5A–C). This is in agreement with previous reports describing the contribution of these species for control of more attenuated strains of F. tularensis [15], [41], [45]. However, in human macrophages inhibition of RNS had no effect on the antibacterial activity mediated by CLDC+MPF on F. tularensis (Figure 5D and E). Rather, generation of ROS was essential for CLDC+MPF mediated killing of F. tularensis. In the past it was commonly believed that human cells could not produce RNS. Thus, one might assume that the dependency on ROS for CLDC+MPF mediated killing in human cells was due to an inherent inability of these macrophages to generate RNS. Indeed, soon after the description of cytokine induced nitric oxide in mouse cells, investigators attempted to reproduce the phenomenon in human macrophages. However, early studies in human cells did not recapitulate the potent nitric oxide response observed in mouse macrophages [46]. This led to the hypothesis that human cells did not express the product responsible for cytokine induced nitric oxide, iNOS. Since those early studies there have been a number of reports demonstrating the presence and function of iNOS in human cells [47]. We now understand that the inability of human macrophages to produce nitric oxide in response to cytokines alone was due to the nature of the cell type (as reviewed, [48]). For example, macrophages differentiated in vitro from resting peripheral blood monocytes of normal donors generally do not express iNOS. In contrast, macrophages differentiated from monocytes obtained from donors with chronic inflammatory disorders or currently battling infection readily express iNOS [49], [50]. Thus, under the appropriate conditions human macrophages, like mouse macrophages can induce RNS. As shown in the present manuscript, IFN-γ induced RNS is capable of controlling SchuS4 replication in human macrophages, thus confirming the ability of human cells to generate effective RNS responses. Therefore, the role for ROS in CLDC+MPF mediated killing of SchuS4 in human macrophages is not due to the inability of the cells to generate RNS, but is a unique feature of CLDC+MPF stimulation of this cell type.
Understanding and identifying the different requirements and redundancy for RNS and ROS in mouse and human cells is important for several reasons. First, identification of the mechanisms by which human and mouse cells control infections with virulent bacteria is essential for monitoring the potential effectiveness of novel drugs. Second, identification of the specific killing pathways will also aid in development of other novel therapeutics. Lastly, understanding the different requirements for bacterial killing in human and mouse macrophages may reveal new pathways used by host cells to effectively combat invading pathogens. As described above, unlike stimulation with IFN-γ, CLDC+MPF revealed a difference in the mechanism by which human and mouse macrophages control virulent F. tularensis. Thus, this reagent may also serve as a useful tool to dissect the relative roles of these pathways in the effective control of infections with virulent bacteria. This in turn may result in superior immunoprophylaxis and therapeutics for infections mediated by diverse groups of bacteria.
Human blood cells were collected from anonymous volunteers under a protocol reviewed and approved by the NIH Clinical Center Institutional Review Board. Signed, informed consent was obtained from each donor acknowledging that their donation would be used for research purposes by intramural investigators throughout NIH.
Francisella tularensis strain SchuS4 was kindly provided by Jeannine Peterson, Ph.D. (Centers for Disease Control, Fort Collins, Colorado), F. tularensis strain LVS was provided by Jean Celli, Ph.D. (Rocky Mountain Laboratories, Hamilton, Montana). SchuS4 and LVS were cultured in modified Mueller-Hinton (MMH) broth at 37°C with constant shaking overnight, aliquoted into 1 ml samples, frozen at −80°C and thawed just prior to use as previously described [37]. Frozen stocks were titered by enumerating viable bacteria from serial dilutions plated on modified Mueller-Hinton agar as previously described [51], [52]. The number of viable bacteria in frozen stock vials varied less than 5% over a 10 month period.
Yersinia pestis strain 195/P expressing GFP was provided by B. Joseph Hinnebusch, Ph.D. (Rocky Mountain Laboratories, Hamilton, Montana). 195/P-GFP was cultured overnight at 21°C in BHI broth followed by subculture at 37°C as previously described [43]. Bacterial titer was estimated by optical density of the culture at 600nm. Inoculum titers were confirmed following enumeration of viable bacteria from serial dilutions plated on blood agar plates as previously described [43].
Burkholderia pseudomallei strain DD503 expressing GFP was provided by David DeShazer, Ph.D. and Mary Burtnick, Ph.D. (USAMRIID, Fort Detrick, MD and University of South Alabama, Mobile AL, respectively). DD503-GFP was cultured in LB broth overnight at 37°C. Three hours before use, DD503-GFP was diluted 1∶25 in TSBDC culture medium [53] and incubated at 37°C. Bacterial titer was estimated by optical density of the culture at 600nm. Immediately prior to use bacteria were diluted in tissue culture medium and added to cells as described below. Inoculum titers were confirmed following enumeration of viable bacteria from serial dilutions plated on LB agar as previously described [54].
Brucella abortus strain 2308 expressing GFP was kindly supplied by Jean Celli, Ph.D. (Rocky Mountain Laboratories, Hamilton, MT). GFP-B. abortus was cultured on Tryptic Soy agar (TSA) plates for 48 h at 37°C. Individual colonies were then transferred to Tryptic Soy Broth (TSB) and bacteria were cultured overnight at 37°C with constant shaking. The number of bacteria present on broth cultures was determined by OD 600nm. Actual numbers of viable bacteria were confirmed by plating an inoculum on TSA plates as previously described [55].
LVS was grown in MMH broth as described above. Following overnight culture, LVS was centrifuged for 15 minutes at 8000×g. The resulting pellet was resuspended in breaking buffer (50 mM Tris/HCl, 0.6 µg/ml DNase, 0.6 µg/ml RNase, 1 mM EDTA [all from Sigma] and 1 Complete EDTA free tablet [Roche]) and the bacteria were centrifuged again for 15 minutes at 8000×g. Pelleted bacteria were then resuspended in breaking buffer. To break open LVS, the bacteria were added to Fast Prep Lysing Matrix B tubes and processed in a FastPrep24 (MPBio) for 10 cycles of 45 seconds with 2 minute rest periods on ice in between each cycle. The resulting slurry was then centrifuged at 10,000 rpm for 10 minutes. The supernatant was collected and centrifuged twice at 100,000×g for 4 h. The pellet was resuspended in buffer containing 50 mM Tris/HCl, 1 mM EDTA and dialyzed against PBS using 3000 MW cutoff Slide-A-Lyzer cassettes (Pierce). Protein concentration of LVS membrane protein fraction (MPF) was determined using a BCA Protein Assay Reagent Kit according to the manufacturer's instructions. Endotoxin levels were determined using Limulus Amebocyte Lysate (LAL) assay. Endotoxin levels were <0.1 EU/µg of protein. MPF was then aliquoted, irradiated to render it sterile, and stored at -80°C. SchuS4 LPS was generated as previously described [23]. LVS LPS was obtained from the BEI Resources (Manassas, VA).
CLDC was provided by Juvaris Biotherapeutics. Formulation of CLDC has been previously described [56]. CLDC was prepared by Juvaris Therapeutics under good manufacturing conditions (GMP). Briefly, DOTIM∶cholesterol liposomes were combined to form a liposome intermediate. pMB75.6 plasmids were generated from the non-pathogenic strain of E. coli DH5α. The plasmid contains the following elements: (i) the cytomegalovirus immediate early (CMV-IE) promoter/enhancer, (ii) a polyadenylation signal derived from simian virus 40 (SV40), (iii) a kanamycin-resistance gene, and (iv) an origin of replication (pUCori/f1ori). There are no genes expressed by the plasmid. pMB75.6 plasmids were added to liposomes at a ratio of 9.8∶1 (lipid∶DNA). CLDC were aliquoted, lyophilized, and stored at 4°C. Endotoxin levels were determined using Limulus Amebocyte Lysate (LAL) assay. Endotoxin levels of CLDC were <0.1 EU/mg DNA.
Immediately prior to use, CLDC was hydrated using 500 µl endotoxin free, water (Cape Cod Associates Incorporated, E. Falmouth, MA). CLDC were allowed to rehydrate for approximately 5 minutes at room temperature. CLDC were then diluted 1∶3 in 5% dextrose water (D5W; Baxter Healthcare, Deerfield, IL). For in vitro experiments, MPF was thawed, vortexed and added to CLDC at a final concentration of 3.52 µg/ml. As indicated, MPF and CLDC were also tested individually. In these experiments MPF was diluted in 5% dextrose water to 3.52 µg/ml. All CLDC, MPF and CLDC+MPF mixtures were added to cells in a volume of 50 µl/well and were used immediately following preparation.
For in vivo experiments, MPF or LPS was thawed, vortexed, and diluted in 5% dextrose water. In experiments testing MPF or LPS alone, MPF and LPS were diluted to 50 µg/ml. In experiments testing combination of CLDC and MPF or LPS, the reagents were added to CLDC at a final concentration of 50 µg/ml. In these experiments MPF was diluted in 5% dextrose water to 50 µg/ml. All CLDC, MPF and CLDC+MPF mixtures were used immediately following preparation. Mice were injected with 200 µl of prepared CLDC+MPF.
Bone marrow derived macrophages were generated as previously described [51] with the exception that cells were differentiated into macrophages in tissue culture plates with or without glass coverslips in the presence of 10 ng/ml recombinant murine M-CSF (Peprotech, Rocky Hill, NJ). All cells were used on day 6 of culture.
Human monocyte derived macrophages were differentiated from apheresed peripheral blood monocytes as previously described [57]. Briefly, apheresed monocytes were enriched using Ficoll-paque (GE Healthcare). CD14+ progenitor cells were enriched via negative selection using Dynabeads MyPure Monocytes Kit for untouched human cells per manufacturer's instructions (Invitrogen). Cells were resuspended at 3×105/ml in cRPMI supplemented with 10 ng/ml M-CSF (Peprotech) plated at 1 ml/well in 24-well plates with or without glass coverslips and incubated at 37°C/5%CO2. On day 2 and 5 of culture medium was replaced with cRPMI supplemented with 10 ng/ml M-CSF. All cells were used on day 6 of culture. Human blood cells were collected from anonymous volunteers under a protocol reviewed and approved by the NIH Clinical Center Institutional Review Board. Signed, informed consent was obtained from each donor acknowledging that their donation would be used for research purposes by intramural investigators throughout NIH.
Macrophages were treated with CLDC, MPF or CLDC+MPF, prepared as described above, 18 h prior to infection with bacteria. As indicated, 1 or 6 hr prior to addition of CLDC+MPF human and mouse macrophages were pretreated with 10 mM (human) or 3 mM (mouse) N-acetyl-L-cysteine (NAC) (Sigma), 3 mM (human) or 1 mM (mouse) L5-[imino(methylamino)methyl]-L-ornithinemonoacetate (L-NMMA), (Caymen Chemicals), respectively.
Following treatments described above, macrophages were infected with either F. tularensis SchuS4 (MOI = 50), B. abortus (MOI = 50), B. pseudomallei (MOI = 1), or Y. pestis (MOI = 2). Briefly, medium was removed and reserved, and then F. tularensis was co-incubated with macrophages at 37°C in 7% CO2 for 1.5 h followed by treatment with gentamicin (Invitrogen) at 500 µg/ml for 1 h. Then, cultures were washed extensively and reserved medium was replaced. The infection inoculum was confirmed by plating serial dilutions of stock F. tularensis on MMH agar plates immediately prior to addition to cell cultures.
B. pseudomallei was co-incubated with macrophages at 37°C in 5% CO2 for 1 h followed by addition of kanamycin (Invitrogen) at 250 µg/ml for the remainder of the experiment.
Y. pestis was co-incubated with macrophages at 21°C in 5% CO2 for 1 h followed by treated with gentamicin at 80 µg/ml for 1 h. Then, cultures were washed extensively and reserved medium supplemented with 8 µg/ml gentamicin was replaced.
Macrophages were infected with B. abortus as previously described [55]. At the indicated time points, cells were either lysed for enumeration of intracellular bacteria or isolation of RNA, or were fixed with 3% paraformaldehyde in PBS for 20 min at 37°C/5%CO2 prior to analysis for bacteria as described below.
At the indicated time points, medium was removed and cells were washed extensively. Then, cells were lysed following incubation in sterile water. Cell lysates were immediately serially diluted and plated on either MMH, LB, blood, or TSA agar plates. Agar plates were incubated at 37°C/7% CO2 for 48–72 h for enumeration of bacterial colonies.
Macrophages were grown on coverslips, treated, and infected with SchuS4, GFP-B. pseudomallei, GFP-Y. pestis or GFP-B. abortus as described above. Cells were fixed in 3% paraformaldehyde for 20 minutes at 37°C/5%CO2. Cells were washed with PBS and stained for LAMP-1 as previously described [10], [57]. SchuS4 was detected using Alexa Fluor488 goat conjugated anti-F. tularensis (US Biological, Swampscot, MA) as previously described [57]. Samples were observed on a Carl Zeiss (Thornwood, NY) Axio Imager.M1 epifluoresence microscope for quantitative analysis. Approximately 75–100 cells/field and a minimum of three fields per coverslip were analyzed for presence of intracellular bacteria. Percent of infected cells was calculated as follows: (number infected cells/total number of cells) ×100.
Confocal images of 1,024×1,024 pixels were acquired and assembled using Adobe Photoshop CS2 software (Adobe Systems, San Jose, CA).
RNA was isolated and converted to cDNA using RT2 qPCR Grade RNA Isolation Kit and RT2 First Strand Kit (both from SA Biosciences, Frederick, MD) according to manufacturer's instructions. Samples were assessed for ROS and RNS associated genes expression using nitric oxide RT2 Profiler PCR Arrays for mouse and human per manufacturer's instructions (SA Biosciences). Gene expression was quantitated using RT2 Profiler PCR Array Data Analysis Software (SA Biosciences). Accession numbers for genes in which expression was increased or decreased compared to untreated controls are listed in Table S1.
Specific-pathogen-free, 6–8 week old male C57Bl/6 mice (wild type) (n = 10/group) were purchased from Jackson Laboratories (Bar Harbor, MI). nos2/phox gp91 deficient mice (nos2/gp91−/−) mice were bred at the NIAID/Rocky Mountain Laboratories. All research involving animals was conducted in accordance with Animal Care and Use guidelines and animal protocols were approved by the Animal Care and Use Committee at RML.
CLDC, MPF and CLDC+MPF were prepared as described above. At the time points indicated, mice were injected intravenously (i.v.), intraperitoneally (i.p.) or subcutaneously (s.c.) with 200 µl of each preparation prior to challenge. For intranasal (i.n.) administration mice were anesthetized via intraperitoneal injection with 100 µl of ketamine (12.5 mg/ml) + xylazine (3.8 mg/ml) solution and the reagents were delivered in a total volume of 25 µl evenly distributed between the nares at the indicated time points prior to challenge. Then, mice were anesthetized via intraperitoneal injection with 100 µl of ketamine (12.5 mg/ml) + xylazine (3.8 mg/ml) solution and intranasally infected with 25 CFU SchuS4 diluted in a final volume of 25 µl of PBS. Inoculating doses were confirmed by plating inoculum on MMH agar. This inoculum routinely results in 100% mortality and a mean time to death of 5 days following infection in naive animals.
For in vitro studies, statistical differences between two groups were determined using an unpaired student t test with the significance set at p<0.05. For comparison between three or more groups, analysis was done by one-way ANOVA followed by Tukey's multiple comparisons test with significance determined at p<0.05. For in vivo studies, significance in survival was assessed using log-rank (Mantel Cox) test with significance set at p<0.05.
|
10.1371/journal.pntd.0001156 | Host Alternation Is Necessary to Maintain the Genome Stability of Rift Valley Fever Virus | Most arthropod-borne viruses (arboviruses) are RNA viruses, which are maintained in nature by replication cycles that alternate between arthropod and vertebrate hosts. Arboviruses appear to experience lower rates of evolution than RNA viruses that replicate in a single host. This genetic stability is assumed to result from a fitness trade-off imposed by host alternation, which constrains arbovirus genome evolution. To test this hypothesis, we used Rift Valley fever virus (RVFV), an arbovirus that can be transmitted either directly (between vertebrates during the manipulation of infected tissues, and between mosquitoes by vertical transmission) or indirectly (from one vertebrate to another by mosquito-borne transmission).
RVFV was serially passaged in BHK21 (hamster) or Aag2 (Aedes aegypti) cells, or in alternation between the two cell types. After 30 passages, these single host-passaged viruses lost their virulence and induced protective effects against a challenge with a virulent virus. Large deletions in the NSs gene that encodes the virulence factor were detectable from the 15th serial passage onwards in BHK21 cells and from the 10th passage in Aag2 cells. The phosphoprotein NSs is not essential to viral replication allowing clones carrying deletions in NSs to predominate as they replicate slightly more rapidly. No genetic changes were found in viruses that were passaged alternately between arthropod and vertebrate cells. Furthermore, alternating passaged viruses presenting complete NSs gene remained virulent after 30 passages.
Our results strongly support the view that alternating replication is necessary to maintain the virulence factor carried by the NSs phosphoprotein.
| Arthropod-borne viruses are transmitted among vertebrate hosts by insect vectors. Unusually, Rift Valley fever virus (RVFV) can also be transmitted by direct contacts of animals/humans with infectious tissues. What are the molecular mechanisms and evolutionary events leading to adopt one mode of transmission rather than the other? Viral replication is implied to be different in a vertebrate host and an invertebrate host. The alternating host cycle tends to limit virus evolution by adopting a compromise fitness level for replication in both hosts. To test this hypothesis, we used a cell culture model system to study the evolution of RVFV. We found that freeing RVFV from alternating replication in mammalian and mosquito cells led to large deletions in the NSs gene carrying the virulence factor. Resulting NSs-truncated viruses were able to protect mice from a challenge with a virulent RVFV. Thus, in nature, virulence is likely maintained by continuous alternating passages between vertebrates and insects. Thereby, depending on the mode of transmission adopted, the evolution of RVFV will be of major importance to predict the outcome of outbreaks.
| Most arthropod-borne viruses (arboviruses) are RNA viruses, although they use a variety of strategies to ensure their replication and transmission. The feature that best distinguishes RNA genomes from DNA ones is the high mutation rate of the former during replication. Misincorporation errors during replication have been estimated to occur within the range of 10−3–10−5 substitutions per nucleotide and per round of copying [1]. The main factor contributing to such high mutation rates is a lack of proof-reading repair activities that is associated with RNA replicases [2]. Another source of mutations results from the spontaneous deamination of Cytidine residues to Uracil. In DNA genomes, this reaction is repaired by a Uracil-glycosylase, but this cannot function on an RNA template [3]. The resulting complex mixtures of closely related RNA genomes are termed quasispecies [4], [5] whose existence allow RNA viruses to adapt rapidly to fluctuating environments [6], [7]. Indeed, mutation rates per nucleotide site of around 10−4 mean that for a 10 kb genome, an average of one mutation is incorporated each time the genome is copied, and it is this, together with short replication times and large population sizes, that ensure the existence of the quasispecies genome pool.
However, sequence comparisons reveal that RNA arboviruses are relatively stable in nature, suggesting that the alternating host cycle (between vertebrate and invertebrate hosts) constrains viral evolution by a strong conservative sequence selection. This sequence stability may result from the requirements for replication in two separate hosts that present conflicting niches for replication and adaptation [8]. Furthermore, low rates of evolution do not necessarily reflect the adaptive compromise of a virus to the alternating host cycle [9], but could be principally related to other biological constraints including the need to maintain virulence [10], [11]. Effectively, more virulent viral strains are generally at a competitive advantage in mixed-strain infections [12]. Virulence can be considered as a consequence of virus efforts to maximize its fitness: the virus must replicate extensively in a host to ensure its transmission to the next host. However, viral replication damages host tissues, leading to host death, which can be considered as seriously deleterious for virus survival [13], [14]. Alternation may play a significant role in maintaining the genetic stability of arboviruses; setting aside one or several selective filters may lead to accelerate evolution. For such, we used a cell culture system for these studies, as in vitro systems are convenient to investigate the evolution of arboviruses.
Contrary to most arboviruses, Rift Valley fever virus (RVFV), a member of the Phlebovirus genus within the Bunyaviridae family can be transmitted through direct contact with body fluids or aborted fetuses. It constitutes an interesting model to study host alternating cycling as cause of genetic stability of arboviruses in nature. RVFV is a tri-segmented negative-stranded RNA virus composed of the L segment that codes for the RNA-dependent RNA polymerase, the M segment that codes for the GN and GC glycoprotein precursor, and the S segment that has an ambisense strategy, coding for the N nucleoprotein and the NSs phosphoprotein [15]. The NSs phosphoprotein plays a key role in RVFV pathogenesis in the mammalian host. A natural isolate defective for the NSs protein (Clone 13; [16]), was found to be avirulent for mice [17] and to induce an interferon response in mammalian cells, in contrast to virulent RVFV strains [18].
Here, we present results showing that genetic material non-essential to viral replication such as the phosphoprotein NSs is rapidly eliminated leading to the loss of virulence.
The Institut Pasteur animal facility has received accreditation from the French Ministry of Agriculture to perform experiments on live mice in appliance of the French and European regulations on care and protection of the Laboratory Animals. This study was approved by the relative IACUC at the Institut Pasteur.
We used two cell lines: a mammalian cell line derived from hamster kidney (BHK21) and an insect cell line derived from Aedes aegypti larvae (Aag2). BHK21 cells defective in IFN-a/b signaling were grown at 37°C with 5% CO2 in Glasgow's minimal essential medium (G-MEM) containing 5% fetal bovine serum (FBS), 1000 units/mL penicillin, 1 mg/mL streptomycin, tryptose phosphate broth 1× and HEPES 0.01M. Monolayer cultures of Aag2 cells were grown at 28°C in Schneider Drosophila medium supplemented with 10% heat-inactivated FBS, L-Glutamine 0.4×, 1000 units/mL penicillin and 1 mg/mL streptomycin.
The parental P strain was derived from the ZH548 strain originally isolated in 1977 from a human case in Egypt [19] and passaged three times in Vero cells. The virulence of this strain is related to the phosphoprotein NSs, which is responsible for a general inhibition of cellular RNA synthesis by interacting with the p44 subunit of the TFIIH transcription factor [20]. Furthermore, NSs strongly antagonizes IFN-ß production [17], [18], [21], [22]. The titer of the frozen virus stock was 106.8 PFU (plaque forming unit)/mL. Moreover, the other virus strains were produced at high titers: Z30Alt at 108.3 PFU/mL, Z30B at 107.6 PFU/mL and Z30A at 108.5 PFU/mL. Viral titers were estimated by serial 10-fold dilutions on Vero cells.
In addition, biological clones Z30AC and Z30BC were produced by plaque purification and amplification in Vero cells from the 30th serial passage of parental P strain in Aag2 cells (Z30A strain) and BHK21 cells (Z30B strain), respectively. Briefly, six-well plates containing confluent monolayers of Vero cells were infected with serial 10-fold dilutions of virus. Cells were incubated for five days under an overlay consisting of Dulbecco's MEM (DMEM), 2% FBS, antibiotics and 1% agarose at 37°C. The lytic plaques were localized and removed by suction using a pipette. Each agarose plug that contained an individual clone was dissolved overnight at +4°C in DMEM supplemented with 10% FBS before being re-amplified in BHK21 or Aag2 cells, respectively [23]. Both clones were produced at high titers: 108.8 PFU/mL for Z30BC and 108.7 PFU/mL for Z30AC.
The parental P strain virus was subjected to 30 serial passages in BHK21 cells or Aag2 cells, or 30 passages that alternated between BHK21 and Aag2 cells (i.e. 15 passages in BHK21 cells and 15 passages in Aag2 cells), at a multiplicity of infection (MOI) of 0.1 PFU/cell. Virus was adsorbed for 1 hr onto confluent cell monolayers prepared in plastic flasks of 25 cm2, at 28°C for Aag2 and at 37°C for BHK21. After adsorption, the inoculum was removed, cells were washed with medium, 13 mL of maintenance medium (with 2% FBS) was added and cells were incubated at the appropriate temperature. Cell supernatants were harvested when titers reached a plateau (Figure S1): at 48 hr p.i. for BHK21 and 96 hr p.i. for Aag2 cells. At each passage, supernatants were harvested and stored in aliquots at −80°C for titration on Vero cells.
Plastic flasks of 75 cm2 containing confluent cell monolayers (Aag2 or BHK21) were infected at a MOI of 0.1 PFU/cell as described above. The supernatant from a flask was harvested every 2 hr from 0 to 12 hr p.i. and every 24 hr from 0 to 120 hr p.i., and titrated on Vero cells by serial 10-fold dilution [24].
Total RNA was extracted from aliquots of supernatants (100 µL) using the Nucleospin RNA II kit (Macherey-Nagel) according to the manufacturer's instructions, and RT-PCR targeting the NSs gene was conducted using the Titan One Tube RT-PCR kit (Roche Applied Science) following the manufacturer's recommendations. Primers were selected in the NSs gene that lies within the S genome segment. The amplification program was performed as follows: reverse transcription at 50°C for 30 min, an inactivation of RT enzyme step at 95°C for 3 min, followed by 35 cycles of 95°C 30 s, 51°C 30 s, 72°C 1 min, and a final step at 72°C for 5 min. The size of the PCR product was 781 bp. PCR products were excised from the gel and eluted using the QIAquick Gel Extraction Kit (Qiagen) as specified by the manufacturer. The recovered DNA was cloned into the Topo TA vector and transformed into Top10 competent cells according to the manufacturer's protocol. Colonies were screened by direct PCR, using insert-specific primers. Plasmid DNA was purified using a QIAprep Spin Miniprep kit (Qiagen), as specified by the manufacturer. Sequencing was carried out using virus-specific primers. In addition, the three segments (S, M and L) of the parental P strain and the selected strains (Z30Alt, Z30B and Z30A) were completely sequenced. For each segment, primers were designed (based on the nucleotide sequence of the reference strain ZH548) in order to obtain around 700 pb RT-PCR amplicons with an overlap of around 100 pb along the entire segment (Table S1). Amplicons were obtained using SuperScript One-Step RT-PCR with platinium Taq (Invitrogen) following the manufacturer's recommendations. The amplification program was performed as follows: reverse transcription at 50°C for 30 min, an inactivation of RT enzyme step at 94°C for 2 min, followed by 35 cycles of 94°C 15 s, 50°C 30 s, 72°C 1 min 30, and a final step at 72°C for 10 min. The obtained fragments were purified by ultrafiltration (Millipore). Sequencing reactions were performed using the BigDye Terminator v1.1 cycle sequencing kit (Applied Biosystems) and purified by ethanol precipitation. Sequence chromatograms from both strands were obtained on automated sequence analyzer ABI3730XL (Applied Biosystems). For sequence analysis, contig assembly was performed using the software BioNumerics version 5.1 (Applied-Maths, Sint-Martens-Latem, Belgium). Sequence alignments and computation of substitution tables were also performed using the BioNumerics software. For phylogenetic analysis, maximum-likelihood trees were constructed using MEGA version 4 [25] with the Kimura-2 parameter for corrections of multiple substitutions. Reliability of nodes was assessed by boostrap resampling with 1,000 replicates.
The pathogenicity of the parental P strain, and the selected strains, Z30Alt, Z30B, Z30A, Z30BC or Z30AC, was assayed in 4- to 5-week-old female Swiss mice (OF-1; Charles River, France) by inoculating 104 PFU intraperitoneally into each mouse. The control was inoculated with DMEM supplemented with 10% FBS. Mice surviving at the end of the observation period were bled and their sera tested for IgG by enzyme-linked immunosorbent assay (ELISA) [26]. Two experiments were carried out. In the first, each batch of five mice was infected with a different virus strain: P, Z30Alt, Z30B, Z30A, Z30BC or Z30AC. One batch was used as the control. Mice were kept under observation for 21 days post-inoculation or until death occurred. In the second experiment, we aimed to detect the IgG protective capacity induced by selected clones in inoculated mice. Batches of 12 mice were inoculated with the clones Z30BC and Z30AC, and one batch was used as control. At day 14 post-inoculation, one half of surviving mice from each batch was challenged with 104 PFU of the parental P strain and the other half was inoculated with DMEM. Mice were then observed for 21 days after challenge.
Figure 1 describes the strategy adopted to test the hypothesis that an alternating host cycle constrains the evolution of RVFV.
To compare the replication of selected RVFV strains (Z30B, Z30A, Z30BC, Z30AC) to the parental P strain, replication kinetics were determined in BHK21 cells and in Aag2 cells. When examining replication rates in BHK21 cells (Figure 2A), all strains gave similar patterns of viral growth, with an initial exponential growth phase until 24 hr post-infection (p.i.) followed by a plateau. However, the Z30BC clone exhibited higher titers than other virus strains, almost 1 log10 PFU/mL higher from 6 hr p.i. until 72 hr p.i. Furthermore, viral replication was detectable two hours earlier for Z30BC than for the other four strains tested. This strain reached a maximum titer of around 9 log10 PFU/mL from 24 hr p.i. onwards. These results suggest that the Z30BC strain was better adapted to BHK21 cells than other strains were, and that the 30th passage in BHK21 cells (Z30B) from which Z30BC was derived encompassed a mixture of different viral clones with variable capacities to replicate in BHK21 cells. When replication rates in Aag2 cells were analyzed (Figure 2B), differences in titers were clear-cut from 24 hr p.i. onwards. Effectively, Z30A and the Z30AC clone derived from this strain replicated to higher titers than the other three strains tested: ∼4 log10 PFU/mL higher at 48 hr p.i. and ∼2 log10 PFU/mL higher at 96 hr p.i. Moreover, replication of Z30A and Z30AC was detectable 24 hr earlier than that of the other strains, suggesting their better adaptation to Aag2 cells. In summary, the two selected clones that resulted from serial passages in either BHK21 or Aag2 cells exhibited an increased replication capacity in the corresponding cell type.
We inoculated four-week-old mice intraperitoneally with 104 PFU of one of six viral strains to evaluate their virulence. Strains tested were: P (the parental strain), Z30Alt (isolated at the 30th alternating passage in BHK21 and Aag2 cells), Z30B (pool harvest at the 30th serial passage in BHK21 cells), Z30A (pool harvest at the 30th serial passage in Aag2 cells), Z30BC (clone selected at the 30th serial passage in BHK21 cells) and Z30AC (clone selected at the 30th serial passage in Aag2 cells). Mouse survival rates were recorded over 21 days post-inoculation (Figure 3A). The control, inoculated with Dulbecco's MEM (DMEM) medium, survived 21 days. All batches of mice inoculated with a given viral strain also survived, except for those inoculated with the parental P strain and the 30th alternating passage strain (Figure 3A). All mice inoculated with P died before day 7 post-inoculation, and 4 of 5 treated with Z30Alt died before day 9 post-inoculation. Thus, the 30th alternating passage strain behaved roughly like the parental P strain after 30 passages. All surviving mice were tested for the presence of IgG against RVFV at day 21 post-inoculation, and showed positive compared to the IgG level in non-infected mice (Table S2).
To test the protective effect of infection by clones Z30BC or Z30AC against infection pool harvest with the parental P strain, further batches of mice were first inoculated with the Z30BC or Z30AC clones and then challenged 14 days later by inoculation with the parental P strain. Mortality rates were scored up to 21 days after challenge (Figure 3B). Before challenge, all mice had survived. The control mice started to die five days after challenge, and no controls survived beyond 9 days after challenge. Batches of mice that received one of the two clones at day 0 survived for 36 days. When mice previously inoculated with the Z30BC clone were challenged with the P strain, all mice survived 21 days after challenge. This suggests protection by the Z30BC clone selected in BHK21 cells. Mice sera were all IgG positive (Table S3). In batches of mice first inoculated with the clone Z30AC, one among 6 mice died 8 days after challenge with the parental P strain. The five others survived 21 days after challenge. The sera of the surviving mice were IgG positive (Table S3), suggesting that primary inoculation with the Z30AC clone selected in Aag2 cells could protect mice against a secondary inoculation with the parental P strain.
As the phosphoprotein NSs has been shown to be responsible for virulence, we used RT-PCR amplification to monitor the NSs gene upon serial or alternating passages through BHK21 and Aag2 cells. Amplification of the parental P strain NSs gene generated an amplicon of ∼780 bp (Figure 4A). Upon serial passage in BHK21 cells, PCR products from the 10th, 20th and 30th passage exhibited different sized amplicons, with a predominant band at 700–800 bp at the 10th and 20th passages and a band at 500–600 bp at the 30th passage (Figure 4A). When examining the RT-PCR profiles for different serial passages in BHK21 cells in detail, it could be seen that the 500–600 bp band was detectable from the 15th passage, reaching maximum expression from the 25th passage onwards, concomitant with a decrease in the expression of 700–800 bp bands (Figure 4B). The shortening of the NSs gene coincided with the loss of cellular lysis in BHK21 cells, evident by the 30th passage. Upon serial passages in Aag2 cells, a band of the expected size corresponding to the NSs gene was found at the 10th passage, whereas a smaller band of 500–600 bp was detected at the 20th and 30th passage (Figure 4A). This smaller band was actually found as early as the 11th passage, and became predominant by the 21st passage (Figure 4C), since the 700–800 bp band decreased in quantity from the 20th passage. In contrast, when virus was subjected to alternating passages, only a major band at 700–800 bp was found, irrespective of passage number (Figure 4A).
To characterize more precisely the molecular events associated with the emergence of viral variants, clones were analyzed by RT-PCR and sequencing of the NSs gene. Virus from the 30th passage in BHK21 or Aag2 cells showed large deletions in NSs, while viruses passaged alternately through the two cell types showed no nucleotide changes, suggesting the maintenance of NSs integrity during alternating cell-type passages. 48 clones isolated from the 30th serial passage in BHK21 presented two deletions: (i) a deletion of 259 nucleotides (nt) at position 124 that leads to a shift in the NSs open reading frame (ORF), introducing a stop codon at position 437, and (ii) a smaller deletion of 6 nt at position 650. Thus, the Z30BC clone has a 533 nt NSs gene. 48 clones isolated after 30 passages in Aag2 cells presented two deletions in the NSs gene: (i) a deletion of 73 nt at position 374, inducing a shift in the ORF with the introduction of a stop codon at position 474, and (ii) a deletion of 157 nt at position 536. The Z30AC clone thus presents a 568 nt NSs gene. Figure 5 summarizes the cartography of the deletions found in the NSs gene of clones Z30BC and Z30AC compared to the parental P strain. In contrast, Z30Alt which was isolated from the 30th alternating passage in BHK21 and Aag2 cells had no deletion in the Nss gene. These results suggest that all clones containing a shorter NSs gene than the parent resulted from a single molecular event in both BHK21 and Aag2 cells; in our experiment, all 48 clones examined from the 30th serial passage presented the same deletions. To further explore the molecular features of the parental P strain and the selected strains (Z30Alt, Z30B and Z30A), the complete sequencing of the three segments S, M and L was achieved. Except for the previously described deletions in the NSs gene for the strains Z30B and Z30A, no other deletions or mutations were observed in the S segment for the parental P strain, the Z30B and the Z30A strains. Only one silent mutation was found in the NSs gene for the Z30Alt strain (Table S4). Concerning the segments M and L, no deletion event was found for the four strains, and the phylogenetic analysis () confirmed a cluster in the genetic lineage A around the reference strain ZH548 originally isolated in 1977 in Egypt [27]. Nevertheless, the segments M and L presented some mutations, for most non silent, in the three selected strains. As expected, the segment M was the most variable with a total of 11 amino-acid substitutions compared to the segment L with only five amino-acid changes (Table S4). To note, the segment M of the three selected strains have retained the previously described five in-frame AUG-methionine start codons and the different amino-acids involved in glycosylation [27].
Results presented in this paper support the hypothesis that an alternating viral cycle comprising of infection of two distinct hosts constrains the evolution of RVFV. The study model used consisted of abolishing the alternate host environment using a cell culture system. Serial passages in a single cell type led to the loss of virulence. Large deletions were observed in the NSs gene, non-essential to viral replication. The rapid emergence of NSs-deleted variants in the course of serial passages is likely to result from a selective advantage in their replication rates.
After 30 serial passages in either mammalian BHK21 cells defective in IFN-a/b signaling or mosquito Aag2 cells, single-host adapted viruses from mosquito cells (Z30A and Z30AC) replicated better in mosquito cells. These mosquito-cell adapted viruses reached higher titers (2 log10 PFU/mL higher) than non-adapted viruses, and their replication was detectable earlier (24 hr earlier). Interestingly, full adaptation to mammalian cells needed more passages as shown in the Figure 4B where the deleted variant was not totally predominant at the 30th serial passage in BHK21 cells. Replication patterns of the Z30B strain in mammalian cells were similar to those of the parental P strain and of viruses that had bypassed this host. However, the Z30BC clone isolated from the 30th serial passage in BHK21 cells presented the highest replication rate in mammalian cells (replication was detectable 2 hr earlier and reached a titer ∼2 log10 PFU higher than for other viruses). This result suggested that a higher number of passages would favor adaptation to BHK21 cells which are defective in IFN-a/b signaling. Viral clones from single-cell passages showed a consistent fitness advantage over the parental P strain in the cell type used for their selection: the Z30BC clone exhibited a fitness advantage in BHK21 cells and the Z30AC clone in Aag2 cells. Thus, our clones isolated from single-host adapted passages present fitness gains in their specific host, and no fitness changes in the bypassed host. Similar results have been obtained for members of various arbovirus families using cell culture model systems [28]–[31] or in vivo systems [28], [30], [32]–[35].
Surprisingly, when sequencing the NSs gene, that encodes the virulence factor responsible for a general inhibition of cellular RNA synthesis and IFN-ß production, we found that single-host adapted viruses presented large deletions in this gene. Virus passaged in mammalian BHK21 cells defective in IFN-a/b signaling had a deletion of 259 nt introducing a stop-codon in the NSs gene. The resulting protein was shortened to 60 amino-acids instead of 265. Virus passaged in mosquito Aag2 cells showed a deletion of 73 nt, again with a stop-codon, causing a shortening of the NSs protein to 131 amino-acids. These two distinct deletions in the NSs gene following 30 serial passages in each cell type suggest that the viral genome may function differently depending on whether replication is in mammalian or mosquito cells. Thus, while serial passages of RVFV in a single cell type selected for a virus with a truncated NSs gene specific to that cell type, alternating passages did not allow the emergence of deletions in the NSs gene. Indeed, like the parental P strain, the 30th alternating passage virus Z30Alt did not present any major genetic changes in the NSs gene. The fact that the NSs gene is dispensable in both single host systems suggests that other mutations are involved in the host adaptation process. Thus, the different non-synonymous mutations identified in the segments M and L should be further explored in this context, particularly since they are mostly specialized depending on host. It is likely that these kinds of deletion events take place spontaneously during viral replication and are selected only in the absence of alternation. Further experiments involving independent serial passages would permit to evaluate the frequency at which the phenomenon occurs. Deletions in the NSs gene have been described in a naturally attenuated RVFV (Clone 13) purified from a nonfatal human case in the Central African Republic [16]. This strain has a large internal deletion of 549 nt in the NSs gene (∼70% of its length). Animals can survive high infectious doses of Clone 13, up to 106 PFU, without developing any symptoms. Furthermore, Clone 13 has been tested as a vaccine candidate in sheep and cattle that can indeed then elicit a protective response against challenge with a virulent RVFV strain. When the clones we obtained after 30 serial passages in a single cell type were inoculated into mice (at 104 PFU), the animals survived for 21 days and developed protective IgG against challenge with a virulent strain of RVFV. Moreover, most mice inoculated with Z30Alt (isolated at the 30th alternating passage) died 2 days later than did those treated with the parental P strain, and one mouse survived virus inoculation. Having said that, it is known that the outcome of infection is mainly determined by a balance between the rate of viral replication and the immune response, which together limit viral spread [36]. This might explain why one of five mice recovered from infection. Nevertheless, the 30th alternating passage virus Z30Alt retained roughly the same level of virulence as the parental P strain owing to the maintenance of NSs integrity.
From our results, we have provided new insights as to how biological constraints such as host alternation are necessary to maintain RVFV integrity and virulence. Contrary to most other arboviruses, RVFV could escape from the alternating host replication cycle to evolve more rapidly, like single-host animal RNA viruses. Thus, our single-host adapted RVFV present large deletions in the NSs gene associated with a loss of virulence when inoculated into mice, which develop a long-lasting immunity. In contrast to persistent non-cytolytic replication in insects, arboviruses must replicate to high titers in the mammalian host. This increases the probability of transmission during a blood-meal [37], [38]. By its transfer from wild animals (e.g. buffalo [39]) to livestock, RVFV may intensify direct transmission by contacts with infectious tissues or fluids hosting high viral loads. From our results, viruses selected on mammalian cells may favor attenuation via NSs alteration, leading to the maintenance of avirulent RVFV strains. Surprisingly, mosquito cell-specific selection also leads to large deletions in the NSs gene, with similar phenotypic consequences. In both cases, virulence will only be restored when alternation between both cell types is initiated in conditions that would reconstitute a complete viral genome by reassortments with a virulent RVFV strain. Such in vitro studies should be consolidated with studies using in vivo systems. Indeed, vertebrates are subjected to acute infections, with clearance of the virus being triggered by the immune defense system, whereas insect vectors sustain persistent viral replication and are the site of such genetic changes as reassortment or recombination upon co-infection [40]. Such rearrangements may restore virulence, upon the acquisition of a complete NSs gene in the course of virus replication [41]. Finally, our results suggest that subtle modifications of selective filters can lead to major genetic changes within a viral population.
|
10.1371/journal.pgen.1000492 | Root Suberin Forms an Extracellular Barrier That Affects Water Relations and Mineral Nutrition in Arabidopsis | Though central to our understanding of how roots perform their vital function of scavenging water and solutes from the soil, no direct genetic evidence currently exists to support the foundational model that suberin acts to form a chemical barrier limiting the extracellular, or apoplastic, transport of water and solutes in plant roots. Using the newly characterized enhanced suberin1 (esb1) mutant, we established a connection in Arabidopsis thaliana between suberin in the root and both water movement through the plant and solute accumulation in the shoot. Esb1 mutants, characterized by increased root suberin, were found to have reduced day time transpiration rates and increased water-use efficiency during their vegetative growth period. Furthermore, these changes in suberin and water transport were associated with decreases in the accumulation of Ca, Mn, and Zn and increases in the accumulation of Na, S, K, As, Se, and Mo in the shoot. Here, we present direct genetic evidence establishing that suberin in the roots plays a critical role in controlling both water and mineral ion uptake and transport to the leaves. The changes observed in the elemental accumulation in leaves are also interpreted as evidence that a significant component of the radial root transport of Ca, Mn, and Zn occurs in the apoplast.
| The root system is a highly specialized plant organ that works to get both water and essential mineral nutrients from the changing chemically and physically complex environment of the soil. Roots do this by both controlling the uptake of water and essential mineral ions, as well as regulating their movement to the central vascular system of the plant for long distance transport to the shoot. To allow the cellular control of water and mineral ion uptake and transport via specialized transport proteins, plant roots contain a waxy layer of suberin that acts to seal connections between cells, preventing uncontrolled leakage of water and mineral ions between cells. By screening thousands of mutant A. thaliana plants, we were able to identify a plant with elevated levels of suberin in the root. Using this mutant, we were able to uncover the importance of suberin in sealing connections between root cells to regulate water movement through the plant and accumulation of various essential and nonessential minerals in leaves, including sodium, sulfur, potassium, calcium, manganese, zinc, arsenic, selenium, and molybdenum.
| The plant root is a specialized organ that allows uptake of water and selective uptake of solutes from the soil environment, to support normal plant growth and development. To accomplish this function roots take up water and solutes at the root surface and transport them across the root to the xylem vessels in the central vascular tissue in the stele, where they are transported to the shoot. Transport across the root to the central xylem vessels can occur through the extracellular space (apoplastic) or via cell-to-cell transport (symplastic). Specificity of both water and solute transport is generally thought to be provided by transport proteins at the plasma membrane of root cells. However, to achieve such specificity, the non-specific apoplastic transport pathway between cells needs to be regulated. A long held model to explain this regulation includes the function of the extracellular biopolymer suberin acting as a barrier to limit the apoplastic transport of water and solutes. Plant roots contain two such suberin barriers, at the exodermis and the endodermis. Endodermal suberin is thought to prevent the apoplastic movement of water and solutes into the stele, whereas exodermal suberin blocks apoplastic transport at the root surface. Though central to our understanding of how roots function, a careful review of the literature reveals a lack of genetic evidence to support this foundational model.
The effects of altered root suberin content on water relations and ion transport at both the exodermis and endodermis have been tested experimentally by modifying the cultivation conditions or utilizing roots at different developmental stages [reviewed in 1 & 2]. For example, roots of maize grown hydroponically lack a suberized exodermis when compared to aeroponically grown plants [2]. Such differences have been utilized to identify negative correlations between suberin content and root hydraulic conductivity, and radial transport of ABA [1],[3],[4]. Higher suberin content within mature endodermis also results in decreased Ca translocation at different developmental stages along the root of Cucurbita pepo [5]. Further, in species with an exodermis and endodermis (Zea mays, Allium cepa, and Helianthus annuus), or just an endodermis (Vicia faba and Pisum sativum), higher root suberin, in mature regions of the root, correlates with lower apoplastic transport of fluorochromes [6] and root water loss [7]. In concluding that quantitative differences in suberin content, in either the exodermis or endodermis of the root, determines the permeability of the apoplast to both water and solutes the implicit assumption is that differences observed in these physiological parameters are related to the altered suberin contents observed. It has previously been suggested that qualitative differences in the compositions of suberin also need to be considered [8]. Furthermore, given the varied growth conditions, developmental stage or species used in these studies it is quite feasible that differences other than suberin are either directly or indirectly responsible for the observed effects on water relations and ion transport.
Direct evidence for a role of suberin in regulating apoplastic radial transport requires a plant harboring a mutation that specifically alters the amount of suberin deposited in the root. Currently, there are only three reports of Arabidopsis thaliana mutants with altered suberin. As it is generally accepted that A. thaliana roots do not contain an exodermis [9], suberin in these mutants is likely altered in the endodermis. Knockout alleles of GPAT5 (encoding an acyl-CoA: glycerol-3-phosphate acetyltransferase) in A. thaliana were recently shown to contain 20–50% less C20–C24 aliphatic monomers in root suberin [10]. However, the effect of loss-of-function of GPAT5 on either water relations or ion-transport was not directly measured. The recently isolated A. thaliana mutants horst-1 and horst-2 containing knockout alleles of the cytochrome P450 fatty acid ω-hydroxylase CYP86A1 gene contain 60% less total aliphatic suberin monomers in roots than wild-type plants [11],[12]. Knockout alleles of the DAISY locus, which encodes a β-ketoacyl-CoA synthase involved in the formation of docosanoic acid, also have altered aliphatic suberin monomer composition, including a decrease in C22 and C24 fatty acid derivatives and an increase in C16, C18 and C20 fatty acid derivatives [13]. No evidence for impacts on water relations or ion-transport for either the horst or DAISY mutants was reported.
Given the assumed role of suberin in regulating radial transport of solutes in the root, we would predict that perturbations in suberin content and/or its composition would affect the elemental composition, or ionome, of the shoot. High-throughput screening of the shoot ionome may therefore identify mutants with altered root suberin. We have identified numerous ionomics mutants in a previously reported screen [14], and here we report on the cloning and characterization of one of these ionomics mutants, originally reported as mutant 14501 [14]. Here we report that 14501 results in a doubling of all the aliphatic monomer components of root suberin, and we renamed this mutant enhanced suberin1 (esb1-1). This elevated suberin, most likely in the endodermis given that A. thaliana roots lack an exodermis, results in a reduction in transpiration, reduced wilting after water withdrawal, and a perturbation of the shoot ionome, including increases in Na, S, K, As, Se and Mo, and decreases in Ca, Mn and Zn. Grafting experiments confirm that these phenotypes are the result of elevated root suberin. This report provides direct evidence that suberin acts in the root as a barrier to the extracellular transport of both water and mineral ions.
The A. thaliana leaf ionomic mutant 14501, now termed esb1-1, was originally reported to have elevated leaf K, Mo and Cd and reduced Ca and Fe compared to wild-type plants [14]. Here we extend this observation and establish with two independent alleles (esb1-1 and esb1-2) that this mutant shows significant (P≤0.01) increases in shoot concentrations of Na, S, K, As, Se and Mo and reductions in Ca, Mn and Zn. We also observed a significant (P≤0.05) reduction in Fe (Table 1). We further note that during earlier experiments characterizing this mutant, with a different batch of the same soil type (in trays 533–535 and 590–594, see www.ionomicshub.org) the shoot concentration of B was also significantly (P≤0.01) reduced by approximately 25–40% compared to wild-type plants. Even though the shoot ionomes of esb1-1 and esb1-2 are significantly altered compared to wild-type plants we observed no major phenotypic differences.
Ionomic analysis of leaf tissue sampled from an F2 population derived from the backcross Col-0 x esb1-1 revealed that 20 F2 plants from a total of 117 showed the mutant ionomic phenotype scored as the percentage difference from wild-type in leaf concentrations of B, Ca and Mo. Figure 1 shows that these mutant F2 plants cluster with the esb1-1 mutant. The number of F2 plants showing the mutant phenotype is consistent with the hypothesis that the ionomic phenotype of esb1-1 is caused by a recessive monogenic mutation.
To map the causal locus of esb1-1 an outcross to A. thaliana Ler-0 was made and the ionomic phenotype of 175 F2 plants determined, and represented as the percentage difference from wild-type Col-0 for B, Ca and Mo (Figure 2). To obtain a rough map position a bulk segregant analysis (BSA) experiment [15] was performed with microarray detection of genetic markers [16],[17], using (esb1-1 x Ler-0) F2 plants. Plants with the lowest Ca and B content (n = 41), that clustered with esb1-1, and plants with B and Ca shoot contents similar to Col-0 (n = 41) were pooled separately, and genomic DNA from each pool hybridized to the Affymetrix Arabidopsis ATH1 microarray. Using the oligonucleotide probes on the DNA microarray that show differential hybridization between Ler-0 and Col-0 as genetic markers (Single Feature Polymorphisms or SFP), the locus responsible for the ionomics phenotype in esb1-1 was mapped to an area centered at 11 Mb on chromosome 2 (Figure 3A). To identify the causal DNA polymorphism in esb1-1 we hybridized DNA from esb1-1 and wild-type Col-0 to the Affymetrix Arabidopsis ATTILE 1.0R microarray and compared the hybridization at probes that represent sequence between 9 and 13 Mb on chromosome 2 (Figure 3B). Nineteen probes covering the region from 12,308,779 to 12,309,878 showed log2 intensity differences ranging from 0.23 to 3.57 with an average difference of 1.8. Sequencing of this region revealed a 1097 base pair deletion starting at 12,308,627 and ending at 12,309,723. This deletion lies within a genomic region containing the 2 putative open reading frames (ORFs) At2g28670 and At2g28671 (Figure 3C), neither of which are predicted to contain introns. The deletion is within the predicted promoter region of At2g28670 and the 3′ end of At2g28671. Esb1-2 (GABI_858D03) has a T-DNA insertion at 12,308,657. We are aware that At2g28670 has been previously called DIR10 [18]. However, this naming is not based on any published functional evidence, but rather unpublished phylogenetic data. At2g28670 is annotated as a “disease resistance-responsive family protein/fibroin-related” gene, however there are currently no published studies describing the function of this gene. The second predicted ORF in this genomic region At2g28671 has been recently automatically annotated at The Arabidopsis Information Resource (TAIR) (www.arabidopsis.org) as encoding a protein containing the signature prichextensn domain of proline-rich extensins, based on similarity to the maize extensin protein [19]. However, none of the 54 ESTs reported in TAIR (www.arabidopsis.org) from this genomic region are consistent with expression of At2g28671, but rather they all support expression of At2g28670. Furthermore, no promoter is predicted for the putative At2g28671 ORF using TSSP [20], whereas a strong promoter is predicted for At2g28670. We therefore concluded that At2g28670 is the most likely functional gene at this locus.
Expression of At2g28670 in wild-type, esb1-1 and esb1-2 plants was quantified in root and shoot tissue using quantitative RT-PCR (Figure 4A). At2g28670 was strongly expressed in roots of wild-type plants with little or no expression in shoots. Expression of At2g28670 was lost in both esb1-1 and esb1-2. Searching the Arabidopsis Gene Expression Database (AREX) [21],[22] for information on the expression pattern of At2g28670 revealed that expression of this gene is primarily in the root endodermis with lower levels of expression in the quiescent center, stele and cortex (Figure 4B).
To establish that loss-of-function of At2g28670 is driving the shoot ionomic phenotype observed in esb1-1 we measured the elemental composition of leaf material from esb1-1, esb1-2 and wild-type plants grown together. Using principal component analysis (PCA) of the leaf elemental composition of each genotype we determined that esb1-1 and esb1-2 plants clustered together and formed a single group that was distinct from wild-type Col-0 (Figure 5). Furthermore, esb1-1 and esb1-2 show the same changes in the elemental composition of leaf tissue compared to wild-type plants (Table 1).
Expression of At2g28670 is primarily in the root endodermis, the major site of suberin deposition in the primary root, and the localization of the Casparian strip. Staining of the aliphatic components of suberin using Fluoral Yellow revealed a clear increase in staining in roots from esb1-1, when compared to wild-type roots. In particular, increased staining is observed in the two rows of cells in the central portion of the root, suggesting increases in endodermal suberin (Figure 6A & B). Slightly elevated staining is also observed more diffusely in the root tip. Total aliphatic monomer content in both esb1-1 and esb1-2 was double that of wild-type roots (Figure 6C). However, there was no difference in total lignin content of roots of esb1-1 or esb1-2 (Figure 6D). Further analysis of the individual aliphatic monomer components of suberin revealed a doubling of all the acids, alcohols, ω-hydroxyacids, α,ω-diacids and ferulic acid components measured in both esb1-1 and esb1-2 compared to wild-type roots (Figure 7 and Supplemental Table 1).
Given that At2g28670 is primarily expressed in the root, and that loss-of-function of this gene results in a doubling of root suberin and a corresponding alteration in the leaf ionome, we hypothesized that root processes are responsible for the leaf ionomic phenotype of esb1. To test this hypothesis we performed a reciprocal grafting experiment in which we grafted wild-type scions onto esb1 rootstocks and esb1 scions onto wild-type rootstocks. Grafting of wild-type scions onto esb1 rootstocks (both esb1-1 and esb1-2) revealed that the leaf ionomic phenotype of esb1 is dependent on the rootstock. PCA of the leaf elemental composition indicated that plants with esb1 and wild type roots clustered separately, regardless of scion (Figure 8A & B).
Suberin has been proposed to form a barrier to water movement through the root apoplast. Elevated suberin, as found in esb1, should therefore enhance this barrier, reducing both water movement to the shoot and water loss from the root back to the soil. To test this hypothesis we measured both transpiration and wilting after water withdrawal in esb1. Peak daytime transpiration rates in esb1-1 and esb1-2 plants were approximately 15% of those of wild-type plants (Figure 9). The stomatal index (ratio of guard cells to epidermal pavement cells) was unaltered in esb1 compared to wild-type (Figure 10A), establishing that the reduction in transpiration in esb1 is not driven by a simple alteration in stomatal density. However, the stomatal pore width in esb1 was reduced by approximately 10% compared to wild-type (P<0.01) (Figure 10B), and the magnitude of this change is consistent with the reduction in transpiration observed.
To test the rate of wilting in esb1, watering was stopped after 5-weeks of growth and the plants were monitored for wilting. Both esb1-1 and esb1-2 showed reduced wilting after water withdrawal compared to the wild type plants (Figure 11). Using reciprocal grafting experiments we further determined that this enhanced wilting resistance is determined by the root. Plants with esb1 rootstock (both esb1-1 and esb1-2) grafted to wild-type scions showed reduced wilting similar to esb1 self-grafted plants, whereas plants with wild-type rootstock all wilted within 11 days (Figure 11). Water use efficiency (WUE) of a plant relates water transpired to biomass produced, and is an important parameter for the assessment of the impact of reductions in transpiration on plant productivity. We have established that elevated suberin in esb1 roots is related to a decrease in transpiration and increased water stress tolerance. These reductions in root-controlled water loss in esb1 cause a small (∼14%) reduction in dry biomass production, but an overall increased in the amount of biomass produced per unit of water transpired, with both esb1-1 and esb1-2 showing small (8%) but significant (P<0.01) increases in WUE (Figure 12).
Radial rather than axial transport is known to limit water loading by roots [23]. The force driving water across the root is provided by tension (negative pressure) created by transpiration, and flow is thought to be primarily apoplastic. However, during periods of reduced transpiration, at night or in dry or saline soils, it is root pressure, caused by osmotic water flow driven by the uptake of nutrients into the xylem that drives water transport to the shoots. This osmotic pathway for water is primarily cell-to-cell. It has been proposed that this “Composite Model” for water uptake allows plant roots to take up water under various conditions and balance it with demands [24]. However, because the Casparian strip is thought to be somewhat permeable to water, but not to mineral ions, transport of water and solutes are differentiated in the endodermal apoplast, with the Casparian strip forming the main barrier to radial ion movement [25]. However, to date no mutants with elevated root suberin have been described to help elucidate this model further.
The esb1 mutant described here has significantly higher root suberin compared to wild type plants, and grafting experiments establish that this increase in suberin is responsible for the observed changes in the shoot ionome. The final accumulation of mineral elements in the shoot would be expected to be a balance between the negative effect of suberin reducing radial apoplastic solute transport, and the positive effect of the reduced transpiration of esb1 on the xylem concentration of solutes. The reductions in total shoot Ca, Mn and Zn observed in esb1 suggest that a significant component of the radial root transport of these elements is via the apoplast. This evidence confirms the earlier conclusion of White [26], based on modeling of physiological and biochemical parameters, that a significant root apoplastic bypass pathway exists for Ca. It is also consistent with the observation that increased endodermal cytoplasmic Ca during cooling is inversely proportional to the level of suberin deposition [27]. Further, the conclusion by White and coworkers [28] that Zn may also reach the xylem via an apoplastic pathway is also supported by our data on esb1.
The increase in shoot S, K, As, Se and Mo in esb1 suggests that these elements may be primarily transported radially in the root via a cell-to-cell pathway, making them resistant to changes in endodermal suberin, and more prone to the influence of increasing xylem concentrations of solutes due to the reduced transpiration driven water flow in esb1. We also observe Na concentration increasing in the shoot of esb1, suggesting the absence of an apoplastic bypass pathway for this element at the low external Na concentrations present in the potting mix used in these studies. However, elevated suberin has been implicated in reducing Na accumulation and increasing Na tolerance in rice [29],[30]. Such evidence suggests that at low external Na apoplastic bypass flow is limited. However, at elevated external Na concentrations apoplastic bypass flow becomes significant, and reduction in this flow can lead to elevated Na tolerance.
It has previously been reported that Ca and Mg shoot concentrations are positively correlated across taxa [31], within-species and genetically [32]. We also observe a strong correlation between Ca and Mg in both wild-type and esb1 (R2 = 0.75 and 0.68, respectively. Data taken from trays 533, 534 and 535, n = 30 for each genotype; see www.ionomicshub.org). Such correlations suggest that the mechanisms driving Ca and Mg accumulation are related. However, the fact that in esb1 Ca is reduced without significantly changing Mg suggests that the regulatory mechanisms controlling Ca and Mg can be separated.
Elevated root suberin in esb1 is also associated with a root-dependent increase in time to wilting during water stress. Resistance to wilting can be achieved via a reduction in water loss to the environment, from both the root and leaves, or through an increase in the ability to take up water from the soil. Delayed wilting in esb1 appears to be related to the reduced stomatal aperture and reduced daytime transpiration rates in this mutant. However, grafting experiments establish that this delayed wilting in esb1 is a root-dependent phenomenon, suggesting signaling from root to shoot. A possible explanation for this signal is that roots of esb1 are constitutively responding to water stress, due to increased suberin causing enhanced hydraulic resistance to radial transport of water. This stress is transmitted to the shoot via ABA, or directly as a hydraulic signal [33]. Such a model remains to be tested.
Reduced transpiration is often correlated with reduced biomass accumulation, as observed here for esb1. However, biomass reduction in esb1 is less than the reduction in transpiration, leading to an overall increase in water use efficiency of esb1. Increased root suberin may therefore present new opportunities for developing enhancing drought resistance in crop plants.
Here we establish that loss-of-function of the endodermally expressed gene At2g28670 leads to a doubling of all the aliphatic components of root suberin. The enzymatic function of the protein encoded by At2g28670, and how this function affects suberin, remains an open question. Importantly, the elevated root suberin observed in two independent At2g28670 loss-of-function mutants (esb1-1 and esb1-2) directly affects the shoot ionome, causing several changes, including a 50% reduction in Ca and a 40% increase in Na. These changes are also associated with decreased transpiration and increased wilting resistance.
Overall, these observations provide strong experimental support for the standard model that suberin acts as an extracellular transport barrier limiting apoplastic radial transport of water and solutes. Our observations suggest that elevation of root suberin may represent another approach to the development of drought resistant crops with improved WUE. Furthermore, manipulation of suberin may also provide new opportunities for the development of plant-based foods with altered mineral nutrient contents.
All A. thaliana lines were obtained from the ABRC or Lehle seeds. All T-DNA lines analyzed were homozygous for the T-DNA insertion. Unless stated otherwise all plants were grown in a controlled environment, 8 h light∶16 h dark (90 µmol m−2 s−1 photosynthetically active light) and 19 to 22°C as previously describe [14]. Briefly, seeds were sown onto moist soil (Sunshine Mix LB2; Carl Brehob & Son, Indianapolis, Indiana, United States) with various elements added at subtoxic concentrations (As, Cd, Co, Li, Ni, Pb, and Se [14]) and stratified at 4°C for 3 d. Plants were bottom-watered twice per week with 0.25× Hoagland solution in which iron was replaced with 10 µM Fe-HBED [N,N′-di(2-hydroxybenzyl)ethylenediamine-N,N′-diacetic acid monohydrochloride hydrate; Strem Chemicals, Inc., http://www.strem.com]. For elemental analysis after 5-weeks, plants were nondestructively sampled by removing one or two leaves. The plant material was rinsed with 18 MΩ water and placed into Pyrex digestion tubes.
Tissue samples were dried at 92°C for 20 h in Pyrex tubes (16×100 mm) to yield approximately 2–4 mg of tissue for elemental analysis. After cooling, seven of approximately 100 samples from each sample set were weighed. All samples were digested with 0.7 ml of concentrated nitric acid (OmniTrace; VWR Scientific Products; http://www.vwr.com), and diluted to 6.0 ml with 18 MΩ water. Elemental analysis was performed with an ICP-MS (Elan DRCe; PerkinElmer, http://www.perkinelmer.com) for Li, B, Na, Mg, P, S,K, Ca, Mn, Fe, Co, Ni, Cu, Zn, As, Se, Mo, and Cd. All samples were normalized to calculated weights, as determined with an iterative algorithm using the best-measured elements, the weights of the seven weighed samples, and the solution concentrations, described in [14], and implemented in the PiiMS database [34]. An Excel implementation of this algorithm is available at www.ionomicshub.org along with validation data sets.
DNA microarray-based BSA was performed as previously described [17],[35]. Briefly, SFPs were identified between Col-0 and Ler-0 by hybridizing labeled genomic DNA from each one of the accessions to Affymetrix ATH1 microarrays and comparing them to Col-0 hybridizations downloaded from http://www.naturalvariation.org/xam. Two genomic DNA pools from an F2 population of a cross between Ler-0 and the esb1-1 mutant in the Col-0 background were created and hybridized to separate DNA microarrays. Each one of the pools contained plants with either shoot Ca and B contents similar to Col-0 (“control” pool) or low shoot Ca and B contents similar to esb1-1 (“low Ca and B” pool). At loci unlinked to the low Ca and B phenotype, the pools should have equivalent amounts of each genotype, and the hybridization signal at each SFP should be intermediate between the two parent accessions, for an average difference between the two DNA microarrays of zero. At linked loci, the difference between the two DNA pools should be approximately two-thirds the difference between the parent accessions. By smoothing the signal across multiple SFPs, noise is reduced and the peak of the differences in hybridization signal will correspond to the chromosomal region of the loci controlling the low Ca and B trait. Raw hybridization data (.CEL files) for each probe on the ATH1 DNA microarrays used in these experiments have been submitted to the Gene Expression Omnibus and are accessible through GEO Series accession number GSE15655 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE15655). For the deletion mapping, DNA was extracted from Col-0 and esb1-1 plants and hybridized to the ATTILE 1.0R array using the same protocols as described above. After quantile normalization, the difference in hybridization intensity for each probe between 9 and 13 Mb was visually inspected to identify the causal locus.
Seedlings were grafted as previously described [16]. Plants were harvested for shoot elemental analysis 4 weeks after transfer to soil. Postharvest analysis of graft unions was performed under the stereoscope to identify any adventitious root formation from grafted individuals. Individuals with adventitious roots emerging at or above the graft union or without a clear graft union were eliminated from subsequent analyses.
Plants were first analyzed by ICP-MS and further used to determine esb1 transcript levels as described previously [16]. For esb1 (At2g28670) transcript quantification the following primers were used: forward primer 5′-ATGTCCCTTTCCTCGTTGGA-3′ and reverse primer 5′-GCCACTAGCAACAGGGAAACC-3′. Three reactions were done per biological sample and three independent replicate samples per genotype were used to evaluate the transcript abundance of esb1. Data was analyzed using the SDS software (Applied Biosystems version 1.0), following the method of Livak and Schmittgen [36]. CT values were determined based on efficiency of amplification. The mean CT values were normalized against the corresponding ACTIN 1 gene (At2g37620) and CT values calculated as (CT esb1- CT Actin1). The expression of esb1 was calculated using the 2∧( ΔCT) method. The final error was estimated by evaluating the 2∧( ΔCT) term using 2∧( ΔCT) plus standard deviation and 2∧( ΔCT) minus the standard deviation [36].
In vivo staining of suberin was performed using Fluorol Yellow, following the method of [37]. Roots of 1 week old seedlings, grown on 1/2 MS agar plates, were incubated in a freshly prepared 0.01%(w/v) solution of Fluorol Yellow 088 (Sigma) in lactic acid at 70°C for 1 h. Parallel stained roots of Col-0 and esb1-1 mutant plants were placed adjacent on a microscope slide and observed under UV-light using an Axioplan microscope (Zeiss, Germany). For quantitative chemical suberin analysis [38] roots of 35 d old soil-grown plants were incubated in 1% (v/v) cellulase (Celluclast, Novozymes, Germany), 1% (v/v) pectinase (SIHA, Novozymes, Germany) in 10 mM citric buffer pH 3 containing 10 mM NaN3. After 10 d the root cell wall material was washed and incubated in 10 mM sodium tetraborate (pH 9.0) for 2 d and extracted with chloroform∶methanol (1∶1, v/v) to remove unbound lipids. The cell wall material of 4–5 plants was depolymerized by transesterification in 1 ml borontrifluorid in methanol (10%, Fluka) for 16 h at 70°C. After adding 10 µg dotriacontane (internal standard) the methanolysate was transferred into 2 ml saturated NaHCO3/H2O and suberin monomers were subsequently extracted in chloroform. Free hydroxyl and carboxyl groups were derivatized with bis-(N,O-trimethylsilyl)-tri-fluoroacetamide (BSTFA, Macherey-Nagel, Germany) (20 µl BSTFA+20 µl Pyridin, 40 min, 70°C) prior to GC–MS/FID analysis. Suberin monomers were injected on-column (DB-1 (J&W Scientific), 30 m×0.32 mm, 0.1 µm), separated and identified on an Agilent 6890N gas chromatograph coupled with an Agilent 5973N quadrupole mass selective detector (70 eV, m/z 50–700). The following temperature gradient was applied: 2 min at 50°C, 10°C/min to 150°C, 1 min at 150°C, 3°C/min to 310°C, 15 min at 310°C. Quantitative determination of the components was carried out with an identical GC-system coupled with a flame ionisation detector based on the internal standard.
Lignin content in roots was determined by the acetyl bromide method using 3–5 mg root material as used for suberin analysis. 1 ml acetyl bromide/acetic acid (1∶3) was added to the root cell wall material and incubated at 70°C for 30 min. After cooling to 15°C 0.9 ml 2 M NaOH and 5 ml acetic acid was added. Finally 0.1 ml 7.5 M hydroxylamine-HCl and acetic acid was added to a final volume of 10 ml. The lignin content was calculated from the A280 using the extinction coefficient 24 g−1 L cm−1.
Plants were grown for 5 weeks in 2 inch pots with 12 hr of photosynthetically active light (80–100 µmol/m2/s), with mean day and night temperatures of 22 and 18°C, respectively. For analysis of transpiration rates, pots were covered with plastic wrap (Saran wrap) to avoid water loss from the soil, and placed on one of twenty balances (EK-410i, A&D) to monitor changes in weight. Weights of pots were automatically recorded using balances connected to computers through WinWedge software (TAL technologies Inc.) at 5 min intervals for 2 days and 3 nights. A total of six to seven plants per genotype were analyzed. At the end of the experiment total leaf area for each plant was determined by digitally recording images of all leaves and using ImageJ [39] to determine leaf area.
Five week old plants grown under the conditions described above for measurement of transpiration were used for measuring stomatal index and aperture. To determine stomatal index, stomata were observed on the abaxial surface of leaves using a scanning electron microscope (JSM-840, JEOL) at 250× magnification. Stomatal index was calculated as the ratio of the number of stomata to the total number of cells (epidermal cells and stomata) in an area of 0.18 mm2. To determine stomatal aperture we modified the protocol of Hilu and Randall [40]. Briefly, clear nail polish was applied to the abaxial surface of leaves, peeled when dry and stomatal impression in the nail polish observed under a light microscope (Vanox-S, Olympus) at 400× magnification and images recorded digitally. Stomatal aperture width was measured using ImageJ software [39].
Grafting was done as described above. To measure wilting resistance grafted plants were transferred to soil and grown for three weeks, after which watering was stopped and plants observed for symptoms of wilting for 11 days.
Plants were grown in inverted brown 50 mL Falcon tubes (Greiner Bio-One) filled with soil (1∶1 proportion of Premier ProMix PGX and calcined clay (Turface MVP)) and covered with a screw cap with a mesh insert to allow for water uptake. A 4–5 mm diameter hole was made on the narrow end of the Falcon tube for seed germination and growth. Plants were grown in a growth chamber with 10 hr of photosynthetically active light (80–100 µmol/m2/s), with mean day and night temperatures of 22 and 18°C, respectively. Watering was done with alternating clean or fertilized (Miracle Gro Excel, Scotts) water. To monitor water use tubes containing plants, and also control tubes with no plants, were weighted (Balance- GH-252, A&D) before and after each watering, over a period of five weeks. At the end of the experiment whole shoots were harvested for each plant, dried at 70°C and the dry weight determined. Water use efficiency was calculated by dividing total shoot dry weight by the amount of water utilized by the plants over the complete growth cycle.
|
10.1371/journal.pgen.1006991 | Defective erythropoiesis caused by mutations of the thyroid hormone receptor α gene | Patients with mutations of the THRA gene exhibit classical features of hypothyroidism, including erythroid disorders. We previously created a mutant mouse expressing a mutated TRα1 (denoted as PV; Thra1PV/+ mouse) that faithfully reproduces the classical hypothyroidism seen in patients. Using Thra1PV/+ mice, we explored how the TRα1PV mutant acted to cause abnormalities in erythropoiesis. Thra1PV/+ mice exhibited abnormal red blood cell indices similarly as reported for patients. The total bone marrow cells and erythrocytic progenitors were markedly reduced in the bone marrow of Thra1PV/+ mice. In vitro terminal differentiation assays showed a significant reduction of mature erythrocytes in Thra1PV/+ mice. In wild-type mice, the clonogenic potential of progenitors in the erythrocytic lineage was stimulated by thyroid hormone (T3), suggesting that T3 could directly accelerate the differentiation of progenitors to mature erythrocytes. Analysis of gene expression profiles showed that the key regulator of erythropoiesis, the Gata-1 gene, and its regulated genes, such as the Klf1, β-globin, dematin genes, CAII, band3 and eALAS genes, involved in the maturation of erythrocytes, was decreased in the bone marrow cells of Thra1PV/+ mice. We further elucidated that the Gata-1 gene was a T3-directly regulated gene and that TRα1PV could impair erythropoiesis via repression of the Gata-1 gene and its regulated genes. These results provide new insights into how TRα1 mutants acted to cause erythroid abnormalities in patients with mutations of the THRA gene. Importantly, the Thra1PV/+ mouse could serve as a preclinical mouse model to identify novel molecular targets for treatment of erythroid disorders.
| Patients with mutations of the THRA gene exhibit erythroid disorders. The molecular pathogenesis underlying erythroid abnormalities is poorly understood. In Thra1PV/+ mice expressing a dominant negative mutant TRα1PV, we found abnormal red blood cell indices similar to patients. Total bone marrow cells, the clonogenic potential of erythrocytic progenitors, and terminal differentiation of erythrocytes were markedly decreased in Thra1PV/+ mice. We elucidated that Gata-1, a key erythroid gene, was directly positively regulated by TRα1. The erythroid defects in Thra1PV/+ mice were due, at least partly, to the TRα1PV-mediated suppression of the Gata-1 gene and its down-stream target genes. Over-expression of Gata-1 rescued impaired terminal differentiation. Our studies elucidated molecular mechanisms by which TRα1 mutants caused erythroid disorders in patients. The present study suggests that therapies aimed at GATA1 could be tested as a potential target in treating erythroid abnormalities in patients.
| Thyroid hormones have long been known to play an important role in erythropoiesis. Early in vitro studies demonstrated that L-thyroxine (T4) stimulates the biosynthesis of hemoglobin [1]. Experimental animal models also showed that T4 enhances red blood cell formation and stimulates hemoglobin synthesis [2]. Studies in humans have shown a causal association between hypothyroidism and anemia [3]. Persons with subclinical hypothyroidism had lower hemoglobin levels [4, 5] and a higher prevalence of anemia than euthyroid persons. Treatment of patients with subclinical hypothyroidism with thyroid hormone resulted in a significant increase in hemoglobin content [6] or erythropoietin levels [7]. However, how hypothyroidism results in erythroid disorders at the molecular level remains largely unknown.
Recently, patients with mutations of the thyroid hormone receptor α gene (THRA) have been reported to exhibit some of the classical symptoms and signs of hypothyroidism with impaired growth and delayed bone development [8–11]. These patients also display anemia. These findings suggested that erythropoietic disorders in humans are mediated by TRα1 mutants and support the critical role of TRα1 in erythropoiesis. The availability of a mouse model (Thra1PV/+ mice), harboring a mutated TRα1 (designated as TRα1PV) [12], has made it possible for us to elucidate the role of TRα1 mutants in erythroid disorders. TRα1PV has a C-terminal mutated sequence (398-PPFVLGSVRGLD- 409) [12], similar to the truncated C-terminal sequence in two patients (398-PPTLPRGL -405) [9]. The PV mutation was first identified from a patient with severe resistance to thyroid hormone (RTHβ), characterized by elevated thyroid hormone levels accompanied by normal TSH, short stature, goiter, and tachycardia [13]. The PV mutated sequence was targeted to the Thra gene at the corresponding position as in THRB gene to assess the functional consequence of TRα1 mutations at the time when no patients with the mutations of the THRA gene was discovered. Thra1PV/+ mice exhibit displayed phenotype of hypothyroidism as in the patients with severe growth retardation [12] and delayed bone development [12, 14, 15]. The Thra1PV/+ mouse has been used as a preclinical model to test the effectiveness of T4 treatment for the correction of impaired bone development due the actions of mutated TRα1 [16]. These studies further validate the usefulness of the Thra1PV/+ mouse to understand how mutations of the THRA gene result in deleterious abnormalities in patients.
In the present study, we first characterized the erythroid phenotypes in Thra1PV/+ mice and showed that Thra1PV/+ mice exhibit anemia with decreased red blood cells and reduced hemoglobin content similar to patients with mutations of the THRA gene [11]. We further identified the genes that were abnormally regulated by TRα1PV, resulting in defective erythropoiesis. Thus, our study has provided direct molecular evidence to show that mutations of the THRA gene could impair erythropoiesis and has uncovered novel molecular actions of TRα1 in the erythroid differentiation and development.
We first analyzed peripheral blood composition to characterize erythropoietic phenotypes in Thra1PV/+ mice. Complete blood count revealed that major indices for erythrocytes, such as red blood cell count (RBC), hemoglobin levels (Hb) and hematocrit (HCT), were significantly lower in Thra1PV/+ mice than in wild-type (WT) mice. As shown in Fig 1, RBC, Hb, and HCT were decreased by 14% (panel A), 13% (panel B), and 10% (panel C), respectively. Other erythrocyte peripheral indices, namely, the mean corpuscular hemoglobin concentration (MCHC), red cell distribution width (RDW), and platelets (PLT), were 3% (panel D), 3% (panel E), and 30% (panel F) lower in Thra1PV/+ mice than in WT mice. These results indicated that mutations of the Thra gene could affect different lineages leading to defects.
After birth and throughout adult life, the bone marrow remains the major hematopoietic organ in mice [17]. We therefore examined the histology of H & E stained femur sections to assess the cellularity of the bone marrow. As shown in Fig 2A, fewer bone marrow cells with a higher fat deposit were apparent in Thra1PV/+ mice than wild-type mice (Fig 2A, compare panel b with panel a). By counting cell numbers, we found that total bone marrow cells were decreased 57% in Thra1PV/+ mice (Fig 2A-c). We have measured the fat areas in the bone marrow of wild-type mice (n = 3) and Thra1PV/+ mice (n = 3). The quantitative data is now shown as Fig 2A-d, indicating that the fat (% of total area) was 6.2-fold higher in the bone marrow of Thra1PV/+ mice than in wild-type mice.
We next assessed which abnormal population of progenitors in the bone marrow cells of Thra1PV/+ mice contributed to anemia. It is known that a small population of hematopoietic stem cells (HSC) gives rise to multipotent progenitors (MPP), which subsequently differentiate into common myeloid population cells (CMP population cells) and common lymphoid progenitors (CLP). CMP is differentiated into megakaryocytic/erythroid (MEP) and granulocyte-myeloid (GMP) progenitors. Because patients with mutations of the THRA gene exhibit anemia, we focused on the analysis of the lineage derived from MEP, and its subsequent downstream progenitors: burst-forming unit-erythroid (BFU-E) to colony-forming unit-erythroid (CFU-E), then to erythroblasts, and ultimately to mature erythrocytes [18–20]. To assess the effect of TRα1PV mutation on the number of progenitors in the MEP lineage, we analyzed the progenitors by specific cell surface markers in the bone marrow using flow cytometry. Erythroid cells at different developmental stages can be identified by different cell surface markers: Ter119 (erythroid specific glycophorin) and CD71 (the transferrin receptor). Ter119 expression increases as maturation of erythrocyte progresses [21, 22]. CD71 is the transferrin receptor that expresses at high levels in early erythroid precursors, but its levels decrease toward erythroid maturation [21, 22]. cKit (CD117) is a cytokine receptor expressed on the surface of hematopoietic stem cells, MPP, and CMP. Stem cells antigen 1 (Sca1) is expressed in HSC [19, 23]. Using these specific cell markers, we identified which sub-populations were altered in the bone marrow cells of Thra1PV/+ mice.
CFU-E progenitors reside in the lineage negative (Lin-)Ter119-CD71+cKit+ populations of the bone marrow (Fig 2B, panel a). CFU-E progenitors were decreased 24% in Thra1PV/+ mice compared with wild-type mice (Fig 2B, panel d). Mature red blood cells and reticulocyte precursors exhibited characteristic Ter119+CD71- expression on the cell surface (Fig 2B, panel b). We found that erythrocytes were decreased 56% in the bone marrow of Thra1PV/+ mice (Fig 2B, panel e). CFU-Mk progenitors and megakaryocytes reside in Ter119-Sca1-CD41+CD61+ population (Fig 2B panel c). The only difference between CFU-Mk and megakaryocyte is that CFU-Mk expresses cKit as a marker in the progenitors. We found that CFU-Mk progenitors and megakaryocytes were decreased 19% and 52%, respectively, in the bone marrow of Thra1PV/+ mice (Fig 2B, panels f and g, respectively). Notably, the extent of differences between WT and Thra1PV/+ mice was greater in the CFU-E and erythrocytes than in CFU-Mk and megakaryocytes.
We further carried out in vitro colony forming unit assays. BFU-E progenitors were defined as cells that colonize after 14 days under defined-in vitro culture conditions (Fig 2C, panel a). CFU-E progenitors were colonized cells after 2 days in defined culture conditions (Fig 2C, panel b). CFU-Mk progenitors were counted after 7 days in defined culture conditions (Fig 2C, panel c). BFU-E, CFU-E, and CFU-Mk (expressed as relative colonies versus total bone marrow cells) were decreased 68% (Fig 2C-d), 62% (Fig 2C-e), and 68% (Fig 2C-f), respectively, in Thra1PV/+ mice as compared with WT mice. We further carried out colony assays for CFU-GEMM and CFU-GM. The number of CFU-GEMM and CFU-GM colonies as shown in Fig 2D-a and 2D-c, respectively, was 70% and 60% lower, respectively, in Thra1PV/+ mice than in WT mice (Fig 2D-b and 2D-d). These results imply that the capacity of progenitor cells to differentiate from MEP to erythroblasts as well as to megakaryocytes was impaired, leading to erythroid disorders in Thra1PV/+ mice.
TRα1PV is a dominantly negative mutant and cannot bind T3 [12]. That the capacity of the progenitors to form BFU-E, CFU-E, and CFU-Mk was impaired in the bone marrow of Thra1PV/+ mice prompted us to ascertain whether the differentiation from MEP to downstream BFU and CFU was regulated by thyroid hormones. We therefore rendered Thra1PV/+ mice hypothyroid by treating them with PTU and rendered the PTU-treated mice hyperthyroid by T3 treatment. As shown in Fig 3A, PTU treatment was effective in lowering serum total T3 and T4 levels (bar 2 in panels a and b, respectively) in WT mice. In line with the lowering of T3 and T4, TSH was highly elevated (bar 2 in Fig 3A-c). T3 treatment of PTU-treated WT mice led to highly elevated T3 (Fig 3A-a, bar 3) and so suppressed TSH levels (Fig 3A-c, bar 3). Treatment of Thra1PV/+ mice with PTU followed by T3 injection led to similar changes as shown for WT mice (Fig 3A-a, 3A-b and 3A-c, bars 4–6). These findings are consistent with the earlier reports that the feedback loop in the pituitary-thyroid axis was not affected by expressing TRα1PV mutant in Thra1PV/+ mice [12].
We next isolated bone marrow cells from hypothyroid (PTU-treated) and hyperthyroid (T3-treated) mice and carried out colony forming assays. No significant differences in BFU-E (Fig 3B-a, bars 1–2), CFU-E (Fig 3B-b, bars 1–2), and CFU-Mk (Fig 3B-c, bars 1–2) were detected between hypothyroid and euthyroid mice. However, T3 treatment increased the numbers of BFU-E (Fig 3B-a, bar 3), CFU-E (Fig 3B-b, bar 3), and CFU-Mk (Fig 3B-c, bar 3) by 42%, 77% and 25%, respectively from hypothyroid to hyperthyroid mice. In contrast, in Thra1PV/+ mice, compared to WT mice, there was markedly decreased BFU-E (Fig 3B-a), CFU-E (Fig 3B-b), and CFU-Mk (Fig 3B-c) (compare bar 4–6 to bars 1–3). Moreover, the extent of decreases in Thra1PV/+ mice were not affected by T3 treatment (bars 4–6). These results indicated that colony forming units derived from MEP were stimulated by T3 in WT mice, but were not affected by T3 in Thra1PV/+ mice due to the actions of dominant negative TRα1PV mutant.
To further confirm that TRα1PV mutant acted to inhibit the differentiation in the erythroid lineage, we isolated lineage depleted bone marrow cells (Lin-BM) by eliminating mature lineage cells including T cells, B cells, macrophages, granulocytes, and erythrocytes. Using Lin- BM, we compared the maturation of erythrocytes in WT and Thra1PV/+ mice using an in vitro terminal erythropoiesis system [24]. Using an equal number of total bone marrow cells from WT (Fig 4A-a) and Thra1PV/+ mice (Fig 4B-a), we found 40% and 31%, respectively, of Ter119+ with low FSC population (boxed in red). As shown by May-GrÜnwald—Giemsa staining (Fig 4A-d and 4B-d), these populations were enriched with mature erythrocytes (marked by red arrows). After depletion of mature lineage cells as evidenced by markedly reduction of Ter119-positive cells (panel b in Fig 4A and 4B) and enucleated erythrocytes by May-GrÜnwald—Giemsa staining (panel e in Fig 4A and 4B), progenitor cells were stimulated by erythyropoietin to undergo terminal erythropoiesis [24]. After culturing for 3 days, Ter119-positive cells associated with low FSC fraction were detected (panel c in Fig 4A and 4B). The enucleated cells were visualized by May-GrÜnwald—Giemsa staining (panel f in Fig 4A and 4B). The gated Ter119-positive cells associated with low FSC fraction (boxed in red) in panel c were quantified: 30% and 17% of mature erythrocytes were detected for WT and Thra1PV/+ mice, respectively. These results represent a 47% reduction in the capacity of Lin- MB cells from Thra1PV/+ mice to mature into erythrocytes (Fig 4C).
To investigate the molecular mechanisms by which TRα1PV mutant induced erythropoietic disorders, we first analyzed the expression of a major regulator of erythropoiesis, the Gata1 gene (erythroid transcription factor; GATA-binding factor 1). GATA1 is a member of the GATA transcription factor family and is essential for erythroid development by regulating a large ensemble of genes that mediate both the development and function of red blood cells [25, 26]. The loss of Gata1 expression leads to erythroid maturation arrest and embryonic lethality due to anemia [27–31]. We found the expression of the Gata1 gene was ~50% lower in the bone marrow of Thra1PV/+ mice than in WT mice (Fig 5A-a, compare bar 2 with bar 1). Direct western blot analysis shows that the GATA1 protein level in the bone marrow was lower than that in the spleen of WT mice (compare lane 2 with lane 1, Fig 5A-b). However, when the bone marrow cell lysates of WT mice were first enriched by immunoprecipitation followed by western blot analysis (Co-IP), GATA1 proteins were detected (lane 3, Fig 5A-b). In contrast, under identical experimental conditions, GATA1 proteins were not detected (lane 4), indicating that the GATA1 protein level in the bone marrow of Thra1PV/+ mice was lower than that of WT mice. Lanes 5 and 6 were the corresponding negative controls in that an irrelevant IgG was used in the immunoprecipitation step. That no signals were detected in lanes 5 and 6 indicated that the protein detected by Co-IP was specific. GATA1 regulates the expression of the Klf1 gene, which drives erythropoiesis by affecting its downstream target genes critical for maturation of erythrocytes [25, 32, 33]. Consistent with decreased expression of the Gata1 gene, the expression of the Klf1 gene was 53% lower in the bone marrow of Thra1PV/+ mice (Fig 5B, bar 2 versus bar 1). We also analyzed the expression of KLF1 and GATA1 target genes, such as the β-globin (beta major globin) and the dematin genes (erythroid membrane and cytoskeleton related gene). The β-globin protein along with α-globin makes up the most common form of hemoglobin in adult humans. Dematin is a relatively low abundance actin binding and bundling protein associated with the spectrin—actin junctions of mature erythrocytes. Dematin binds to spectrin and dynamically regulates red cell membrane mechanical function [34]. Consistent with the reduced expression of the Klf1 gene, the expressions of the β-globin and dematin genes were also repressed by 70% and 50%, respectively, in the bone marrow of Thra1PV/+ mice (Fig 5C and 5D). We further analyzed the expression of CAII, band 3 and eALAS genes, which are regulated by GATA-1 during erythropoiesis. Moreover, these three genes are known to be directly regulated by TRα1 and T3 in birds [35–37]. The expression of these three genes was inhibited 93%, 90% and 86%, respectively, in the bone marrow of Thra1PV/+ mice (Fig 5E, 5F and 5G, respectively). Taken together, these data indicate that suppression of Gata-1 gene and the known T3- target genes led to impaired erythropoiesis.
That the expression of the Gata1 gene was suppressed by TRα1PV prompted us to ascertain whether the Gata1 gene was directly regulated by TR/T3. We searched for putative thyroid hormone response elements (TREs) in the GATA1 hematopoietic regulatory domain [38], located upstream of the transcription starting site (TSS) of the Gata1 gene (Fig 6A). We also searched for putative TREs in the +4691 intronic sequence between exon I and exon II (Fig 6A). In this -4096 to +4691 region, we found putative TREs-containing regions with the core consensus sequences of the hexa-nucleotide “half-site” (A/G)GGT(C/A/G)A (T#1 –T#8, Fig 6B-I). To identify relevant TREs, we used chromatin immunoprecipitation (ChIP) to ascertain the binding of TRα1 to these eight putative TREs in the bone marrow of WT and Thra1PV/+ mice. When anti-TRα1 antibodies were used, a significantly higher binding of TRα1 to TRE1 (AGTGGGGTCCATT in the region +1410 to +1549 bp) were found in the bone marrow of WT mice than when anti-IgG antibodies (negative controls) were used (bar 3 versus bar 1, Fig 6B-II-a). In the bone marrow of Thra1PV/+ mice, when specific anti-TRα1PV antibodies (rabbit polyclonal antibody T1) were used, a significantly higher binding of TRα1PV to TRE1 than when anti-IgG antibodies (negative controls) were used (bar 6 versus bar 2, Fig 6B-II-a). Similar significantly higher binding of TRα1 (bar 3 versus bar 1, Fig 6B-II-b) and TRα1PV (bar 6 versus bar 2, Fig 6B-II-b) to TRE2 (CCCACGGAGATTCCTGT, in the region -3699 to -3325 bp) were also detected. However, we did not find specific binding of TRα1 and TRα1PV to other regions containing putative TREs (i.e., T#1, T#2, T#3, T#5, T#7, and T#8, Fig 6B-I). These ChIP results suggested that TRα1 and TRα1PV could bind directly to these TRE-containing regions in TRE1 (region T#4) and TRE2 (region T#6) on the intronic region and the proximal promoter, respectively, of the Gata1 gene.
We next used electrophoretic mobility shift assay (EMSA) to demonstrate directly the binding of TRα1 and TRα1PV to TREs in TRE1 and TRE2 regions (Fig 6C-I). As shown in Fig 6C-II-a, TRα1 prepared by in vitro transcription/translation system, bound to TRE1 (the TRE sequence is shown in Fig 6C-I) as heterodimers with the retinoid X receptor (RXR) (lanes 5 and 6 show the increasing concentrations of TRα1). The TRE1-bound TRα1 was supershifted by anti-TRα1 antibodies (lane 7), but not by control anti-IgG antibodies (lane 8). In the presence of unlabeled TRE, no binding to 32P-labeled TRE1 was detected (lane 9). These data indicate that TRE1 bound specifically to TRα1 as heterodimers with RXR. Similarly, TRα1PV also bound to TRE1 as heterodimers with RXR (lanes 11–12), which was supershifted by anti-TRα1PV (lane 13). No specific binding to labeled TRE in the presence of unlabeled TRE was observed (lane 15). Lanes 16–19 were positive controls using a labeled F2 (TRE with inverted two half-sites). Moreover, when TRE1 from AGTGGGGTCCATT was mutated to AGACGGGTCTGTT, no binding was detected by EMSA (Fig 6C-II-b; compare lanes 11–16 with lanes 3–8 in which wild-type TRE was used in EMSA). These EMSA results indicate that TRE1 bound directly and specifically to TRα1 and TRα1PV. Using similar EMSA, we also found specific binding of TRα1 (Fig 6C-III-a, lanes 5–9) and TRα1PV (Fig 6C-III-a, lanes 11–15) to TRE2. The binding was further confirmed by mutational analysis in which, when TRE2 was mutated from CCCACGGAGATTCCTGT to TTTGAGGAGATCTTCTC, no binding was detected by EMSA (Fig 6C-III-b; compare lanes 11–16 with lanes 3–8 in which wild-type TRE was used in EMSA). Taken together, these results indicate that we have uncovered two specific TRα1 and TRα1PV binding TREs in the promoter of the Gata1 gene.
To assess whether these two TREs mediated the TRα1-dependent regulation of the transcription of the Gata1 gene, we constructed reporters in which the expression of luciferase is mediated by TRE1 or TRE2. We cloned the 0.140 kb fragment containing the TRE1 (Fig 7A-I-a and 7A-I-b) and 1.553 kb fragment containing the TRE2 into the pGL4.23 luciferase plasmid (Fig 7A-I-b and 7A-I-c, respectively). It is of interest to note that in the 1.553 kb fragment, one GATA box and two E-boxes, enhancers to regulate the transcription of the Gata1 gene, were also present [39]. These two Luc-reporters were transfected into human erythroleukemia 562 (K562) cells, which express endogenous TRα1 [40]. Indeed, using a Luc-reporter containing palindromic TRE (Pal-Luc), the reporter activity mediated by the endogenous TRα1 was detected in the presence of T3 (Fig 7A-II-a, bar 2 versus bar 1). When the TRα1PV expression vector was transfected into K562 cells, the Pal-Luc reporter activity was suppressed by dominant negative activity of TRα1PV (bar 4 versus bar 2). Transfection of the TRα1 expression plasmid led to an additional 15-fold activation of Pal-Luc reporter activity (bar 6 versus bar 2). This TRα1/T3-mediated activation was repressed by the transfection of the TRα1PV expression plasmid (bar 8 versus bar 6). These results indicate that K562 cells are a good model cell line to evaluate the activity of TRE1 and TRE2 in the regulation of the Gata1 gene by TRα1.
In contrast to Pal-Luc, in the presence of T3, a small, but significant repression of TRE1-Luc activity was observed (Fig 7A-II-b, bar 2 versus 1). Interestingly, transfection of TRα1PV led to a 1.3-fold activation of TRE1-Luc activity (bar 4 versus 2). When TRα1 expression plasmid was transfected into K562 cells, more suppression of TRE1-Luc activity than by endogenous TRα1 was observed (bar 6 versus 2). Remarkably, co-transfection of the TRα1PV expression plasmid with TRα1 expression plasmid into K562 cells resulted in a 2.5-fold activation of TRE1-Luc activity (bar 8 versus bar 6). These results demonstrated that TRE1 mediated the negative regulation activity of TRα1. These findings are reminiscent of the regulation of the Tshα common subunit (α-SU) gene by TR/T3 in the pituitary of Thra1PV/+ mice [12].
In contrast to TRE1-Luc activity, the luciferase activity mediated by TRE2 was activated two-fold by T3 in K562 cells (Fig 7A-II-c, bar 2 versus bar 1). This T3-stimuated reporter activity was totally abolished by transfection of the TRα1PV expression plasmid (bar 4 versus bar 2). Transfection of the TRα1 expression plasmid led to an additional 1.9-fold activation of the reporter activity (bar 6 versus bar 2). This T3-activated reporter activity mediated by TRα1 was totally abolished by co-transfection of the TRα1PV expression plasmid (bar 8 versus bar 6). These results indicate that TRE2, via T3-TRα1, mediates the positive regulation of the Gata1 gene transcription.
The identification of one positive TRE and one negative TRE in the promoter prompted us to assess the overall regulation by TRα1of the Gata1 gene in the bone marrow. We rendered WT and Thra1PV/+ mice hypothyroid by treating them with PTU and then some PTU-treated mice with T3 to make them hyperthyroid (see also Fig 3). The expression of the Gata1 mRNA was ~50% lower in hypothyroid WT mice than in euthyroid WT mice (Fig 7B-a, bar 3 versus bar 1). The expression of the Gata1 mRNA was ~3-fold higher in hyperthyroid WT mice than in hypothyroid WT mice (Fig 7B-a, bar 5 versus bar 3). These results indicate that the Gata1 gene was positively regulated by T3. In Thra1PV/+ mice, Fig 7B-a also shows that the expression of the Gata1 mRNA in the bone marrow of euthyroid Thra1PV/+ mice was ~50% lower than that in euthyroid WT mice (bar 2 versus 1). The Gata1 mRNA expression in hypothyroid Thra1PV/+ mice was also lower than that in hypothyroid WT mice (bar 4 versus 3). However, the expression of the Gata1 mRNA in the hyperthyroid in Thra1PV/+ mice was not significantly increased as compared with that in hypothyroid mice (Fig 7B-a, bar 6 versus bar 4), indicating that TRα1PV expressed in the bone marrow of Thra1PV/+ mice has lost transcription capacity due to its inability to bind T3.
Consistent with the regulation of the Gata1 gene by T3, similar T3-regulatory patterns in the expression of the Klf1 gene was found in the WT mice (Fig 7B-b). The Klf1 mRNA level was lower in the hypothyroid mice than in euthyroid mice (bar 3 versus bar 1), but was higher in hyperthyroid mice than in hypothyroid mice (Fig 7B-b, bar 5 versus bar 3). The Klf1 mRNA level was lower in euthyroid Thra1PV/+ mice than in euthyroid WT mice (bar 2 versus 1). No apparent increase in the expression of the Klf1 mRNA level was found in hyperthyroid Thra1PV/+ mice as compared with hypothyroid Thra1PV/+ mice (Fig 7B-b, bar 6 versus bar 4) due to the loss of T3 binding activity and transcriptional activity of TRα1PV. Taken together, these results indicate that the Gata1 gene is directly positively regulated by T3/TRα1. Mutations of TRα1, such as TRα1PV, suppressed the expression of the Gata1 gene to impair erythropoiesis in Thra1PV/+ mice. To further support this notion, we carried out rescue experiments. We exogenously expressed the GATA1 gene in the Lin-bone marrow cells followed by terminal differentiation assay. As shown in Fig 8A, GATA1-tagged with V5 protein was detected by anti-V5 antibodies (lane 2), whereas it was not detected in the bone marrow cells transfected with on the control plasmid (lane 1). FACS analysis shows that exogenously expressed GATA1 did not further increase mature erythrocytes in WT mice, as indicated by Ter119+ with low FSC population (Fig 8B-a versus 8B-b, areas boxed in red). In contrast, exogenously expressed GATA1 led to increase in mature erythrocytes (39% in the Fig 8C-b versus 27% in the Fig 8C-a, areas boxed in red) in Thra1PV/+ mice. Quantitation graph represent a 43% increase in mature erythrocytes from Thra1PV/+ mice (Fig 8C, panel c). These data demonstrated that the defective terminal erythropoiesis from Lin-BM in Thra1PV/+ mice was partially corrected by exogenous expression of GATA1 protein.
We further evaluated whether the exogenously transfected GATA1 as described above could affect the erythroid genes critical for erythropoiesis. Fig 9 shows the expression of Gata1 (panel A), Klf1 (panel B), β-globin (panel C), dematin (panel D), CAII (panel E), band3 (panel F), and eALAS (panel G) were all increased by the exogenous expression of GATA1 (compare lanes 4 with lanes 3 in all panels). These results indicated that the elevated expression of these erythroid genes contributed to the partial rescue of the defective erythropoiesis and further support the essential role of GATA1 in mediating the impaired erythropoiesis in Thra1PV/+ mice.
In 2002, we created the Thra1PV/+ mouse that displayed aspects of the phenotype of hypothyroidism with retarded growth, delayed bone development, and marginally abnormal thyroid function tests [12]. These phenotypic manifestations are distinct from those of the ThrbPV/+ mice that faithfully reproduce resistance to thyroid hormone (RTH) in patients caused by mutations of the THRB gene [41]. The distinct phenotypic manifestations by Thra1PV/+ mice and ThrbPV/+ mice suggested TR isoform-dependent actions of TR mutants in vivo. However, because of the lack of patients with mutations of the THRA gene at the time, it was not clear whether this postulate held true. The discovery of patients with mutations of the THRA gene in 2012 [8], exhibiting similar hypothyroidism as in the Thra1PV/+ mouse, has validated that the Thra1PV/+ mouse is a valuable model to elucidate the molecular basis of hypothyroidism caused by mutated TRα1. Indeed, Thra1PV/+ mice have been used to predict the outcome of prolonged treatment with a supraphysiologic dose of T4 aimed at ameliorating the skeletal abnormalities in patients [16]. The studies provided valuable information showing that patients with different THRA mutations would display responses to T4 treatment that vary depending on the severity of the causative mutation.
In the present studies, we used Thra1PV/+ mice to understand the molecular actions of TRα1 mutants that lead to erythroid disorders in patients. We found that Thra1PV/+ mice exhibited abnormal red blood cell indices, decreased erythroid lineage progenitors in the bone marrow, and reduction in the terminal differentiation of progenitors in the erythroid lineage. Moreover, for the first time, we have identified a key erythropoietic gene, the Gata1 gene, as a direct TR/T3 target gene by the discovery of thyroid hormone response elements in the promoter region of that gene. Thus, using the Thra1PV/+ mouse, we have provided direct evidence to indicate that mutations of TRα1 could lead to erythroid disorders and that the disorders are mediated, at least in part, by the suppressed expression of a key erythropoietic regulator, the Gata1 gene, by TRα1 mutants. The present studies have shed new light on the molecular basis of the erythroid disorders found in patients with mutations of the THRA gene.
The critical role of TRα1 was also demonstrated in another mouse model, the Thra-/- mouse [42]. Deficiency of TRα1 led to defective fetal and adult erythropoiesis in that erythroid progenitor numbers were decreased in fetal livers; moreover, terminal maturation of erythrocytes was impaired [42]. While the erythroid defects observed in Thra-/- mice were similar to those of Thra1PV/+ mice, the underlying mechanisms would be different. The defective erythropoiesis in Thra-/- mice would be mediated by the lack of TRα1 in regulating the transcription of erythroid-related TRα1 target genes. In contrast, in Thra1PV/+ mice, the impaired erythropoiesis was due to the dominant negative action of TRα1 mutants on erythroid-related TRα1 target genes, such as the suppression of the T3-positively regulated Gata1 gene demonstrated in the present studies. At present, in Thra-/- mice, the erythroid-related TRα1 target genes affected by TRα1 deficiency have not been elucidated. It is difficult to compare the extent and the scope of erythroid abnormalities from the TRα1 deficiency or mutated TRα1 until a comprehensive analysis of global gene expression profiles becomes available. Still, the similar erythroid defects observed in Thra-/- mice and Thra1PV/+ mice clearly highlight the important role of TRα1 in erythropoiesis.
The loss of normal functions of TRα1 due to mutations as in Thra1PV/+ mice or due to TRα1 deficiency as in Thra-/- mice could lead to erythroid disorders as shown in the present studies and in Kendrick et al. [42], respectively. However, that wild-type TRα1 plays a critical role in erythropoiesis was also demonstrated by the findings that thyroid hormones promoted the clonogenic forming ability of BFU-E, CFU-E, and CFU-Mk in hyperthyroid mice shown in the present studies (Fig 3B). These findings are consistent with the observations from clinical studies, indicating that thyroid hormones play a critical role in erythropoiesis. Qualitative and quantitative studies of erythropoiesis in patients with hyperthyroidism exhibit mild erythrocytosis with erythyroid hyperplasia and increased erythropoietic activity in the bone marrow [43]. In addition, erythrocyte counts, serum erythropoietin, and hypoxia-inducible factor 1α levels patients with untreated Graves' hyperthyroidism were significantly higher than those in the age- and sex-matched healthy controls. Methimazole or subsequent radioiodine therapy of patients with hyperthyroidism reduced erythrocytosis and thyroid function returned to normal, suggesting that thyroid hormone promotes erythrocytosis [44]. Moreover, recent population-based studies on euthyroid subjects revealed a significant positive association between thyroid hormones and erythrocyte indices, such as erythrocyte counts, and hemoglobin levels [45]; [46]. These association studies support the notion that thyroid hormones stimulate erythropoiesis. Still, the molecular mechanisms by which thyroid hormones promote erythropoiesis in hyperthyroidism are not clear. The models of Thra1PV/+ and Thra-/- mice would provide valuable tools for such studies.
Early studies have shown that v-erbA is one of the two oncogenes of the avian erythroblastosis virus (AEV) [47], an acute chicken retrovirus that induces lethal erythroleukemia and sarcoma in vivo. V-erbA is a mutated TRα1, which acts in neoplasia by blocking erythroid differentiation and by altering the growth properties of fibroblasts [47]. Similar to TRα1PV, v-erbA functions as a transcription repressor by dominant negative interference with the transcription activity of its normal cellular homolog, c-erbA (TRα1). While the v-erbB locus alone is sufficient to induce erythroleukemia and sarcoma independent of the v-erbA gene, the v-erbA by itself is not capable of independently causing transformation in either erythroid cells or fibroblasts [47, 48], The expression of the v-erbA gene in erythroid cells blocks terminal differentiation and keeps the cells in very immature and highly proliferative stages [47]. Similar to v-erbA, TRα1PV is a potent dominant negative mutant of TRα1. These observations on the functional characteristics of v-erbA raised an important question about whether TRα1PV could act as an oncogene in Thra1PV/+ mice. Up to now, we have not observed any transformed phenotypes in erythroid cells of Thra1PV/+ mice. It is likely that another oncogene such as v-erbB (a mutated version of epidermal growth factor receptor; EGFR) would be needed, as in AEV-induced erythroleukemia and sarcoma, to collaborate with TRα1PV to bring out the transformed phenotypes in erythroid cells of Thra1PV/+ mice. This possibility is testable and awaits future studies.
The present studies have shown that Thra1PV/+ mice exhibited erythroid disorders with abnormal red blood cell indices, decreased total bone marrow cells, and reduced clonogenic potential of erythroid progenitors. The defective erythropoiesis was mediated by TRα1PV-mediated suppression of a key erythropoietic gene (the Gata1 gene), resulting in concurrent repression of other genes involved in the maturation of erythrocytes. Two TREs were identified on the Gata1 gene that responded to T3 differently in the reporter assay. TRα1/T3 via interacting with TRE1 mediated suppression of the transcription whereas TRα1/T3 via interacting with TRE2 activated transcription. It is of interest to point out that located upstream of TRE2 are two E-boxes and one GATA box (GATA1 gene hematopoietic enhancers) critical to the GATA1 gene transcription (see Fig 7A-I). In view of the findings that in vivo, T3 activated the overall transcription of the Gata1 gene as shown in the hyperthyroid WT mice (Fig 7B-a, makes it tempting to speculate that TRE1, though shown to be negatively regulated by T3 in the reporter assay (Fig 7A-II-b), could conceivably be affected by the GATA1 gene hematopoietic enhancers via a long-range looping mechanism, functionally acting as a positive TRE. However, interestingly, in the chicken GATA1 promoter, a negative TRE was identified [49]. TRα1 binds to this TRE as heterodimers with the chicken ovalbumin upstream promoter transcription factor [49], suggesting that the regulatory role of TRα1 in the GATA1 gene transcription is conserved between chicken and mouse.
The identification of the Gata1 gene as a key regulator would suggest that the Gata1 gene or the genes it regulates could potentially be targets for treatment. Moreover, novel TR isoform-specific thyroid hormone analogs are being developed. The Thra1PV/+ mouse would be a valuable model to test the effectiveness of these potential targets to correct the erythroid disorders.
All animal studies were performed according to the approved protocols of the National Cancer Institute Animal Care and Use Committee. The animal study protocol is NCI LMB-036. Mice harboring the mutated Thra1PV gene (Thra1PV mice) were prepared and genotyped by PCR as described earlier [12]. Wild-type (Thra1+/+) and Thra1PV/+ female siblings were used in this study. To induce hypothyroidism, mice were fed a low-iodine diet supplemented with 0.15% propylthiouracil (LoI/PTU) (Cat# TD 95125, Harlan Teklad, Madison, WI) for 10 days. To induce hyperthyroidism, T3 (5 μg; Cat# T2752, Sigma-Aldrich, St. Louis, MO) was injected intraperitoneally to each mouse for 6 days while they were being fed with LoI/PTU diet. The same volume of vehicle (phosphate-buffered saline) was injected in the control group.
Erythroleukemia K562 cell line was maintained in RPMI1650 (Thermo Fisher Scientific, Waltham, MA) with 10% fetal bovine serum (FBS; GE Healthcare life science, Marlborough, MA) with 50 units/ml penicillin G and 50 μg/ml streptomycin (Thermo Fisher Scientific, Waltham, MA). Bone marrow cells were isolated from femurs and tibiae of wild-type and Thra1PV mice (age: 3–5 months). Single cell suspensions were prepared by passing bone marrow through a 70 μM cell strainer.
For analysis of complete blood counts, peripheral blood was collected in a heparinized microtube and analyzed by hematology analyzer (HEMAVET HV950FS, Drew Scientific, Miami Lakes, FL).
The level of TSH in serum was measured as described [50]. Total T4 (TT4) and T3 (TT3) levels were determined by using Gamma Coat T4 and T3 assay radioimmunoassay (RIA) kits according to the manufacturer’s instruction (Cat# 06B256447 and 06B254029, MP Biomedical, LLC, Solon, OH).
To detect burst-forming units-erythroid (BFU-E) colonies, 5 X 104 bone marrow cells were seeded in duplicates in semisolid medium (Methocult M3434; STEMCELL Technologies, Vancouver, BC). To detect colony forming units-erythroid (CFU-E) colonies, 8 X 104 bone marrow cells were seeded in duplicates in semisolid medium (Methocult M3334; STEMCELL Technologies, Vancouver, BC). To detect colony- forming units-megakaryocytes (CFU-Mk) colonies, 1 X 105 bone marrow cells were seeded in duplicates in semisolid medium (Methocult-c, 04974; STEMCELL Technologies, Vancouver, BC) supplemented with 10 ng/ml Interleukin (IL)-3, 20 ng/ml Interleukin (IL)-6, 50 ng/ml thrombopoietin (TPO) (STEMCELL Technologies, Vancouver, BC). To analyze the colonies of multi-potential progenitor cells (CFU-GEMM) and granulocyte/macrophage progenitor (CFU-GM), 4 x 104 bone marrow cells were mixed with semisolid medium (Methocult GF M3434; STEMCELL Technologies, Vancouver, BC) by vortexing. Bone marrow cells (4 x 104 cells) from WT mice (n = 4) and Thra1PV/+ mice (n = 4) were seeded in 6 wells plate (DENVILLE, SCIENTIFIC INC., quadruplicates) which was cultured in 5% CO2 humidified incubator at 37°C. The numbers of colonies were counted under inverted microscope (Primo Vert, Ziess) by morphologic criteria 8 days after plating.
Total RNA was isolated from bone marrow cells using Trizol (Thermo Fisher Scientific, Waltham, MA). RT—qPCR was performed with one step SYBR Green RT-qPCR Master Mix (Qiagen, Valencia, CA). The mRNA level of each gene was normalized to the GAPDH (glyceraldehyde-3-phosphate dehydrogenase) mRNA level. The primer sequences are listed in S1 Table.
ChIP assay with bone marrow cells was performed as described previously [51]. Quantitative PCR was performed to detect the upstream fragment in Gata1 genes (primer sequences are listed in S1 Table). The fold of changes in binding was relative to the control of IgG level as 1.
Oligonucleotide probe containing mouse Gata1 TREs or F2 TRE (positive control) was labeled with [α-32p] dCTP by Klenow fill-in reaction. Assays were performed as described previously [52].
The Gata1 TRE2 and Gata1 TRE1 luciferase constructs were generated by cloning upstream Gata1 promoter fragments into the pGL4.23 luciferase plasmid. The Gata1- TRE2 luciferase construct was made by insertion of a 1.553 kb XhoI-HindIII fragment representing the sequences between -3.741 kb and -2.189 kb. The Gata1-TRE1 luciferase construct was made by insertion of a 140 bp XhoI-HindIII fragment representing the sequences between +1.385 kb and +1.524 kb. Insert sequence validated by DNA sequencing. K562 cells were transfected with the TRE-luc reporters with Genepulse X cell electrophorators (Biorad). Bone marrow cells (2X106 cells) were transfected with the GATA1-pLenti6/V5 plasmid (3 μg) provided by Dr. GP Rodgers (NHLBI; [53]) using 4D nucleofector (Lonza) in accordance with the manufacturer’s instructions. The Gata1 reporter plasmids cloned in pGL3 basic (5 μg), and the expressing plasmid for TRα1 (pcDNA3.1-TRα1; 10 μg) with or without the expression plasmid for TRα1PV (pcDNA3.1-TRα1PV; 80 μg) were transfected into K562 cells according to Guigon et al [54]. Luciferase activity was measured using Victor 3 (PerkinElmer Life and Analytical Sciences, Waltham, MA). Luciferase values were standardized to the ratio of β-galactosidase activity and protein concentration. The fold of changes in activity was based on using the values of negative control (no plasmid transfected cell without T3) as 1.
For lineage depleted bone marrow cell preparation, linage marker positive cells were depleted using the biotin based selection kit (cat# 19856, STEMCELL Technologies, Vancouver, BC) according to the manufacturer's instructions. Lin- BM cells were seeded in fibronectin-coated wells (Corning Inc, Corning, NY). To induce erythropoiesis, Lin- BM cells were cultured as described [24].
For May-GrÜnwald Giemsa stain, cytocentrifuged cells were stained with May-GrÜnwald solution (Cat# MG500, Sigma-Aldrich, St. Louis, MO) for 5 minutes and in Giemsa (Cat# GS500, Sigma-Aldrich, St. Louis, MO) for 20 minutes. For whole bone marrow sections, femurs were fixed in 10% (vol/vol) neutral buffered formalin solution (NBF, approximately 4% formaldehyde) (Sigma-Aldrich, St. Louis, MO). The embedded sections were stained with hematoxylin and eosin (HistoServ, Germantown, MD).
All antibodies used in flow cytometry were from eBiosciences (Thermo Fisher Scientific, Waltham, MA). The sources of antibodies and fluorophore-labeled antibodies used in FACS analyses are listed in S2 Table. The flow cytometry analyses were performed on a BD LSR II flow cytometer (BD bioscience, San Jose, CA) and analyzed with FloJo, LLC (Tree Star Inc, Ashland, OR).
The western blot analysis of bone marrow lysates was carried as described previously [50]. To determine the GATA1 tagged with V5 (The V5 tag is derived from a small epitope found on the P and V proteins of the paramyxovirus of simian virus 5 (SV5). after transfection with GATA1-V5 expression plasmid, anti-V5 antibodies (1:2000 dilution; Thermo Fisher Scientific) was used to detect the expressed GATA1-V5. GAPDH (1:4000 dilution; Cell Signaling Technology (Danvers, MA) was used as a loading control.
For the detection of GATA1 proteins in the bone marrow of WT and Thra1PV/+ mice, bone marrow lysates (600 μg each) were first immunoprecipitated with rat anti-GATA1 antibody (4 μg; Santa Crus Biotecholology, Cat.# Sc-265) or mouse IgG (4 μg; negative controls) followed by pulling down the enriched GATA1-anti-GATA1 antibody-complex with protein G-agarose beads. GATA1 proteins were subsequently detected by western blot analysis as described above using rabbit anti-GATA1 antibody (1:1000 dilution; abcam, Cat.# ab28839).
All statistical analyses and the graphs were performed and generated using GraphPad Prism version 6.0 (GraphPad Software, La Jolla, CA). P < 0.05 is considered statistically significant. All data are expressed as mean ± SEM.
|
10.1371/journal.pgen.1007092 | Flip/flop mating-type switching in the methylotrophic yeast Ogataea polymorpha is regulated by an Efg1-Rme1-Ste12 pathway | In haploid cells of Ogataea (Hansenula) polymorpha an environmental signal, nitrogen starvation, induces a reversible change in the structure of a chromosome. This process, mating-type switching, inverts a 19-kb DNA region to place either MATa or MATα genes under centromeric repression of transcription, depending on the orientation of the region. Here, we investigated the genetic pathway that controls switching. We characterized the transcriptomes of haploid and diploid O. polymorpha by RNAseq in rich and nitrogen-deficient media, and found that there are no constitutively a-specific or α-specific genes other than the MAT genes themselves. We mapped a switching defect in a sibling species (O. parapolymorpha strain DL-1) by interspecies bulk segregant analysis to a frameshift in the transcription factor EFG1, which in Candida albicans regulates filamentous growth and white-opaque switching. Gene knockout, overexpression and ChIPseq experiments show that EFG1 regulates RME1, which in turn regulates STE12, to achieve mating-type switching. All three genes are necessary both for switching and for mating. Overexpression of RME1 or STE12 is sufficient to induce switching without a nitrogen depletion signal. The homologous recombination genes RAD51 and RAD17 are also necessary for switching. The pathway controlling switching in O. polymorpha shares no components with the regulation of HO in S. cerevisiae, which does not involve any environmental signal, but it shares some components with mating-type switching in Kluyveromyces lactis and with white-opaque phenotypic switching in C. albicans.
| The molecular mechanisms of self-fertility (homothallism) vary enormously among fungal species. We previously found that in the yeast Ogataea polymorpha, homothallism is achieved by a novel mating-type switching mechanism that exchanges the locations of MATa and MATα genes between expression and repression contexts. Switching in this species is induced by nitrogen depletion, unlike the analogous process in Saccharomyces cerevisiae. Here, we show that the upstream parts of the genetic pathway controlling the environmental induction of switching in O. polymorpha are the same as the environmental pathway that induces competence for mating in this species.
| In yeast species (unicellular fungi) that can reproduce sexually, the ability of a cell to mate with other cells is governed by which mating-type genes it expresses [1, 2]. In ascomycete yeasts, these genes are located at a single genomic site called the mating-type (MAT) locus. Mating generally occurs between two haploid cells with opposite genotypes (MATa and MATα) at this locus, to form a diploid zygote (MATa/α). In some ascomycete yeasts such as Saccharomyces cerevisiae, haploid cells are able to change their MAT genotypes by a process called mating-type switching [3, 4]. During this process, DNA at the MAT locus is physically replaced, exchanging a MATa allele for a MATα allele or vice versa. Mating-type switching is a form of secondary homothallism [5] because it enables a yeast strain to mate with any other strain of the same species, regardless of their initial mating types, by means of fusion between a-cells and α-cells [6, 7].
The molecular mechanism and regulation of mating-type switching in S. cerevisiae has been elucidated by extensive studies over the past several decades and is well understood [3, 8]. It involves an endonuclease (HO) that cuts the outgoing MAT locus, and two ‘silent cassettes’ (HMR and HML) that contain unexpressed copies of the MATa and MATα DNA sequences. One of the cassettes is chosen to be used as the template for synthesis of new DNA to repair the MAT locus, replacing MAT with a sequence of the opposite genotype. In contrast, until recently little was known about how other ascomycete yeasts switch mating types, other than in Schizosaccharomyces pombe [9] which is a member of a different subphylum. In 2014, Maekawa and Kaneko [10], and our group [11], discovered that haploid cells of Ogataea polymorpha switch mating types by a novel ‘flip/flop’ mechanism that is quite different from the mechanism used by S. cerevisiae. O. polymorpha (formerly called Hansenula polymorpha) is a methylotrophic yeast in the same subphylum as S. cerevisiae (Saccharomycotina, the budding yeasts) but quite distantly related to it (Fig 1A).
O. polymorpha chromosome 3 contains both a MATa locus and a MATα locus, approximately 19 kb apart (Fig 1B). The two MAT loci are beside two copies of an identical 2-kb DNA sequence that form an inverted repeat (IR) on the chromosome. During mating type switching, the two copies of the IR recombine, inverting the orientation of the 19-kb region relative to the rest of the chromosome. The centromere of chromosome 3 is located just to the left of the left copy of the IR (Fig 1B). The MAT locus proximal to the centromere is not transcribed, probably due to silencing by centromeric heterochromatin, whereas the distal MAT locus is transcribed. By inverting the 19-kb region, mating type switching swaps the locations of the MATa and MATα genes, repressing the MAT genes that were previously expressed, and expressing the ones that were previously repressed. Similar flip/flop mating type switching mechanisms are now known in three other Saccharomycotina species (Komagataella phaffii, Pachysolen tannophilus, and Ascoidea rubescens) [4, 11, 12].
Mating type switching in O. polymorpha is induced by an environmental signal, nitrogen depletion [10, 11]. In a culture transferred into media that contains no nitrogen, up to approximately 25% the cells in the culture switch their mating type (Fig 1C). This situation, in which an environmental signal reproducibly induces a DNA rearrangement at a specific chromosomal locus, is unusual in biology and we were motivated to investigate its mechanism. Our aim in the current study was to identify the pathway in O. polymorpha that detects the environmental signal and executes rearrangement of chromosome 3 in response. A priori, we know that the pathway in O. polymorpha must be quite different from the pathway that regulates mating-type switching in S. cerevisiae [3, 13], because switching in S. cerevisiae is not regulated by the environment and occurs even in rich media, and because O. polymorpha has no ortholog of the S. cerevisiae HO endonuclease gene. Therefore, both the upstream (nitrogen-sensing) and downstream (DNA inversion) parts of the pathway in O. polymorpha must be different from S. cerevisiae. Furthermore, since the DNA rearrangements that occur during switching in S. cerevisiae, O. polymorpha and Kluyveromyces lactis are all substantially different but are descendants of a common ancestral switching mechanism [4, 14, 15], we were interested to determine how the pathways that regulate these rearrangements have evolved.
To identify components of the switching pathway in O. polymorpha, we used several strategies including transcriptomic analysis, candidate gene approaches, and mapping the defective gene in a naturally-occurring mutant that is unable to switch mating types. We identified five genes that are required for switching. Although we were unable to deduce all the steps that lead from nitrogen depletion to mating type switching, we infer that O. polymorpha senses nitrogen depletion using the Protein Kinase A (PKA) pathway, which then transmits a signal via Ste12 to induce mating and/or mating type switching, and that recombination between the IRs is mediated by the homologous recombination pathway for DNA repair. We compare the roles of genes in the O. polymorpha pathway to the roles of their orthologs in other species.
Our initial approach to search for genes involved in mating-type switching in O. polymorpha was to look for differences between the transcriptomes of cells that are switching and cells that are not switching. Switching in several methylotrophic yeast species is induced by nitrogen depletion [10–12, 16], and in O. polymorpha we used liquid NaKG media (0.5% NaOAc, 1% KCl, 1% glucose), which completely lacks amino acids or any other source of nitrogen, to induce switching. O. polymorpha grows poorly in NaKG, so to induce switching we first grew ‘pre-induction’ cultures in rich media (YPD) and then transferred the cells, after washing, into NaKG. In the haploid strain NCYC495, recombination between the IRs in the MAT region was induced within 24 hours after transfer into NaKG, whereas no recombination occurred in the YPD pre-induction cultures (Fig 1C).
To examine the transcriptional response induced by nitrogen depletion, we used mRNAseq to compare the transcriptomes of O. polymorpha cells 2 h after transfer from a YPD pre-induction culture into NaKG, to parallel cultures transferred into fresh YPD. Furthermore, because we expect that switching occurs only in haploid cells, we conducted this experiment in parallel on haploid (MATa and MATα isogenic strains) and diploid (MATa/α) cells.
Growth of all three cell types in NaKG resulted in a robust transcriptional response to nitrogen depletion, with a large number of genes significantly up- or down-regulated relative to YPD (S1 Fig; S1 Table). Regardless of cell type, homologs of S. cerevisiae genes for nitrogen starvation responses were induced, such as transporters of amino acids (DIP5, GAP1), urea (DUR3), and allantoate (SEO1), and amidases for the release of amide groups from urea (DUR1,2), pyrimidines (PYD3), or other substrates (AMD2). Ribosomal protein genes were strongly repressed, as expected because of the reduced growth rate in NaKG (S1 Table). However, orthologs of S. cerevisiae genes with mating or sporulation functions were not induced by these nitrogen depletion conditions alone, even though mating (of haploids) and sporulation (of diploids) can be induced by plating cells onto similar nitrogen-depleted solid media [17]. Among the genes strongly upregulated in NaKG were two transcription factors, RME1 and CZF1-like3 (one of three O. polymorpha co-orthologs of C. albicans CZF1, which is a singleton zinc finger gene with no S. cerevisiae ortholog [18]). Both of these genes were uniformly induced in all three cell types (MATa, MATα and MATa/α), with CZF1-like3 upregulated 69- to 93-fold, and RME1 upregulated 26- to 79-fold, upon transfer into NaKG (S1 Table).
In S. cerevisiae, defined sets of a- and α-specific genes that allow haploid cells to identify and respond to the presence of a mating partner are well established [19]. These genes are constitutively expressed in S. cerevisiae cells of the appropriate mating type. Surprisingly, comparison of gene expression between haploid O. polymorpha a-cells and α-cells in either NaKG or YPD media revealed that there are essentially no constitutive a- or α-specific genes in this species, apart from the MAT genes themselves (S1 Fig; S2 Fig). All haploid cells of O. polymorpha contain four MAT genes (MAT α1, α2, a1, and a2), and the orientation of the 19-kb region specifies whether the α1 and α2 genes, or the a1 and a2 genes, are placed at the expression site (Fig 1B). In NaKG, transcription of α1 and α2 was respectively 53-fold and 39-fold higher in α-cells than in a-cells; a2 was 31-fold lower, and a1 was just 2-fold lower. In YPD, α1 and α2 were 4-fold and 9-fold higher, a1 was 6-fold lower, and a1 showed no difference. KAR4, which in S. cerevisiae is a general pheromone-induced gene [20] required for fusion of the haploid nuclei after mating, showed moderately higher expression in a-cells than in α-cells (2 to 3-fold; S2 Fig). No other genes showed more than a 2-fold difference in transcription between a- and α-cells, in either of the two media (S2 Fig; S2 Table). This result contrasts sharply with S. cerevisiae, where for example several a-specific genes such as MFA2, STE2 and BAR1 have more than 10-fold higher expression in MATa than MATα cells in YPD [19]. It is also consistent with previous observations that expression of the pheromone receptors STE2 and STE3 in haploid O. polymorpha is independent of cell type [10].
These experiments also enabled us to identify gene expression differences between haploid and diploid cells. O. polymorpha is haplontic, and its diploid state is normally transient because meiosis is induced by the same conditions (nitrogen depletion) that induce mating. However, diploids can be maintained stably on YPD. We calculated the haploid-to-diploid expression ratio for each gene as the ratio between its transcription in a-cells and a/α-cells. The values of this ratio in different genes were quite consistent between YPD and NaKG media (Pearson’s R = 0.68; S3 Fig). Among the genes showing the strongest bias in YPD towards haploid-specific expression were several transcription factors including CZF1-like1, CZF1-like2, CRZ1, GAT1, and MGA1 (S3 Table). Of these, only CZF1-like2 was also haploid-specific in NaKG. Transcription factor DAL81 appeared diploid-specific in both media (S3 Table).
Because mating-type switching occurs in haploid cells grown in NaKG, but not in haploids grown in YPD, and presumably not in diploids, we anticipated that genes with roles in switching might be identifiable as transcripts that are both haploid-specific and NaKG-specific. However, analysis of the genes fitting this transcription profile did not reveal any strong candidates for the downstream steps in the switching process, such as DNA recombination or endonuclease genes. Instead, most of the genes with this profile had metabolic functions (S4 Fig). The most haploid-specific and diploid-specific genes in the two media are listed in S3 Table.
Next, in an alternative approach to find a component of the switching pathway, we made use of a naturally occurring mutant. When assaying the MAT genotypes of Ogataea strains, we discovered that strain DL-1 is unable to switch mating-types, even after 45 h growth in NaKG, in contrast to strains NCYC495 and CBS4732 (Fig 2A). Strain DL-1 has previously been described as ‘semi-sterile,’ meaning that it is very inefficient at forming diploids under nutrient-limited conditions [21]. The semi-sterility phenotype of DL-1 is therefore likely due to a loss of the signal that is induced by nitrogen depletion, upstream of the steps that normally lead to either mating or switching in response to the signal.
Strains DL-1, NCYC495 and CBS4732 were all historically classified as Hansenula polymorpha but it has recently been recognized, based on sequence divergence, that DL-1 is a different species from the other two. DL-1 is now classified as Ogataea parapolymorpha, whereas NCYC495 and CBS4732 are O. polymorpha [22, 23]. However, the genome sequence of O. parapolymorpha DL-1 [24] is completely collinear with the genome sequence of O. polymorpha NCYC495 [12]. Both species have 7 chromosomes, and there are no translocations or other chromosomal rearrangements between them, even though the genomes are approximately 10% different in nucleotide sequence [11]. The fact that the genomes are collinear suggested to us that the mating-type switching defect in DL-1 could be mapped by using an interspecies genetic cross between it and the O. polymorpha laboratory strain NCYC495.
We used bulk segregant analysis [25] to map the locus causing the switching defect. We first isolated a rare diploid from a cross between DL-1 (leu2 ura3 genotype) and an NCYC495 derivative (ade11 met6 genotype), selecting for prototrophy. We then sporulated the diploid and isolated haploid segregants grown from random spores (Fig 2B). Segregants were screened individually for their ability to switch mating-types after 24 h in NaKG by the same PCR assay used above. We made four pools of segregants: MATa switchers, MATa non-switchers, MATα switchers, and MATα non-switchers, each pool containing between 30 and 52 haploid clones (S5A Fig), and sequenced each pool. Cultures of each clone in a pool were grown individually and then combined into pools in equal cell numbers for DNA extraction and genome sequencing. The sequence reads from the switching and non-switching pools were then mapped to the parental DL-1 and NCYC495 genome sequences (S5B Fig); only reads that were unambiguously derived from one identifiable parent were mapped. We developed an asymmetry metric (see Methods) to detect regions of the genome where biased inheritance of parental alleles correlated with the switching/non-switching phenotype in the expected direction (Fig 2C).
Two peaks of asymmetrical inheritance were detected (Fig 2C). The strongest signal was located on chromosome 6 and was centered near the gene OPOL_95241, which we refer to as O. polymorpha EFG1. It is orthologous to the C. albicans transcription factor EFG1 [26] and to the S. cerevisiae gene pair PHD1 and SOK2 derived from the Whole-Genome Duplication [27–29]. Comparison of the EFG1 sequences from the parental NCYC495 and DL-1 genomes revealed a single-base insertion at nucleotide 512 in the DL-1 gene that causes a frameshift (Fig 2D). The predicted DL-1 Efg1 protein product is truncated to 203 residues, compared to 437 residues in NCYC495. The DL-1 Efg1 protein lacks a DNA-binding domain (APSES domain [26, 30, 31]) that is conserved among Efg1 orthologs in multiple species including C. albicans and S. cerevisiae (S6A Fig). A second region of asymmetrical inheritance occurred on chromosome 7 near coordinate 330 kb (Fig 2C). Comparison of the NCYC495 and DL-1 genomes in this region did not reveal any candidate disabling mutations in genes, or differences in gene content. Considering that this analysis used a cross between two different species, it is possible that the chromosome 7 region contains a gene that interacts with a gene near EFG1, for which an interspecies combination of alleles is inviable, but we did not investigate this region further.
To confirm that EFG1 plays a role in mating-type switching in O. polymorpha, we deleted it from both MATa and MATα strains. Gene deletions were made in ku80Δ derivatives from the NCYC495 genetic background [32]. PCR assays showed that, after 24 h in NaKG, almost no switched MAT locus products were formed in the efg1Δ strains, whereas extensive switching occurred in the wildtype control strains (Fig 3A). Furthermore, the efg1Δ strains were defective in mating, similar to the semi-sterility phenotype of DL-1. Crosses of efg1Δ x efg1Δ strains yielded no progeny, and crosses of efg1Δ x EFG1 strains yielded only a small number of progeny compared to wildtype crosses (Fig 3B). This result indicates that both parents in a cross require EFG1 activity in order to mate.
Because EFG1 is required for both mating and mating-type switching in O. polymorpha (Fig 3A and 3B), we reasoned that it must act in an upstream part of the nutrient-sensing pathway that is shared by these two processes. Such a function is consistent with the known role of EFG1 in C. albicans as the major transcription factor of the PKA pathway [33–36], even though its S. cerevisiae orthologs PHD1 and SOK2 have no role in mating or switching. We therefore searched for O. polymorpha genes whose expression depends on Efg1.
To find genes downstream of EFG1, we compared the transcriptional profiles of wildtype and efg1Δ haploid strains, by mRNAseq in nitrogen-poor and nitrogen-rich conditions. We used two different types of paired media for this experiment. One was a comparison of transcriptomes in NaKG versus NaKG plus 40 mM ammonium sulfate, which we have previously shown abolishes switching [11]. The other, chosen to try to reduce the strong general nutrient depletion signal we observed with NaKG (S1 Fig), was a comparison of transcriptomes in synthetic defined media (SD, which includes 40 mM ammonium sulfate) versus SD lacking this nitrogen source. These mRNAseq experiments identified many nutrient transporters and enzymes that have EFG1-dependent expression in nitrogen-poor media (S4 Table), consistent with EFG1’s expected role in the PKA pathway, but no obvious candidates for direct actors in the flip/flop inversion mechanism, such as DNA recombinases or endonucleases. However, these experiments also showed that one of the genes with the largest EFG1-dependent differences in expression between nitrogen-rich and nitrogen-poor conditions was another transcription factor, RME1.
O. polymorpha RME1 is a gene that is substantially more highly transcribed in nitrogen-poor than in nitrogen-rich conditions, being among the top 2% of genes upregulated in NaKG (Fig 3C; S1 Table). In nitrogen-rich conditions, one of the strongest effects of deleting EFG1 was to increase the expression of RME1 from its low baseline, by factors of 4.4-fold in NaKG + ammonium sulfate, and 6.6-fold in SD, relative to wildtype cells (Fig 3C and 3D). In contrast, in nitrogen-poor conditions RME1 expression was high and unchanged between efg1Δ and wildtype cells (Fig 3D; S4 Table). In the efg1Δ strain, RME1 was expressed in both nitrogen-rich and poor media. Thus, transcription of RME1 in nitrogen-rich conditions is normally repressed by an EFG1-dependent mechanism, which could either be direct or involve intermediate proteins. EFG1 itself showed no difference in expression between NaKG and YPD (Fig 3C). In Kluyveromyces lactis, RME1 (also called MTS1) is required for both mating-type switching and mating [14, 37], so regulation of RME1 by EFG1 in O. polymorpha therefore suggests a mechanism connecting the nitrogen limitation response to switching and mating. In contrast, the main function of RME1 in S. cerevisiae is as a repressor of meiosis via repression of IME1, which has no ortholog in O. polymorpha [38, 39].
The set of genes showing expression changes in the efg1Δ strain also included some ‘white-opaque circuit’ genes. These genes are O. polymorpha homologs of genes that form a feed-forward circuit in C. albicans governing the phenotypic switch between mating-competent (opaque) and mating-incompetent (white) cell states. The C. albicans circuit includes EFG1, WOR1, WOR2 and CZF1 [40]. O. polymorpha has no ortholog of C. albicans WOR1, but has a paralogous gene (OPOL_7784) that we refer to as MIT1 because of its similarity to S. cerevisiae MIT1 [41]. Among the strongest effects of deleting O. polymorpha EFG1 were decreases of expression of CZF1-like2 and MIT1, which occurred in both nitrogen-poor and nitrogen-rich conditions (Fig 3C; S4 Table; all decreases were by less than 4-fold).
In parallel to the experiments with the efg1Δ deletion strain, we also used an overexpression strain to search for genes regulated by EFG1. We placed EFG1 under the control of the O. polymorpha alcohol oxidase promoter (pAOX). Expression from pAOX is robustly induced when cells are switched from growth in glucose to growth in media containing methanol as the carbon source. We used mRNAseq to compare the transcriptome of the pAOX-EFG1 strain to a control strain containing an empty pAOX construct, after overnight growth in methanol media. Methanol induced 7-fold higher transcription of EFG1 in the pAOX-EFG1 strain than in the control (Fig 3C). Overexpression of EFG1 led to changes in transcription of large numbers of genes (221 genes downregulated, and 26 genes upregulated, by factors of at least 8-fold; S5A Table). Notably, RME1 transcription remained unchanged in the EFG1 overexpression strain (Fig 3C).
One of the strongest effects of overexpressing EFG1 (OPOL_95241) was 53-fold repression of a related gene, OPOL_93012 (Fig 3C). Phylogenetic analysis showed that these two APSES domain proteins are the products of a gene duplication that occurred within the genus Ogataea (S6B Fig). This gene duplication is separate from an older duplication that formed the EFG1 homolog EFH1 in the Candida clade [42]. The O. polymorpha Efg1 and OPOL_93012 proteins have 49% amino acid sequence identity. The sister species O. parapolymorpha has orthologs of both EFG1 (with a frameshift) and OPOL_93012, but other budding yeasts including methylotrophs outside the genus Ogataea have only a single gene. EFG1 and OPOL_93012 are both transcribed in both nitrogen-poor and nitrogen-rich conditions. EFG1 has higher expression in haploids than in diploids (2.8- to 3.7-fold), whereas OPOL_93012 shows little difference between cell types (S3 Table). Expression of OPOL_93012 was unaffected in the efg1Δ strain.
Overexpression of O. polymorpha EFG1 also caused changes of expression of some white-opaque genes. CZF1-like2 was down-regulated 20-fold, and CZF1-like3 was up-regulated 7-fold (Fig 3C; S5A Table). WOR3 and WOR4, which are more recently identified components of the white-opaque circuit in C. albicans [43, 44], were down-regulated (10- and 12-fold respectively). The regulatory relationship between EFG1 and CZF1-like2 appears to be complex, because CZF1-like2 was down-regulated by both deletion and overexpression of EFG1.
Based on the results of the EFG1 deletion and overexpression mRNAseq analyses, we tested whether RME1 and the EFG1 paralog OPOL_93012 are required for mating-type switching and/or mating. We also tested STE12, which plays a central role in the mating response in other yeast species, and which shows induction by nitrogen depletion (Fig 3C; S1 Table). MATa and MATα deletion strains for each gene were constructed and tested for their ability to switch mating types, and to mate.
Deletion of RME1 severely reduced mating-type switching, as measured by PCR assay, in both MATa and MATα cells (Fig 4A), similar to the result from EFG1 deletion (Fig 3A). Deletion of STE12 completely abolished switching. Furthermore, deletion of RME1 or STE12 abolished mating, in crosses where both parents were rme1Δ or ste12Δ (Fig 4B). Crossing rme1Δ x RME1 resulted in a low number of diploid colonies, similar to efg1Δ x EFG1 crosses, whereas ste12Δ x STE12 crosses did not produce any colonies (Fig 4B). In contrast to these three genes, deleting the EFG1 paralog OPOL_93012 had no effect on switching or mating (Fig 4A and 4B).
Because EFG1 is a component of the white-opaque circuit in C. albicans, and because several O. polymorpha homologs of white-opaque genes were found to be differentially regulated in our transcriptome analyses as mentioned above (Fig 3C), we also made deletion strains of four O. polymorpha ‘white-opaque’ genes: CZF1-like2, CZF1-like3, MIT1 (WOR1), and WOR2 (S7 Fig). However, none of these deletions had any effect on either switching or mating (S7 Fig).
Since EFG1, RME1 and STE12 are all necessary for switching, we investigated whether high expression of any of them is also sufficient to induce switching, even in the absence of a nitrogen depletion signal. We constructed methanol-inducible pAOX-RME1 and pAOX-STE12 strains similar to the pAOX-EFG1 strain described above. Switching was induced when strains containing pAOX-RME1 or pAOX-STE12 were transferred from glucose to methanol (Fig 5A), demonstrating that overexpressed RME1 and STE12 are each sufficient to induce switching. The pAOX-EFG1 strain, and a control strain containing the pAOX vector alone, did not switch under the same conditions (Fig 5A). The latter result is consistent with the observation that EFG1 is transcribed in nitrogen-rich as well as nitrogen-poor conditions, and indicates that EFG1 requires additional factors in order to induce switching.
The results we have presented so far show that EFG1, RME1 and STE12 are each necessary for switching, and that EFG1 acts upstream of RME1. To determine where STE12 fits into the pathway, we constructed strains that combined overexpression of one gene with deletion of another. We introduced the pAOX-RME1, pAOX-STE12, and pAOX-EFG1 constructs individually into the deletion strains rme1Δ, ste12Δ, and efg1Δ in all possible combinations (Fig 5B). Methanol induction of EFG1 was again unable to induce switching in any background, whereas STE12 overexpression induced switching in all backgrounds. RME1 overexpression, although sufficient for switching in the rme1Δ and efg1Δ backgrounds, did not induce switching in the ste12Δ strain. This result indicates that RME1 acts upstream of STE12 in the switching pathway.
To test whether RME1 binds to the promoter of STE12, we performed ChIPseq using 3xHA-tagged Rme1, expressed from its native chromosomal locus. In addition to binding to its own promoter and 3’ UTR, Rme1 bound to the promoter of STE12 (Fig 5C). Furthermore, mRNAseq analysis of the pAOX-RME1 strain shows that overexpression of RME1 results in an increase in expression of STE12 (S8 Fig; S5C Table). Together, these data suggest that RME1 directly activates transcription of STE12 by binding to its promoter, which leads to switching.
Since overexpression of either RME1 or STE12 induces switching, we tried to identify components further downstream in the switching pathway by transcriptome analysis of the pAOX-RME1 and pAOX-STE12 overexpression strains, after overnight growth in methanol media. We found that the major consequence of overexpressing these transcription factors was strong induction of genes in the mating response pathway (S8 Fig; S5 Table), consistent with the essential roles of RME1 and STE12 in O. polymorpha mating (Fig 4). The genes induced included orthologs of the S. cerevisiae haploid-specific genes (STE4, GPA1, STE18, STE5, FAR1, FUS3) required for transmission of the pheromone signal. Our pAOX overexpression strains were constructed in a haploid MATa background, and we detected methanol-induced transcription of a-specific genes (BAR1, AXL1, ASG7, RAM1, RAM2, STE6) that are required for production of a-factor and modulation of the α-factor signal. The mating pathway induction by STE12 overexpression was so strong that it enabled us to annotate the a-factor gene (MFa) of O. polymorpha for the first time (S9 Fig). We also observed induction of the α-specific genes MATα1, MATα2 and the α-factor gene MFα, which is likely due to expression in cells that had successfully switched mating-type from MATa to MATα in the cultures (Fig 5A). In addition to the mating pathway genes, overexpression of RME1 (but not STE12) also induced transcription of genes with roles in sporulation such as RIM4, IME2, MUM2, and MEI2 (S5C Table). In contrast to RME1 and STE12, EFG1 overexpression did not significantly induce expression of mating pathway genes (S8 Fig).
Disappointingly, the RME1 and STE12 overexpression mRNAseq analyses did not reveal any clear candidates for genes that act downstream in the switching pathway. They did however show that RME1 and STE12 form a positive feedback loop. Overexpression of RME1 induced STE12 by 14-fold, and overexpression of STE12 induced RME1 by 9-fold (S5B Table, S5C Table). They also showed that overexpression of RME1 induced expression of CZF1-like3 and repressed expression of CZF1-like2 (S5A Table), similar to overexpression of EFG1, so the effect of EFG1 on these white-opaque genes is probably mediated through RME1.
Although the transcriptomic and ChIPseq experiments did not identify obvious candidates for downstream roles in the switching pathway, such as homologs of known DNA recombinases or endonucleases, they did uncover several O. polymorpha genes of unknown function whose patterns of transcription were consistent with the profile we expected switching pathway genes to have. We chose 21 candidate O. polymorpha genes for deletion and testing of switching phenotypes, including (i) genes of unknown function with appropriate transcription profiles, (ii) orthologs of S. cerevisiae genes that interact with STE12, such as TEC1, FUS3 and KSS1, and (iii) orthologs of S. cerevisiae genes with roles in mating-type switching, homologous recombination or DNA repair, such as ASH1, RAD51 and PMS1. Deletion strains of each of the 21 genes in an NCYC495 ku80Δ MATα background were tested for their ability to switch mating-types, of which 19 had no phenotype (S10 Fig).
We found that mating-type switching was almost completely abolished in strains with deletions of the orthologs of two S. cerevisiae genes in the homologous recombination pathway, RAD51 and RAD17 (Fig 6). In S. cerevisiae, Rad51 is a single-stranded DNA binding protein that mediates strand exchange during homologous recombination [45], and is necessary for mating-type switching [46]. The O. polymorpha RAD51 gene has previously been reported to partially complement an S. cerevisiae rad51 mutant, and the protein catalyzes DNA strand exchange in vitro [47]. Rad17 is a component of the checkpoint signaling clamp called 9-1-1 in humans or Ddc1-Mec3-Rad17 in S. cerevisiae [45]. The Mec3 component of the clamp is not necessary for mating-type switching in S. cerevisiae [48], but whether Rad17 is necessary has not been investigated. The requirement for RAD51 and RAD17 in O. polymorpha switching shows that the homologous recombination pathway for repair of DNA breaks is involved in the interaction between the IRs.
Our experimental results suggest a model for how mating-type switching and the mating response to pheromone are both controlled in in O. polymorpha (Fig 7). In the presence of a nitrogen source, an EFG1-dependent mechanism represses transcription of RME1 and the whole pathway is inactive. In the absence of a nitrogen source, RME1 is active and a positive feedback loop between RME1 and STE12 expression develops. If pheromone is detected, STE12 activates the mating response pathway and mating ensues. We postulate that if no pheromone is detected, STE12 instead activates mating-type switching, which could then lead to mating with a cell of the original mating-type. The final steps in switching utilize the homologous recombination pathway, but the intermediate steps connecting STE12 to the RAD genes remain unknown.
As shown in Fig 7, RME1 also has a role in regulating ‘white-opaque’ transcription factors. We do not know if a regulatory loop similar to the C. albicans white-opaque circuit exists in O. polymorpha, but in any case our gene deletion experiments (S7 Fig) show that white-opaque genes other than EFG1 have no role in switching or mating. In C. albicans, EFG1 is the main activator of the mating-incompetent (white) state [40], whereas in O. polymorpha EFG1 is required for mating competence. Our model may be oversimplified because the connection between EFG1 and RME1 seems to be complex and may involve intermediate steps. By analysis of the efg1Δ strain we found that in nitrogen-rich conditions EFG1 causes repression of RME1, whereas in nitrogen-poor conditions EFG1 was essential for switching and mating, suggesting conversely that it causes activation of RME1. Since overexpression of EFG1 had no effect on RME1 transcription, the activity of EFG1 may depend on other factors such as the presence of partner proteins, or post-translational modification of Efg1. The Efg1 proteins of some yeast species are known to be phosphorylated [49–51]. C. albicans Efg1 can act as both a repressor and an activator [26, 52], so it is possible that O. polymorpha Efg1 can both positively and negatively affect RME1 transcription in different conditions.
A fundamental difference between O. polymorpha and S. cerevisiae is that in S. cerevisiae, detection of pheromone is the only signal necessary to trigger a mating response, whereas in O. polymorpha a nitrogen depletion signal is needed as well. Ste12 is the probable point of integration of these two signals in O. polymorpha (Fig 7), with the nitrogen-depletion signal (communicated through Efg1 and Rme1) increasing the level of STE12 transcription, and the pheromone-induced MAP kinase cascade activating Ste12 by releasing the ortholog of the inhibitor proteins Dig1/Dig2 [53, 54]. We suggest that O. polymorpha cells initiate switching if Ste12 protein becomes abundant but no pheromone has been detected.
Comparing the networks that contain EFG1, RME1 and STE12 in different ascomycete species shows that there has been extensive reorganization during evolution (Fig 8). These networks are complex because they integrate information about the cell’s nutrient status (from the PKA pathway), the presence of pheromone (from the MAPK pathway), and the cell’s ploidy (from the a1/α2 repressor), to decide whether the cell responds by mating, switching, sporulating, or filamentous growth [55]. In S. cerevisiae, the nutrient status of the cell is primarily signaled by modulating PKA activity, which occurs via cyclic AMP for glucose sensing, and independently of cAMP for sensing other nutrients such as nitrogen [36, 56]. Much of the transcriptional response to changes in PKA activity in S. cerevisiae is mediated by the stress-response transcription factors Msn2 and Msn4 [56, 57]. However, there is no ortholog of Msn2/4 in O. polymorpha, and in C. albicans the major PKA-regulated transcription factor is Efg1, not Msn2/4 [33–36, 58]. It is likely that in O. polymorpha PKA regulates EFG1 to signal nitrogen depletion, because PKA is known to regulate Efg1 orthologs in Eremothecium (Ashbya) gossypii [59] and S. cerevisiae [28, 57], as well as C. albicans [51, 60]. One of the functions of S. cerevisiae Sok2 is to repress the master inducer of meiosis IME1 in rich conditions [49], but IME1 also has no ortholog in O. polymorpha. Thus the role of O. polymorpha EFG1 may be quite unlike the roles of S. cerevisiae SOK2 and PHD1. The MAPK and PKA pathways may be more interconnected in other ascomycetes than in S. cerevisiae, because C. albicans Efg1 also plays a role in mating and interacts with the Dig1/2 ortholog [54].
The mating response pathway of O. polymorpha is similar to that in K. lactis in the sense that Rme1 conveys the nutrient depletion signal that is needed to activate STE12 for mating [20, 37], but their pathways for induction of switching are different (Fig 8). In K. lactis, Ste12 has no known role in switching, and upon nutrient depletion Rme1 induces switching either by activating transcription of KAT1 or by binding to the α3 locus, depending on the direction of switching [14, 15]. Furthermore, the connection between nutrient signaling and switching in K. lactis has been proposed to occur via Msn2 rather than Efg1 [61]. Nevertheless, the pathway that regulates switching in O. polymorpha has more similarity to that in K. lactis than to that in S. cerevisiae. Switching in S. cerevisiae via HO endonuclease is highly regulated in terms of cell cycle and cell lineage [3, 4, 13], but has no connection to PKA signaling or STE12 (Fig 8).
Is there an endonuclease or site-specific recombinase for mating-type switching in O. polymorpha? At the outset of this project we assumed that the flip/flop mechanism would employ a specific enzyme to initiate recombination between the two IRs, but we have been unable to find such an enzyme. In retrospect, we realize that a site-specific recombinase is unlikely because recombinases generally recognize sites that are much shorter than the 2-kb IRs [62]. It now seems probable that during switching a site-specific DNA break is induced in one copy of the IR, followed by repair by recombination with the other copy, which can be resolved as either a crossover (inversion of the 19-kb region) or a non-crossover (no switching). Site-specific breaks are made in S. cerevisiae by HO, and in K. lactis by Kat1 and α3, but the O. polymorpha genome contains no homologs of any of these proteins. One possible hypothesis for O. polymorpha is that a site-specific break might be formed during attempted replication of a fragile DNA site, similar to switching in Schizosaccharomyces pombe [4, 9], but if this is correct, the site must be fragile only in nitrogen-poor conditions. Alternatively, O. polymorpha might use a recombinase or endonuclease that is activated post-transcriptionally. Further characterization of the switching mechanism in O. polymorpha may require biochemical approaches or genetic screens to identify mutants that switch constitutively.
Strains and plasmids used in this study are listed in S6 Table. Constructs for gene deletions contained 700–1000 base pairs of sequence flanking the target locus and an antibiotic resistance marker. Flanking and marker sequences were amplified using a high-fidelity DNA polymerase (Phusion or Q5, New England Biolabs), purified (PCR Purification Kit, Qiagen), and assembled by fusion PCR. PCR products were introduced into cells by electrotransformation, as described previously [63]. Gene deletions were made in ku80Δ backgrounds to increase efficiency of homologous recombination [32]. Successful integration was tested by antibiotic selection on YPD plates containing 200 μg/mL G418, 200 μg/mL hygromycin B, 100 μg/mL nourseothricin, or 100 μg/mL zeocin, as appropriate. Colony PCR was performed on resistant colonies to test for integration at the correct locus (GoTaq G2 polymerase, Promega). Plasmids for overexpression were constructed using pHIPH4 [32]. EFG1, STE12, and RME1 coding sequences were amplified using a high fidelity polymerase and primers containing restriction enzyme sites (SbfI, XmaI, HindIII, or XbaI). The purified PCR products and plasmid were digested, ligated, and transformed into E. coli. Clones were purified and digested with StuI enzyme overnight at 37°C for electrotransformation into O. polymorpha.
To induce mating-type switching, overnight ‘pre-induction’ cultures grown at 37°C in YPD were centrifuged at 3400 x g for 2 min, washed once in NaKG (0.5% sodium acetate, 1% potassium chloride, 1% glucose), and resuspended in 10 mL NaKG at A600 0.5. NaKG cultures were then incubated on a shaker at 28°C for 24 h. DNA was isolated from the pre-incubation and NaKG-grown cultures by phenol:chloroform extraction. The MAT locus orientation was determined by PCR amplification using GoTaq G2 polymerase (Promega) for 30 cycles with 55°C annealing temperature and 3 min elongation. PCR products were visualized on 1% agarose gel with ethidium bromide staining.
Cells were streaked in parallel lines on YPD agar and crossed on MEMA (2.5% maltose, 0.5% malt extract, 2% agar) by replica plating. MEMA plates were incubated at 28°C for 24 h before replica plating to SD agar. SD plates were incubated at 37°C for 48 h to observe growth of diploids.
Induction of expression from the AOX promoter was achieved by growing overnight cultures at 37°C in mineral media [64] containing 0.5% glucose (MMG). Overnight cultures were diluted in fresh MMG to A600 0.2 and grown to A600 >1.5. Cultures were diluted again in fresh MMG to A600 0.2 and grown to A600 >2.0. Cultures were diluted in mineral media + 0.4% methanol (MMM) to A600 0.2 and grown on shaker overnight at 37°C. RNA samples were isolated from these cultures with two biological replicates by hot acid phenol extraction and DNase I (Invitrogen) treatment.
A diploid prototrophic colony obtained from the O. polymorpha NCYC495 x O. parapolymorpha DL-1 interspecies cross was sporulated by streaking on ME agar (2% malt extract, 2% agar) and incubating at 25°C. Random spores were isolated by ether treatment: sporulating culture was suspended in sterile water before addition of an equal volume of diethyl ether and incubation at 30°C for 45 min. Ether-treated cells were diluted, plated on YPD agar, and incubated at 37°C for 48 h. Haploid clones grown from spores were tested for the ability to switch mating types using the PCR assay described above, with the following modification: DNA extractions were performed by treatment of cells with 700 units lyticase, incubation at 37°C for 30 min, followed by extraction using a Promega Maxwell 16 according to manufacturer’s instructions. Clones were identified as MATa or MATα, and as switchers or non-switchers, by PCR assay. Clones with clear phenotypes were assigned to four pools for sequencing: MATα switchers (30 clones), MATα non-switchers (35 clones), MATa switchers (35 clones) and MATa non-switchers (52 clones) (S5A Fig). Clones for each pool were grown individually and the pools were then made by combining equal A600 units for phenol:chloroform DNA extraction. DNA was also extracted from the parental strains NCYC495 and DL-1. All DNA samples were purified using a Genomic DNA Clean and Concentrator kit (Zymo). Genomic DNA library preparation and Illumina HiSeq 2500 sequencing were performed at the University of Missouri DNA Core Facility.
We first created new reference genome sequences for our O. polymorpha NCYC495 (ade11 met6) and O. parapolymorpha DL-1 (leu2 ura3) parental strains, by mapping the reads from these strains onto the published genome sequences [12, 24] using BWA [65]. We did this because we discovered that our ade11 met6 derivative of NCYC495 (obtained from Dr. Kantcho Lahtchev, Bulgarian Academy of Sciences) contains regions with significant numbers of differences relative to the reference sequence of strain NCYC495 leu1.1 (obtained from Prof. Andriy Sibirny, National Academy of Sciences of Ukraine) that was sequenced by Riley et al. [12]. Our ‘NCYC495’ ade11 met6 stock appears to be the product of a cross between a genuine NCYC495 background and an O. polymorpha strain with a slightly divergent genome, possibly strain CBS4732.
We then mapped the Illumina reads from each of the four pools to these O. polymorpha and O. parapolymorpha reference genome sequences. Only reads that had a single perfect match to one species, but no perfect match to the other, were retained for analysis. We divided the Ogataea genome into 7824 segments, where each segment is either a pair of orthologous genes in NCYC495 and DL-1, or a pair of ‘intergenic’ regions in the interval between two consecutive pairs of orthologs. These ‘intergenic’ regions can include genes that are present in one species but absent in the other. For each segment in each species, we calculated the numbers of reads from each pool, and from the parental strains, that mapped to it. Preliminary analysis showed no significant differences between the two mating types, so we merged the data from MATa and MATα clones. We then calculated four ratios for each genomic segment:
These four ratios are plotted in S5B Fig. We defined the Asymmetry metric (Fig 2C) of a genomic segment as
Asymmetry=max(SWNCYC495-1,0)*max(1-NSNCYC495,0)*max(NSDL-1-1,0)*max(1-SWDL-1,0)
This metric has a value of zero, except in genomic segments where four criteria are met simultaneously: the proportion of NCYC495-derived DNA is higher than expected by chance in the switcher pool but lower than expected in the non-switcher pool, and the proportion of DL-1-derived DNA is higher than expected by chance in the non-switcher pool but lower than expected in the switcher pool.
Strains for nitrogen depletion samples and efg1 deletion samples were grown in YPD at 37°C overnight, diluted to an OD600 0.1 and grown to log phase (OD600 1.0). Cultures were pelleted by centrifugation, washed once in YPD, NaKG, NaKG + 40mM ammonium sulfate, SD, or SD minus ammonium sulfate, before resuspending in the same media and culturing for 2 h at 28–30°C. RNA samples from these cultures were prepared with the MasterPure Yeast RNA Purification Kit (Epicentre, Illumina) or by hot acid phenol extraction and DNase I (Invitrogen) treatment. Nitrogen depletion samples were performed in triplicate, overexpression and efg1 deletion samples were performed in duplicate.
Chromatin immunoprecipitation (ChIP) was performed by formaldehyde crosslinking of log phase cells for 20 min with glycine addition used to stop the reaction. Cells were lysed using glass beads and chromatin was fragmented by sonication with a Bioruptor Standard (Diagenode). EZview Red Anti-HA Affinity Gel (Sigma-Aldrich) was used to immunoprecipitate chromatin fragments, and bound DNA was eluted using HA peptide (Sigma-Aldrich). Crosslink reversal was followed by phenol:chloroform extraction. ChIP samples were performed with three biological replicates that were pooled prior to sequencing.
Stranded mRNAseq and ChIPseq library preparation and sequencing services were performed at the University of Missouri DNA Core Facility. 50–51 bp unpaired Illumina reads (RNAseq and ChIPseq) were mapped to the Ogataea polymorpha (NCYC495 leu1.1 [12]) genome using Bowtie v1.1.2 using the following options: -v = 3, to report end-to-end hits with < = 3 mismatches; -k = 10, to report up to 10 good alignments per read;—best, so hits guaranteed best stratum with ties broken by quality; -M = 1, to report just 1 random hit out of the good alignments for a read; -S, to write hits in SAM format; -p = 10, to use 10 processors. Aligned hits were split into reads that mapped to the forward and reverse strand (SAM FLAG = 0 and 16) before proceeding. Samtools v0.1.12a (r862) was used to create sorted and indexed BAM files of the results. Bedtools v2.19.0 was used to create genome coverage Bedgraph files, which were converted to BigWig files using bedGraphToBigWig v4 for visualization as tracks in Jbrowse v1.11.2.
For RNAseq data htseq-count v0.6.0 was used to calculate for each feature the number of reads mapping to it. We mapped to a feature file based on the original JGI NCYC495 annotation with extensive manual modification. We counted against both the forward and reverse strand mapping SAM files, creating sense and antisense counts for each feature, but only retained sense counts for further analysis. We then calculated Transcripts Per Million (TPMs) [66] for each feature and used DESeq2 in R v3.2.1 to calculate differential expression between conditions.
|
10.1371/journal.pgen.1006711 | Phenome-wide heritability analysis of the UK Biobank | Heritability estimation provides important information about the relative contribution of genetic and environmental factors to phenotypic variation, and provides an upper bound for the utility of genetic risk prediction models. Recent technological and statistical advances have enabled the estimation of additive heritability attributable to common genetic variants (SNP heritability) across a broad phenotypic spectrum. Here, we present a computationally and memory efficient heritability estimation method that can handle large sample sizes, and report the SNP heritability for 551 complex traits derived from the interim data release (152,736 subjects) of the large-scale, population-based UK Biobank, comprising both quantitative phenotypes and disease codes. We demonstrate that common genetic variation contributes to a broad array of quantitative traits and human diseases in the UK population, and identify phenotypes whose heritability is moderated by age (e.g., a majority of physical measures including height and body mass index), sex (e.g., blood pressure related traits) and socioeconomic status (education). Our study represents the first comprehensive phenome-wide heritability analysis in the UK Biobank, and underscores the importance of considering population characteristics in interpreting heritability.
| Heritability of a trait refers to the proportion of phenotypic variation that is due to genetic variation among individuals. It provides important information about the genetic basis of complex traits and indicates whether a phenotype is an appropriate target for more specific statistical and molecular genetic analyses. Recent studies have leveraged the increasingly ubiquitous genome-wide data and documented the heritability attributable to common genetic variation captured by genotyping microarrays for a wide range of human traits. However, heritability is not a fixed property of a phenotype and can vary with population-specific differences in the genetic background and environmental variation. Here, using a computationally and memory efficient heritability estimation method, we report the heritability for a large number of traits derived from the large-scale, population-based UK Biobank, and, for the first time, demonstrate the moderating effect of three major demographic variables (age, sex and socioeconomic status) on heritability estimates derived from genome-wide common genetic variation. Our study represents the first comprehensive heritability analysis across the phenotypic spectrum in the UK Biobank.
| The heritability of a trait refers to the proportion of phenotypic variance that is attributable to genetic variation among individuals. Heritability is commonly measured as either the contribution of total genetic variation (broad-sense heritability, H2), or the fraction due to additive genetic variation (narrow-sense heritability, h2) [1]. A large body of evidence from twin studies has documented that essentially all human complex traits are heritable. For example, a recent meta-analysis of virtually all twin studies published between 1958 and 2012, encompassing 17,804 traits, reported that the overall narrow-sense heritability estimate across all human traits was 49%, although estimates varied widely across phenotypic domains [2]. Over the past decade, the availability of genome-wide genotyping has enabled the direct estimation of additive heritability attributable to common genetic variation (“SNP heritability” or hSNP2) [3–5]. These estimates do not capture non-additive genetic effects such as dominance or epistasis, and provide a lower bound for narrow-sense heritability because they also do not capture contributions (e.g., from rare variants) that are not assayed by most genotyping microarrays and are not well tagged by genotyped variants. Nevertheless, estimates of SNP heritability can provide important information about the genetic basis of complex traits such as the proportion of phenotypic variation that could be explained by common-variant genome-wide association studies (GWAS).
However, heritability is not a fixed property of a phenotype but depends on the population in which it is estimated. As a ratio of variances, it can vary with population-specific differences in both genetic background and environmental variation [1]. For example, twin data have documented variations in the heritability of childhood IQ by socioeconomic status (SES) [6], highlighting that different environment may have different relative contributions to the variance of a phenotype. In addition, heritability estimates for a range of complex phenotypes have been shown to vary according to the sex and age distributions of the sampled populations [2]. Identifying variables that may affect the heritability of complex traits has implications for the design of GWAS, highlighting subgroups and environmental conditions in which common-variant contributions may be diminished or magnified. To date, however, studies of complex trait heritability and the effect of modifying variables have produced mixed results likely due to sample size limitations and population-specific differences in genetic and environmental variance that may be operating in different cohorts.
The UK Biobank (http://www.ukbiobank.ac.uk) provides a unique opportunity to estimate the heritability of traits across a broad phenotypic spectrum in a single population sample. The UK Biobank is a large prospective population-based cohort study that enrolled 500,000 participants aged 40–69 years between 2006 and 2010 [7]. The study has collected a wealth of phenotypic data from questionnaires, physical and biological measurements, and electronic health records as well as genome-wide genotype data. However, this rich data source also presents analytic challenges. For example, with the large sample size, existing heritability estimation methods such as genome-wide complex trait analysis (GCTA) [3–5] and LD (linkage disequilibrium) score regression [8] become computationally expensive and memory intensive, and thus can be difficult to apply. Here we implemented a computationally and memory efficient approach to estimate the heritability for 551 complex traits derived from the interim data release (152,736 subjects) of the UK Biobank, comprising both quantitative phenotypes and disease categories. We then examined how heritability estimates are modified by three major demographic variables: age, sex and socioeconomic status (SES). Our results underscore the importance of considering population characteristics in estimating and interpreting heritability, and may inform efforts to apply genetic risk prediction models for a broad range of human phenotypes.
We report the heritability for 551 traits that were made available to us through the interim data release of the UK Biobank (downloaded on Mar 3, 2016) and had sufficient sample sizes to achieve accurate heritability estimates (standard error of the heritability estimate smaller than 0.1; 15 disease codes excluded) using a computationally and memory efficient heritability estimation method (see Methods and S1 Text).
The 551 traits can be classified into 11 general phenotypic domains as defined by the UK Biobank to group individual data fields into broadly related sets: cognitive function (5 traits), early life factors (7 traits), health and medical history (60 traits), hospital in-patient main diagnosis ICD-10 codes (194 traits), life style and environment (88 traits), physical measures (50 traits), psychosocial factors (40 traits), self-reported cancer codes (9 traits), self-reported non-cancer illness codes (79 traits), sex-specific factors (14 traits), and sociodemographics (5 traits). ICD-10 (the International Classification of Diseases, version-10) is a medical classification list published by the World Health Organization (WHO), which contains thousands of diagnostic codes. Fig 1 shows the percentage of each domain that makes up the 551 traits we analyzed. Using the top-level categories and chapters of the self-reported disease and ICD-10 coding tree, we can further break down self-reported non-cancer illness codes and ICD-10 codes into different functional domains (S1 Fig). We note that since we only analyzed disease codes that had prevalence greater than 1% in the sample, distribution of the disease traits across functional domains was skewed. For example, we investigated a large number of gastrointestinal and musculoskeletal traits, while diseases that have low prevalence in the sampled population such as psychiatric disorders were not well represented.
Table 1 lists the top heritable traits in each domain (the most heritable trait and traits with heritability estimates greater than 0.30). S1 Table and S2 Table show the heritability estimates, standard error estimates, sample sizes, covariates adjusted, prevalence in the sample (for binary traits) and other relevant information for all the traits we analyzed. Common genetic variants appear to have an influence on most traits we investigated, although heritability estimates showed heterogeneity within and across trait domains. Complex traits that exhibited high SNP heritability (larger than 0.40) included human height (0.685+/-0.004), skin color (very fair/fair vs. other, 0.556+/-0.008), ease of skin tanning (very/moderately tanned vs. mildly/occasionally/never tanned, 0.454+/-0.006), comparative height at age 10 (taller than average, 0.439+/-0.007; shorter than average, 0.405+/-0.008), rheumatoid arthritis (0.821+/-0.046), hypothyroidism/myxedema (0.814+/-0.017), malignant neoplasm of prostate (0.426+/-0.093), and diabetes diagnosed by doctor (0.414+/-0.016), among others. On the other end of the spectrum, traits such as duration of walks/moderate activity/vigorous activity, frequency of stair climbing, ever had stillbirth, spontaneous miscarriage or termination, painful gums, stomach disorder, fracture, injuries to the head/knee/leg, and pain in joint had zero or close to zero heritability estimates, indicating that their phenotypic variation is largely determined by environmental factors, or there is widespread heterogeneity or substantial measurement error in these phenotypes. SNP heritability estimates for several phenotypes, including diseases with known immune-mediated pathogenesis (rheumatoid arthritis, psoriasis, diabetes, hypothyroidism), were markedly reduced when the major histocompatibility complex (MHC) region was excluded from analysis (S4 Table), and thus need to be interpreted with caution (see Discussion).
A substantial fraction of the phenotypes we examined were based on self-report illness codes or diagnostic (ICD-10) codes, which may be noisy and have low specificity. However, the SNP heritability estimates for 14 pairs of self-reported illness and ICD-10 codes that represent the same or closely matched diseases were largely consistent and had a Pearson correlation of 0.78 (Table 2), indicating that both phenotypic approaches captured useful and comparable variations in these phenotypes.
Heritability analysis stratified by sex identified a number of traits whose heritability showed significant difference in males and females after multiple testing correction (Fig 2). For example, the analyses of diastolic and systolic blood pressure, and self-reported hypertension and high blood pressure provided consistent evidence that the heritability of blood pressure related traits and diseases is significantly higher in females than in males.
A majority of physical measures showed decreasing heritability with age (S3 Table). More specifically, 33 out of 50 physical measures had a significant decreasing trend in heritability estimates after accounting for multiple testing correction (mean slope of the 33 traits -0.0035, i.e., heritability estimates decrease by 3.5 percent per decade). The age-varying SNP heritability estimates and their standard errors for 12 traits that showed both significant slopes and significantly different heritability estimates between the first (40–49 years) and last age range (64–73 years) are shown in Fig 3A. S2 Fig shows the mean and standard deviation of the 12 traits in each age range.
When we stratified heritability by the Townsend deprivation index, a proxy for SES, education (has college or university degree or not) was the only trait on which SES had a significant moderating effect after accounting for multiple testing correction. Fig 3B shows that the heritability of education increases with increasing SES.
Estimating the heritability of complex, polygenic traits is an important component of defining the genetic basis of human phenotypes. In addition, heritability estimates provide a theoretical upper bound for the utility of genetic risk prediction models [9]. We calculated the common-variant heritability of 551 phenotypes derived from the interim data release of the UK Biobank, and confirmed that common genetic variation contributes to a broad array of quantitative traits and human diseases in the UK population. Two aspects of our work are particularly notable. First, we developed a computationally and memory efficient method that enabled us to calculate the most extensive population-based survey of SNP heritability to date. Second, we found that the heritability for a number of phenotypes is moderated by major demographic variables, demonstrating the dependence of heritability on population characteristics. We discuss each of these advances and the limitations of the biobank data and our analyses below.
Classical methods to estimate SNP heritability, such as GCTA (also known as the GREML method), rely on the restricted maximum likelihood (ReML) algorithm [3–5], which can give unbiased heritability estimates in quantitative trait analysis and non-ascertained case-control studies, and is statistically efficient when the trait is Gaussian distributed [10]. However, ReML is an iterative optimization algorithm, which is computationally and memory intensive, and thus can be difficult to apply when analyzing data sets with hundreds of thousands of subjects. An alternative and widely used SNP heritability estimation method is LD score regression, which is based on GWAS summary statistics and an external reference panel for the LD structure [8]. The approach can thus be easily applied to complex traits on which large-scale GWAS results are available, and allows meta-analysis of heritability estimates from different studies. Recently, LD score regression has been extended to partition heritability by functional annotation [11], and to estimate the genetic correlation between two traits [12,13]. However, when applying LD score regression to novel phenotypes in a large cohort, conducting GWAS is often time-consuming.
In the present study, we implemented a computationally and memory efficient moment-matching method for heritability estimation, which is closely related to the Haseman-Elston regression [14–16] and phenotype-correlation genetic-correlation (PCGC) regression [10], and produces unbiased SNP heritability estimates for both continuous and binary traits. The moment-matching method is theoretically less statistically efficient than the ReML algorithm (i.e., produces larger standard error on the point estimate) when analyzing quantitative traits, but the power loss is expected to be small [17] and is less of an issue given large sample sizes, such as in the UK Biobank. The moment-matching method is also mathematically equivalent to LD score regression if the following conditions are satisfied: (1) the out-of-sample LD scores estimated from the reference panel and the in-sample LD scores estimated from individual-level genotype data are identical; (2) the intercept in the LD score regression model is constrained to 1 (i.e., assuming that there is no confound and population stratification in the data); and (3) a particular weight is used in the LD score regression (more specifically, the reciprocal of the LD score, which is close to the default setting in the LD score regression software) [18]. Here, since we have constrained our analysis to a white British (Caucasian) sample and have accounted for potential population stratification by including top PCs of the genotype data as covariates, the two methods should produce similar estimates. See Box 1 for an empirical comparison between the moment-matching method, LD score regression and GCTA.
Using the moment-matching method, we found that a large number of traits we examined display significant heritability. For traits whose heritability has been intensively studied, our estimates are generally in line with prior studies. For example, twin and pedigree studies have estimated the heritability of human height and body mass index (BMI) to be approximately 80% and 40–60% [see e.g., 19–21], respectively, although recent studies have shown that heritability may be overestimated in family studies due to, for instance, improper modeling of common environment, assortative mating in humans, genetic interactions, and suboptimal statistical methods [10,22–25]. Using genome-wide SNP data from unrelated individuals, it has been shown that common SNPs explain a large proportion of the height and BMI variation in the population, although SNP heritability estimates are lower than twin estimates [4,5,26]. Specifically, the first GCTA analysis estimated the SNP heritability of human height to be 0.45 using relatively sparse genotyping data (approximately 300,000 SNPs) and showed that the estimate could be higher if imperfect LD between SNPs and causal variants are corrected [4]. A more recent study leveraging whole-genome sequencing data and imputed genetic variants concluded that narrow-sense heritability is likely to be 60–70% for height and 30–40% for BMI [27]. Here, we estimated the SNP heritability of human height and BMI to be 0.685+/-0.004 and 0.274+/-0.004, respectively, which are comparable to the expected range. The SNP heritability estimates of other complex traits of interest, such as age at menarche in girls (0.239+/-0.007), diastolic (0.184+/-0.004) and systolic (0.156+/-0.004) blood pressures, education (has colleague or university degree or not, 0.294+/-0.007), neuroticism (0.130+/-0.005), smoking (ever smoked or not, 0.174+/-0.006), asthma (0.340+/-0.010) and hypertension (0.263+/-0.007) were also more modest and lower than twin estimates, as expected [2].
Heritability is, by definition, a ratio of variances, reflecting the proportion of phenotypic variance attributable to individual differences in genotypes. Because the genetic architecture and non-genetic influences on a trait may differ depending on the population sampled, heritability itself may vary. Examples of this have been reported in the twin literature. In one well-known study, Turkheimer and colleagues [6] reported that the heritability of IQ is moderated by SES in a sample of 320 7-year-old twin pairs of mixed ancestry. In that study, the heritability of IQ was essentially 0 at the lowest end of SES but substantial at the highest end. Subsequent studies of the moderating effects of SES on the heritability of cognitive ability and development using twin designs have produced mixed results [28–33]. In our analysis, using SNP data, we observed no moderating effect of SES (as measured by the Townsend deprivation index) on the heritability of cognitive traits (including fluid intelligence), possibly due to the age range of participants in the UK Biobank (middle and old age) in contrast to many previous studies targeting childhood or early adulthood, and the cross-national differences in gene-by-SES interaction on intelligence as shown by a recent meta-analysis [34]. In addition, the brief cognitive tests available in the UK Biobank may have had limited sensitivity for capturing individual differences in IQ (see discussion below). On the other hand, the heritability of education showed significant interactions with SES, with increasing heritability at higher SES levels. Prior evidence has suggested that education has substantial genetic correlation with IQ and may be a suitable proxy phenotype for genetic analyses of cognitive performance [35]; thus our results may indirectly support earlier studies of the SES moderation of IQ heritability.
With two exceptions, significant sex differences we observed indicated greater heritability for women compared to men. Our results are consistent with findings from some twin studies but not others. For example, we found that women exhibited significantly greater heritability for measured waist circumference and blood pressure. Twin studies have also reported greater female heritability for waist circumference [36] but no substantial sex difference in heritability of blood pressure [2,37]. A substantial difference between the heritability of rheumatoid arthritis (RA) in males compared to females was observed, although the MHC region has a large impact on the SNP heritability estimates of autoimmune diseases, and thus this finding needs to be interpreted with caution (see discussion below). While RA is known to be more common in women, a twin analysis found no sex difference in heritability among Finnish and UK twin pairs, though power was limited in that analysis [38]. Intriguingly, greater heritability was observed among men for the personality trait of miserableness, a component of neuroticism, suggesting that environmental factors may be more influential for this trait among women or that measurement error differs by sex.
We examined age effects on heritability for a subset of variables and found that a number of physical measures indexing body size, adiposity, height, as well as systolic blood pressure and lung function, showed declining heritability with age. Age-related declines in heritability may reflect the cumulative effect of environmental perturbations over the lifespan. Prior twin studies of age effects on the heritability of anthropometric traits in adults have had inconsistent results [39–41]. Haworth and colleagues showed that the heritability of BMI increases over childhood [42]. A recent meta-analysis of 32 twin studies documented a non-monotonic relationship between BMI heritability and age (from childhood to late adulthood), with a peak around age 20 and decline thereafter [43]. An age-related decline in indices of body size may reflect a decreasing contribution of genetically-regulated growth processes over the lifespan. However, we were unable to assess the entire trajectory of heritability due to the age range (40–73 years) of the UK Biobank participants. Some but not all studies have also suggested varying or declining heritability with age for blood pressure, lung function and age at first birth [39,44–50].
Our results should be interpreted in light of the limitations associated with the biobank data. First, the UK Biobank is restricted to middle and old age groups, which may be subject to sample selection bias. For example, older and physically/cognitively impaired subjects may be underrepresented in the study, which may have an impact on the heritability estimates stratified by age. Mortality selection can also alter the results of genetic analyses as shown by recent analyses [51]. In addition, the UK Biobank participants comprised a relatively high proportion of well-educated, skilled professionals [52], potentially leading to the underrepresentation or restricted range of certain traits such as smoking relative to other cohorts. Therefore, our heritability estimates may be specific to this UK population and may not generalize to other settings or ancestry groups. Second, although the UK Biobank has collected a wealth of phenotypes, measurements associated with a particular phenotypic domain may not be comprehensive. For example, only five cognitive tests were included in the UK Biobank. The reasoning task (fluid intelligence test) was brief and had a narrow range; the reaction time was averaged from a small number of trials; and the visual memory test (pairs-matching test) had a significant floor effect (a large number of participants made zero or very few mistakes, and thus the scores do not fully reflect individual differences). In addition, all cognitive tests had relatively low reliability across repeat measurements [53]. These noisy measurements may thus downwardly bias heritability estimates of cognition. The Townsend deprivation index, which we used to stratify phenotypes, was calculated based on the national census output area of each participant in which their postcode was located at the time of recruitment, and thus can only serve as a proxy for SES. Third, the phenotypes were limited to those for which we had sufficient data to estimate heritability with adequate precision. Therefore, diseases with low prevalence in the sampled population were not well represented in our analysis. We expect to analyze traits with lower prevalence (e.g., 0.5%) when the genetic data for all UK Biobank participants becomes available. We also assumed in our analysis that the population prevalence of a binary trait is identical to the observed sample prevalence, but diseases such as schizophrenia and stroke are naturally under-ascertained and thus their sample prevalence is often lower than population prevalence. In addition, we note that since we used medical history to define cases and controls, the prevalence of many diseases we investigated reflected lifetime prevalence, which may be different from cross-sectional prevalence used in other studies. We also binarized categorical (multinomial or ordinal) variables to facilitate analysis, but this might not optimally represent variation in these variables with respect to heritability. Fourth, a substantial fraction of the phenotypes we examined were based on self-report or diagnostic (ICD-10) codes, which may or may not validly capture the phenotypes they represent. For example, a recent UK Biobank study shows that 51% of the participants who reported RA were not on RA-relevant medication, a proxy measure of valid diagnosis [54]. However, our head-to-head comparison of the heritability estimates between self-reported illness and ICD-10 codes showed largely consistent results, indicating that both phenotypic approaches at least captured comparable variations in these phenotypes. Prior research evaluating phenotypes derived from electronic health records (EHR) indicate that greater phenotypic validity can be achieved when diagnostic codes are supplemented with text mining methods [55–58]. The specificity of the disease codes might also be improved by leveraging the medication records in the UK Biobank.
Methodologically, our SNP heritability estimation approach, despite its superior computational and memory performance compared to existing methods, also has several limitations. First, heritability estimation always relies on a number of assumptions on the genetic architecture. For example, the moment-matching method we used here, as well as the established GCTA and LD score regression approaches, implicitly assumes that the causal SNPs are randomly spread over the genome, which is independent of the MAF spectrum and the LD structure, and the effect sizes of causal SNPs are Gaussian distributed and have a specific relationship to their MAFs. Although it has been shown that SNP heritability estimates are reasonably robust to many of these modeling assumptions [59], the estimates can be biased if, for instance, causal SNPs are rarer or more common than uniformly distributed on the MAF spectrum, or are enriched in high or low LD regions across the genome. For example, the heritability estimates for some autoimmune diseases such as psoriasis and RA dropped dramatically when the MHC region (chr6:25-35Mb) was removed when constructing the genetic similarity matrix, indicating, as expected, that causal variants for these diseases are disproportionally enriched in the MHC region. S4 Table lists all the traits whose heritability estimates decreased by 0.2 or more when the MHC region was taken out, and thus need to be interpreted with caution. Methods to correct for MAF properties and region-specific LD heterogeneity of causal variants have been proposed [27,59,60]. For example, we can stratify MAF and LD structure into different bins, compute a genetic similarity matrix within each bin, and fit a mixed effects model with multiple variance components [27,60]. This approach can give heritability estimates that are more robust to properties of the underlying genetic architecture, but has the downside of increased computational burden and reduced statistical power. A different direction to explore is to estimate SNP heritability using imputed data (in contrast to the genotype data here), which might capture more genetic variation from rare variants, or common variants that are not well tagged by the genotyped SNPs, and thus lead to increased heritability estimates. Second, heritability analysis models, including the one we employed in the present study, typically assume that genetic and environmental effects are independent, i.e., no gene-by-environment (GxE) interaction exists. This is certainly a simplification of the real world where GxE interactions are expected for many complex traits. Recent computational studies have also shown that ignoring GxE interactions in heritability analysis can produce biased estimates [25]. However, modeling GxE would require collecting relevant environmental variables for each phenotype and more sophisticated statistical modeling approaches, e.g., incorporating multiple random effects in the heritability analysis model [5,61]. Due to the limited measurements of environment collected by the UK Biobank, and the extensive analyses we have conducted across the phenotypic spectrum, explicitly modeling the environmental factors and GxE interactions is not feasible. We therefore took an alternative approach to examine the moderating effects of three major demographic variables on heritability estimates by stratifying samples. Of note, consistent heritability estimates across different levels of the stratifying variable do not completely eliminate the potential existence of GxE interactions. Specifically, recent studies have identified genetic heterogeneity in human traits such as BMI and fertility [62,63], indicating that the genetic architecture of a trait may be different across environments (i.e., the genetic correlation of a trait in different environments may be significantly smaller than 1) even if the overall heritability estimates are similar. Dissection of common and unique environmental influences and their interactive effects with genetics on different complex traits, and the shared and unique genetic effects across environments are important future directions to explore. Lastly, as reviewed in [64], a number of empirical genetic similarity measurements computable from genome-wide SNP data have been proposed, which, when utilized in heritability analysis, can give different estimates with different interpretations. In addition, recent studies have argued that estimation error associated with genetic similarity measurements and the ill-posedness of the empirical genetic similarity matrix may produce unstable and unreliable SNP heritability estimates [65]. However, this is an area under active investigation and debate [64–67]. Here, as the first study to screen all UK Biobank variables and provide an overview of the distribution of SNP heritability across different trait domains, and to examine the effect of potential modifying variables on heritability estimates, we used a straightforward and classical modeling approach that is most widely used. To obtain more insights into the genetic architecture and find the most appropriate and robust model for each individual trait, more systematic investigation is needed.
In sum, using a computationally and memory efficient approach, we provide estimates of the SNP heritability for 551 complex traits across the phenome captured in the population-based UK Biobank. We further identify phenotypes for which the contribution of genetic variation is modified by demographic factors. These results underscore the importance of considering population characteristics in interpreting heritability, highlight phenotypes and subgroups that may warrant priority for genetic association studies, and may inform efforts to apply genetic risk prediction models for a broad range of human phenotypes.
This study utilized deidentified data from the baseline assessment of the UK Biobank, a prospective cohort study of 500,000 individuals (age 40–69 years) recruited across Great Britain during 2006–2010 [7]. The protocol and consent were approved by the UK Biobank’s Research Ethics Committee.
The UK Biobank collected phenotypic data from a variety of sources including questionnaires regarding mental and physical health, food intake, family history and lifestyle, a baseline physical assessment, computerized cognitive testing, linkage with health records, and blood samples for biochemical and DNA analysis. Details about the UK Biobank project are provided at http://www.ukbiobank.ac.uk. Data for the current analyses were obtained under an approved data request (Ref: 13905).
The interim release of the genotype data for the UK Biobank (downloaded on Mar 3, 2016) comprises 152,736 samples. Two closely related arrays from Affymetrix, the UK BiLEVE Axiom array and the UK Biobank Axiom array, were used to genotype approximately 800,000 markers with good genome-wide coverage. Details of the design of the arrays and sample processing can be found at http://biobank.ctsu.ox.ac.uk/crystal/refer.cgi?id=146640 and http://biobank.ctsu.ox.ac.uk/crystal/refer.cgi?id=155583.
Prior to the release of the genotype data, stringent quality control (QC) was performed at the Wellcome Trust Centre for Human Genetics, Oxford, UK. Procedures were documented in detail at http://biobank.ctsu.ox.ac.uk/crystal/refer.cgi?id=155580. We leveraged the QC metrics made available by the UK Biobank and removed samples that had mismatch between genetically inferred sex and self-reported sex, samples that had high genotype missingness or extreme heterozygosity not explained by mixed ancestry or increased levels of marriage between close relatives, and one individual from each pair of the samples that were 3rd degree or more closely related relatives. We restricted our analysis to subjects that were self-reported white British and confirmed by principal component analysis (PCA) to be Caucasians. We further filtered out genetic markers that had high missing rate (>1%), low minor allele frequency (<1%), significant deviation from Hardy-Weinberg equilibrium (p<1e-7), and subjects that had high missing genotype rate (>1%). 108,158 subjects (age 40–73 years; female 52.84%) and 486,175 SNPs remained for analysis after QC. S4 Fig shows the age distribution of the subjects that passed QC. The genetic similarity matrix was computed using all genotyped autosomal SNPs. All genetic analyses were performed using PLINK 1.9 (https://www.cog-genomics.org/plink2) [68].
We analyzed every trait available to us that had a sufficient sample size to produce a heritability estimate with its standard error smaller than 0.1. The traits can be classified into the following 11 domains as defined by the UK Biobank: cognitive functions, early life factors, health and medical history, life style, physical measures, psychosocial factors, sex-specific factors and sociodemographics. For continuous traits, we excluded samples that were more than 5 standard deviations away from the population mean to avoid extreme outliers and data recording errors. We only analyzed binary traits that had prevalence greater than 1% in the sample, so that we had enough statistical power to get reliable heritability estimates. We typically binarized categorical variables at a meaningful threshold close to the median and then analyzed them as binary traits. For the specific cutoff-points used to binarize each categorical variable, see S1 Table.
We also analyzed a large number of self-reported illness codes and hospital in-patient diagnosis codes. Self-reported cancer and non-cancer illness codes were obtained through a verbal interview by a trained nurse at the UK Biobank assessment center on past and current medical conditions. Hospital in-patient diagnoses were obtained through medical records and were coded according to the International Classification of Diseases version-10 (ICD-10). Disease codes for each domain (self-reported cancer, self-reported non-cancer illness, and ICD-10) were organized in a hierarchical tree structure; codes closer to the root of the tree are often less specific and have larger prevalence, while codes closer to the leaves are more specific but have lower prevalence. We analyzed every disease code that had prevalence greater than 1% in the sample. 15 ICD-10 codes were excluded due to small sample sizes and large standard errors (>0.1) on heritability estimates (N70-N77 Inflammatory diseases of female pelvic organs; O20-O29 Other maternal disorders predominantly related to pregnancy; O80-O84 Delivery; N50 Other disorders of male genital organs; N80 Endometriosis; N81.1 Cystocele; N81.2 Incomplete uterovaginal prolapse; N83 Noninflammatory disorders of ovary—Fallopian tube and broad ligament; N83.2 Other and unspecified ovarian cysts; N84.1 Polyp of cervix uteri; N92.1 Excessive and frequent menstruation with irregular cycle; N93 Other abnormal uterine and vaginal bleeding; O68 Labour and delivery complicated by foetal stress [distress]; O70.1 Second degree perineal laceration during delivery; O80 Single spontaneous delivery). We also employed a data-driven approach to determine if a disease is sex-specific. More specifically, if the sample prevalence of a disease in males was more than 100 times larger than the sample prevalence in females, we defined the disease as male-specific and the analysis was restricted to males. The same approach was used to find female-specific diseases. See S2 Table for all the disease codes we analyzed.
We consider the linear random effect model y = g + e, where an N-dimensional trait y is partitioned into the sum of additive genetic effects g and unique (subject-specific) environmental effects e. The covariance structure of y is cov[y]=σg2K+σe2I, where K is the empirical genetic similarity matrix for each pair of individuals estimated from genome-wide SNP data [4,5], I is an identity matrix, σg2 and σe2 are the total additive genetic variance captured by genotyped common SNPs and the variance of unique environmental factors across individuals, respectively. SNP heritability is then defined as hSNP2=σg2/(σg2+σe2)=σg2/σp2, which measures the total phenotypic variance σp2 that can be explained by total additive genetic variance tagged by genotyped SNPs, and is a lower bound for the narrow-sense heritability h2. When covariates need to be incorporated into the model, i.e., y = Xβ + g + e, where X is an N × q covariate matrix and β is a vector of fixed effects, an N × (N − q) matrix U always exists, which satisfies UTU = I, UUT = P0, UTX = 0, and P0 = I − X(XTX)−1XT. Applying UT to both sides of the model removes the covariate matrix [69,70].
To obtain unbiased estimates of σg2 and σe2, we used a computationally efficient moment-matching approach [70,71], which is closely related to the Haseman-Elston regression [14–16] and phenotype-correlation genetic-correlation (PCGC) regression [10], and mathematically equivalent to the LD score regression under certain conditions [8,18]. Specifically, we regress the empirical estimate of the phenotypic covariance onto the matrices K and I: vec[yyT]=σg2vec[K]+σe2vec[I]+ϵ, where vec[⋅] is the matrix vectorization operator that converts a matrix into a vector by stacking its columns, and ϵ is the residual of the regression. The ordinary least squares (OLS) estimator of this multiple regression problem can be explicitly written as σ^g2=1vKyT(K−τI)y and σ^e2=1vKyT(κI−τK)y, where τ = tr[K]/N, κ = tr[K2]/N, and vK = N(κ − τ2). SNP heritability is then estimated as h^SNP2=σ^g2/(σ^g2+σ^e2)=σ^g2/σ^p2.
To estimate the sampling variance of h^SNP2, we follow Visscher et al. [17] and make two assumptions: (1) the off-diagonal elements in the empirical genetic similarity matrix K are small, such that K ≈ I and V=cov[y]=σg2K+σe2I≈σp2I; and (2) the phenotypic variance σp2 can be estimated with very high precision. We thus have var[σ^g2]=2vK2tr[(K−τI)V(K−τI)V]≈2σp4/vK, and var[h^SNP2]≈2/vK. This estimator coincides with existing results in the literature [17]. We note that the calculation of the variance of σ^g2 relies on an additional assumption that the trait y is Gaussian distributed and thus may be suboptimal for binary traits. However, Visscher and colleagues have empirically shown that this sampling variance approximation is accurate for both continuous and binary traits when the sample size is large [17] (also see S3 Fig).
We note that for large sample size N, the N × N genetic similarity matrix K and residual forming matrix P0 can be very large, making the computation memory intensive. We have developed a memory efficient algorithm that can iteratively load columns (or block columns) of K into the memory to compute the SNP heritability estimate, and does not need to explicitly compute P0 or any other N × N matrices. See S1 Text for details. Matlab and Python implementations of the algorithm are available at https://github.com/chiayenchen/mmhe.
For binary traits, the above calculation gives a heritability estimate on the observed scale, which is dependent on prevalence of the trait in the population. We transformed this heritability estimate to the underlying liability scale under the assumption of a classical liability threshold model [72,73], which makes heritability estimates independent of prevalence and thus comparable across traits. Specifically, heritability estimate on the liability scale can be obtained using a linear transformation of the heritability on the observed scale: h^SNP,L2=ch^SNP2, where c = P(1 − P)/φ(t)2, P is the population prevalence, t = Φ−1(1 − P) is the liability threshold, Φ is the cumulative distribution function of the standard normal distribution, and φ is the density function of the standard normal distribution [3,74]. Since the UK Biobank is not designed to be ascertained for particular diseases, we assumed that population prevalence is identical to sample prevalence. The sampling variance of the heritability estimate can be transformed accordingly: var[h^SNP,L2]=c2var[h^SNP2].
In all heritability analyses, we included genotyping array, UK Biobank assessment center, age at recruitment and top 10 principal components (PCs) of the genotype data as covariates. Other covariates such as sex and handedness (e.g., when analyzing the grip strength of the left/right hand) were adjusted where appropriate. See S1 Table for the set of covariates we included in the model when estimating the heritability for each trait. To compute PCs of the genotype data, we performed pairwise linkage disequilibrium (LD) based SNP pruning at R2>0.02 and excluded SNPs in the major histocompatibility complex (MHC) region (chr6:25-35Mb) and chromosome 8 inversion (chr8:7-13Mb). Top PCs were then computed using flashPCA [75] on the pruned data, which employs an efficient randomized algorithm and is thus scalable to large data sets with hundreds of thousands of individuals.
To examine how heritability estimates are modified by sex, we estimated heritability for each non-sex-specific trait in males and females separately. For binary traits, sample prevalence was calculated in each sex. To test if heritability estimates are significantly different by sex, we assumed that the two SNP heritability estimates to be contrasted, h^A and h^B, are independent and approximately Gaussian distributed, and computed the z-score of their difference: z=(h^A−h^B)/se^A2+se^B2, where se^A2 and se^B2 are standard error estimates of h^A and h^B, respectively. A p-value can then be computed as p = 2 ∙ Φ(−|z|), where Φ is the cumulative distribution function of the standard normal distribution.
To examine whether SNP heritability estimates vary with age, we used a sliding window approach and estimated heritability for every age range of 10 years (i.e., 40–49 years, 41–50 years, …, 64–73 years) by stratifying the samples. For binary traits, sample prevalence was calculated in each age range separately. We assessed whether heritability estimates exhibited a linear trend with age by fitting a regression model, h^k2=α+age¯kγ+ϵk, where h^k2 is the heritability estimate in the k-th age range, age¯k is the mean of the age range, α is an intercept, γ is the slope and ϵk is the residual of the regression, and testing whether γ is significantly different from zero. We weighted heritability estimates by the inverse of their standard errors when fitting the regression model, and thus put more emphasis on estimates with better precision. We only analyzed physical and cognitive measures, and did not consider disease codes and medical history in age stratification analyses because age at recruitment does not reflect disease onset.
Similarly, we used a sliding window approach to estimate the SNP heritability for each trait from the bottom 1/3 quantile to the top 1/3 quantile of the Townsend deprivation index at recruitment, a measure of material deprivation within the population of a given area. For binary traits, sample prevalence was calculated in each SES bin separately. For traits that do not reflect the status of participants at the time of recruitment (e.g., medical history and early-life factors), we have implicitly made an assumption that the SES of participants had not changed dramatically throughout their lives.
To account for multiple testing in our stratification analyses, we corrected the p-values using the effective number of independent traits we analyzed. Specifically, for each stratification analysis (sex, age and SES), we calculated the Pearson correlation coefficient for each pair of the traits using their overlapping samples. The correlation between traits that had no sample overlap, e.g., male- and female-specific factors, was set to zero. We then conducted a principal component analysis (PCA) to the constructed phenotypic correlation matrix, and estimated the effective numbers of independent traits that explained 99% of the total phenotypic variation in sex, age and SES stratification analyses to be 400, 31 and 440, respectively. Finally, we multiplied uncorrected p-values by the corresponding effective number of independent traits to obtain corrected p-values.
|
10.1371/journal.pgen.1006134 | Next-Generation “-omics” Approaches Reveal a Massive Alteration of Host RNA Metabolism during Bacteriophage Infection of Pseudomonas aeruginosa | As interest in the therapeutic and biotechnological potentials of bacteriophages has grown, so has value in understanding their basic biology. However, detailed knowledge of infection cycles has been limited to a small number of model bacteriophages, mostly infecting Escherichia coli. We present here the first analysis coupling data obtained from global next-generation approaches, RNA-Sequencing and metabolomics, to characterize interactions between the virulent bacteriophage PAK_P3 and its host Pseudomonas aeruginosa. We detected a dramatic global depletion of bacterial transcripts coupled with their replacement by viral RNAs over the course of infection, eventually leading to drastic changes in pyrimidine metabolism. This process relies on host machinery hijacking as suggested by the strong up-regulation of one bacterial operon involved in RNA processing. Moreover, we found that RNA-based regulation plays a central role in PAK_P3 lifecycle as antisense transcripts are produced mainly during the early stage of infection and viral small non coding RNAs are massively expressed at the end of infection. This work highlights the prominent role of RNA metabolism in the infection strategy of a bacteriophage belonging to a new characterized sub-family of viruses with promising therapeutic potential.
| The increase of the proportion of multidrug resistant bacterial strains is alarming and alternative ways to treat infections are necessary such as the use of the natural enemies of bacteria, also known as phage therapy. However, explorations of the molecular mechanisms underlying the viral cycle of bacteriophages have been so far restricted to a small number of viruses infecting model bacteria such as Escherichia coli. By combining next-generation transcriptomics and metabolomics approaches, we have now demonstrated that the virulent bacteriophage PAK_P3, infecting the opportunistic pathogen Pseudomonas aeruginosa, directly interferes with specific host metabolic pathways to complete its infection cycle. In particular, it triggers a dramatic degradation of host RNAs and stimulates bacterial pyrimidine metabolism to promote a nucleotide turnover. Overall, we found that upon PAK_P3 infection, host metabolism is redirected to generate the required building blocks for efficient viral replication. We also showed that PAK_P3 gene expression relies on RNA-based regulation strategies using small non coding RNAs and antisense RNAs. Our findings highlight the molecular strategies employed by this virulent phage, which is a representative of a new subfamily of viruses shown to display promising therapeutic values.
| The threat of antibiotic resistance has renewed attention to phage therapy leading to isolation of many bacteriophages (phages) targeting human pathogens such as Pseudomonas aeruginosa and, consequently, an increasing number of phage genome sequences are available [1]. Comparative genomics has allowed the implementation of a genome-based taxonomy for tailed phages which reflects their great diversity. However, the lack of knowledge of molecular mechanisms underlying their infectious cycles is slowing down their global acceptance as valid therapeutics. Indeed, outside basic characterizations (e.g. phage growth parameters, identification of bacterial receptors and phage structural proteins) many questions about their infection strategy remain conspicuously unanswered for most phages, mainly because genome annotations cannot provide hints on the functions of many viral genes.
For Pseudomonas phages, the introduction of whole-transcriptome studies with RNA-Sequencing (RNA-Seq) has recently led to improved genome annotations, discovery of regulatory elements and elucidation of temporal transcriptional schemes, while at the same time looking at the impact on transcription regulation of host genes upon phage infection. For example, giant phage ϕKZ is now understood to infect and lyse its host cell as well as produce phage progeny in the absence of functional bacterial transcriptional machinery [2]. The impact of phage infection on the host can also be observed at the metabolome level. Recently, a high coverage metabolomics analysis comparing several viruses that cover most genera of Pseudomonas phages infecting strain PAO1, revealed specific phage and infection-stage alterations of the host physiology. These changes often appear mediated by phage-encoded auxiliary metabolic genes (AMGs) and by host gene features that are specifically modulated by the phage [3].
One Pseudomonas phage clade that has not yet been studied is comprised of the two newly proposed genera (PAK_P1-like and KPP10-like) belonging to a new subfamily of viruses, Felixounavirinae. Interestingly, these phages display the best therapeutic potential in an experimental murine lung infection model as compared to other P. aeruginosa phages belonging to distinct clades [4]. Aside from structural genes, most of their predicted ORFs could not be associated with a putative function and consequently, no meaningful conclusions about their strategy for hijacking host metabolism could be drawn [5].
In this work, we used synergistic next generation approaches to provide the first parallel transcriptomics and metabolomics analyses on phage PAK_P3, a representative of the KPP10-like genus. We intended to draw a detailed global scheme of PAK_P3 infectious cycle by addressing the following questions: Does PAK_P3 control expression of specific bacterial genes? Does it interfere with bacterial metabolism? How does it regulate its gene expression?
Our major finding is the predominant role of RNA metabolism in PAK_P3 infectious strategy. Beside the dramatic global depletion of host transcripts induced by phage infection, PAK_P3 causes a strong up-regulation of a single specific host operon. Consistently, an increase of pyrimidine metabolism upon infection was revealed by metabolomics analysis showing that, like T-even phages, PAK_P3 actively manages nucleotides scavenged from their hosts [6]. In addition, besides revealing the temporal expression of PAK_P3 genes, we highlighted an unexpected prominent role of RNA-based regulation of phage gene expression. Indeed, PAK_P3 produces early antisense transcripts encompassing structural genes as well as phage-encoded small non coding RNAs.
To study bacterial transcriptional response to PAK_P3 infection, it was first crucial to exhaustively characterize the genome of its host, P. aeruginosa strain PAK. Initially, a draft genome was produced and assembled (6.28 Mbp, 66.3% GC content and 6,267 predicted ORFs). Next, a detailed genome reannotation was performed based on RNA-Seq data generated from exponentially growing and uninfected PAK cells (S1 Table).
Using COV2HTML [7] to visualize transcripts, we manually reannotated 32 open reading frames (ORFs) (by detection of an alternative start codon) and defined 63 new putative coding sequences. Among them, 39 have been previously annotated in other P. aeruginosa genomes while the other 24 are new hypothetical coding sequences that display no homology to sequences in databases and may be considered as strain-specific (S1 Table).
Recent genome-wide studies based on RNA-Seq led to the discovery of a substantial number of non-coding RNAs (ncRNAs), which are now acknowledged as important modulators of various bacterial processes (for a review of ncRNAs in Pseudomonas aeruginosa, see [8]). We identified a total of 75 small ncRNAs encoded in the PAK genome, 26 of which correspond to known functional classes (S1 Fig, S1 Table). Among these 26, 12 are similar to uncharacterized ncRNA conserved within the Pseudomonas genus and the other 14 have predicted functional assignments, according to Rfam. The majority of ncRNAs (49 out of 75) could not be assigned to any functional class (see Methods), and have not been identified in previous RNA-Seq investigations carried out on P. aeruginosa strains PAO1 and PA14 [9–11], suggesting that they may represent novel ncRNAs regulators.
Eighteen long antisense RNAs (asRNAs) were also identified within genes. As they do not display any consistent ORFs, they are not likely to contain overlapping protein-coding genes and may therefore cis-interfere with the expression of gene they are encoded in (S1 Fig, S1 Table). Finally, 32 potential riboswitches were identified by looking at intergenic transcription events starting at a significant distance from a downstream gene, usually involved in a metabolic pathway and displaying a characteristic RNA-Seq pattern. Eleven of them were confirmed by Rfam search (S1 Fig, S1 Table).
With more than 50% of new ncRNAs amongst total ncRNAs identified, along with the identification of new putative riboswitches and evidence of antisense transcripts, strain PAK exemplifies the great diversity of bacterial RNA-based regulation [12]. Such in-depth annotation, including new strain-specific RNA elements, was mandatory for the subsequent transcriptomic analysis of phage infected cells in order to assess the impact of phage infection on host physiology.
To study the dynamics of the transcriptional and metabolic consequences of phage infection, we first selected the most relevant time points, representative of the different steps of the course of infection by determining the growth parameters of PAK_P3. Adsorption assays revealed that ≥90% of PAK_P3 virions adsorbed on strain PAK within 4.6 ±0.7 min (ka = 2.2.10- 9 ±5.1.10−10 mL.min-1) (Fig 1A). A standard one step growth experiment showed that the first functional new virions are rapidly assembled (eclipse period: 12.3 ±0.4 min) and almost immediately released (latency period: 13 ±2.1 min), producing an average of 53 ±21 progeny phages per infected cell (Fig 1B). With a mean infection cycle duration as short as 18 ±0.6 min, PAK_P3, with a genome length ≥80 kb, is faster than the myoviruses ϕKZ (60–65 min, 280 kb) [3] and T4 (25–30 min, 168 kb) [13], therefore being among the most rapid Myoviridae.
Given the short eclipse period duration, we focused on 3.5 min and 13 min time points as representative snapshots of the beginning (early) and the end (late) of one infection cycle at the transcription level. Investigation of the regulation of both viral and host gene expression over a single phage infection cycle by RNA-Seq revealed a progressive and dramatic replacement of host mRNA with phage transcripts. This process eventually results in host transcripts representing fewer than 13% of non-ribosomal RNAs in the cell (Fig 2). However, even in the context of this dramatic depletion of host transcripts, a response to phage infection at the transcription level was observed, suggesting a globally accelerated degradation of unstable mRNA species rather than a global transcriptional repression as described for phage T4 [14].
In addition to providing a transcriptional environment fully co-opted by the phage for optimal infection (i.e. making host RNA polymerase available for viral RNAs for instance), this observed host RNA depletion can be expected to suppress host defenses that require host transcripts to function [15] as well as prophage induction attempts. Indeed, PAK_P3 infection appears to activate the transcription of a P2-like prophage (Fig 3, S2 Table) as corresponding transcripts display a 6.8-fold increase in PAK_P3 infected cells at late time point compared to non-infected cells. However, host transcripts overall were depleted by 7.2 fold at the late time point, which would leave the infected cell with marginally fewer prophage transcripts than during exponential growth, indicating that the transcriptional activation of the prophage is suppressed, although not completely blocked.
Although host transcripts are globally replaced by phage transcripts, we could still analyze the changes in host mRNA population by normalizing the host transcript counts before infection to the counts after infection, artificially depleting counts before infection and enriching reads after infection. This allows us to look for specific differential expression of host gene features in response to the stress of phage infection as well as specific changes in host gene expression imposed by the phage in order to hijack cellular metabolism.
We discovered that one operon, comprising six genes (PAK_4493–4499), has a nearly 80-fold increase in abundance relative to other host genes, which is large enough to strongly enrich its transcript abundance relative to the total RNA in the cell even in the context of global RNA degradation (Fig 3, S2 Table).
RNA-Seq analysis thus provided precise depictions of phage influence on the bacterial transcriptome and host transcriptional response to infection. It also allowed us to decipher the transcriptional strategy adopted by the phage to control its own gene expression (see below). To have a broader view of the consequences of a phage infection on host cell physiology, we performed a complementary metabolomics analysis.
Viruses depend on host cell metabolic resources to complete their intracellular parasitic development [16]. However, the effects of phage infection on host metabolism are still poorly understood. We thus investigated whether the phage completely shuts off host metabolism, as it may burden efficient phage replication, or if it influences specific pathways.
To assess the impact of PAK_P3 infection on strain PAK metabolism, high-coverage metabolomics analysis was applied to monitor metabolite dynamics during infection [17]. Comparison of the metabolite levels at different time points post infection to uninfected samples revealed significant metabolic changes upon phage infection. Within the first 5 min of infection, 22% of measured metabolites display altered levels with 13.8% increased and 8.5% decreased (p-value ≤ 0,05, │Log2(fold change)│ ≥ 0,5). The proportion of metabolites with increased levels gradually rises up to 22% at 25 min post infection, while the proportion of metabolites with decreased levels temporarily drops to 3% to finally increase back to 13% during bacterial lysis (Fig 4). These variations indicate that PAK_P3 does not simply deplete available host metabolites but relies on an active metabolism in agreement with recent observations identifying phage-specific physiological alterations [3,18].
Next, to investigate whether PAK_P3 targets specific metabolic pathways, a metabolite set enrichment analysis was performed. Overall, metabolites from amino/nucleotide sugar and pyrimidine metabolic pathways were found over-represented among increasing metabolites, while amino acid-related pathways were enriched among decreasing metabolites at later stages of infection (Fig 5). Intriguingly, about 50% of the detected (deoxy)nucleotides-phosphates have at least two-fold increased levels during infection (S3 Table).
Among accumulating metabolites belonging to amino/nucleotide sugar metabolism and to lipopolysaccharide biosynthesis pathway (Fig 5), it is worth noting that the levels of cell wall precursors such as UDP-N-acetyl-D-glucosamine or UDP-N-acetyl-D-galactosaminuronic acid show a significant two- and three-fold increase, respectively, during late infection (S4 Table). This increase is not accompanied by altered expression of host genes involved in this pathway (see below).
Most enriched pathways among decreasing metabolites involve amino acid biosynthesis (Fig 5), more specifically Arg, Pro, Ala, Asn, Glu, Cys and Met metabolism were found significantly enriched (p-value < 0.005). These observed decreases may indicate drainage of amino acid pools in the cell during phage particle formation, due to an imbalance between cellular amino acid biosynthesis and consumption by the phage.
We initially hypothesized that the observed changes in metabolome composition upon infection would largely be the result of a differential expression of host genes induced by the phage. This would indicate that PAK_P3 mainly interferes with cellular transcription to alter host physiological processes. To address this question, we investigated if the variations at the metabolome level could be directly linked to transcriptional changes. We thus analyzed all metabolites belonging to pathways highlighted by the pathway enrichment analysis (see above) that display significant variations (│Log2(fold change)│ > 0.5, p-value < 0.05) as well as differential expression of coding sequences related to the corresponding pathways with a stringent cut-off point (│Log2(fold change)│ > 1.3, p-value < 0.05) (Fig 6).
Only few genes linked to these pathways were significantly differentially expressed upon late infection, indicating that the phage influence on host metabolism is not primarily mediated through differential gene expression. In fact, several pathways with increased metabolite levels have a decreased transcription of the involved genes or vice versa (e.g. lipopolysaccharide biosynthesis).
Based on these complementary “-omics” approaches, it can be concluded that PAK_P3 does not otherwise redirect host physiology towards viral reproduction through modification of host gene expression. The general degradation of host RNA observed likely ensures sufficient building blocks for viral genome replication. The metabolic content of PAK_P3 infected cells shows both increased and decreased metabolite levels. We hypothesize these changes are either the direct consequence of an increased viral consumption of metabolites (e.g. amino acid metabolism) or are likely triggered by phage-encoded AMGs (e.g. pyrimidine metabolism).
Besides redirecting host cell physiology, the phage must also control its own gene expression. Here we intended to investigate the transcriptional strategy of PAK_P3 and also discovered unexpected regulatory mechanisms the phage uses to complete its infection cycle.
During the course of infection, early, middle and late transcripts of PAK_P3 genome were identified (Fig 7, S5 Table). The early transcribed region encompasses genes gp74 through gp112, all of which encode hypothetical proteins with low or no sequence similarity to gene products from other bacteriophages (so-called ‘ORFans’). Transcripts produced at middle time point focus on two regions that each contains gene features related to nucleic acid metabolism. As expected, the structural region appears to be mostly transcribed in late infection. Strikingly, five ORFs (i.e. gp34, gp37, gp38, gp45 and gp46), although located in the structural region, are overexpressed early compared to late time point. Finally, all predicted genes are transcribed, except for gp113, which corresponds to the predicted genome terminus. Intergenic transcription is observed throughout the genome, highlighting the great compaction of viral genomes where every single gene is expressed, in contrast to bacterial genomes. This property is further illustrated by the large amount of antisense transcripts detected, as reported below.
Analysis of antisense transcripts of PAK_P3 revealed 20 putative asRNAs, 8 of which are small asRNA (mean length 176±30 bp) and 12 are longer than 300 bp (S6 Table). All but one are encoded within genes, suggesting they may act as cis-encoded antisense RNAs. These asRNA are predominantly (15 out of 20) located in the structural region of PAK_P3 genome and are significantly more strongly transcribed during early infection compared to late infection with fold changes ranging between 2 and 17 (S2 and S3 Figs, S5 Table). These data support the hypothesis that such antisense transcription is used to shut down expression of late structural genes during the early stage of infection.
Following the observation of abundant antisense transcripts, we looked for other unusual transcriptional profiles within PAK_P3 transcriptome and detected two abundant small (~100bp) transcripts during late infection. These two transcripts, hereafter referred as sRNA1 and sRNA2, were found in two neighboring intergenic regions: sRNA1 is encoded within a 200bp-intergenic region between two genes encoding hypothetical proteins, whereas sRNA2 is part of a larger intergenic region between two phage-encoded tRNAs (Fig 7). They are temporally regulated since they display a 91- and 12-fold change, respectively, in their ‘late versus early’ expression. Strikingly, these two small RNAs belong to the most strongly transcribed regions of the phage genome during late infection as they respectively represent the 18th and 24th most expressed gene features over 86 late genes (S5 Table).
We hypothesized that they could be trans-encoded small RNAs, acting by base-pairing on a target mRNA. As such, we looked for potential target regions in both phage and host genomes. The potential targets found on the host genome were not differentially expressed 13 min post infection, indicating that these two phage small RNAs would not act through mRNA degradation but rather have a role in translational silencing, if any.
Interestingly, a stretch of 11 nucleotides on sRNA2 was found to be repeated eight times on the host genome and systematically located within tRNAs, more particularly within the TψC-loop. We propose that it could be involved in translational repression by binding, and eventually blocking, bacterial ribosomes. As this 11bp-stretch is also conserved in closely related phages (PAK_P1-like genus), it may represent a starting point leading to the discovery of new phage non-coding RNAs.
On the phage genome, the only potential targets (11 consecutive nucleotides matching perfectly) are located in the early ORFan product gp78 for sRNA1 and in the late gene encoding the putative ribonucleotide-diphosphate reductase gp67 for sRNA2.
To date, only few phage-encoded small RNAs have been described in the literature and most of them derive from prophages [19,20]. The only examples of phage sRNAs encoded by a virulent phage, T4 band C and band D RNAs, were described in the 1970’s and their functions have remained unknown ever since [21].
Next generation transcriptomics, metabolomics, and classical microbiological techniques have here been integrated for the first time to describe virus/host interactions between the candidate therapeutic bacteriophage PAK_P3 and its host, P. aeruginosa strain PAK. By capturing early, middle and late infection time points, we delineated genomic regions of temporally distinct phage expression. This particularly highlights early gene features, which are typically involved in the shutdown of host metabolism. Like the approximately 50 so called ‘monkey-wrench’ proteins found in phage T4, small early proteins likely have functions reliant on protein-protein interactions to disrupt host systems and could potentially be exploited to aid in small molecule antibiotic design [22–24].
It is well established for model bacteriophages, including T7 and T4, that the temporal regulation of middle and late gene expression is typically the result of a tight regulation driven by phage early proteins through various mechanisms such as redirection of host RNA polymerase to phage middle and late promoters (like phage T4 proteins AsiA-MotA or phage-encoded sigma factor gp28 in SPO1) [16,25]. The early expression of antisense RNAs could represent an additional regulation mechanism preventing transcriptional leaks from strong promoters controlling expression of late structural genes. Consistent with this hypothesis, the temporal distribution and the location of the numerous PAK_P3 asRNAs correlate with the shut-off of structural gene expression observed 3.5 min post infection. Although cis-antisense RNAs appear to be a common form of regulation in bacterial genomes, they have not been extensively described in phage genomes. Beside the regulatory oop RNA reported over 40 years ago (reviewed in [26]), no other asRNAs were reported until recently [27] and exclusively in lambdoid phages. Moreover, such asRNAs have never been reported for virulent phages until 2014 [28]. Therefore, the high number asRNAs reported for PAK_P3 implies that antisense transcription may be a regulatory mechanism used by phages more frequently than previously thought.
From a phage-host interaction point of view, we found that existing host transcripts are rapidly overwhelmed with viral transcription. This may reflect a globally accelerated degradation of RNA in the cell in a way similar to phage T4. Indeed, it has been previously reported that T4 globally alters the stability of existing mRNAs, in addition to repressing the transcription of cytosine containing DNA [29,30]. This hypothesis is further supported by the drastic overexpression of one host operon (PAK_4493–4499) encoding RNA processing-related proteins. In particular, this operon encodes a RNA 3'-phosphate cyclase RtcA (PAK_4496) that has been described as being involved in the processing of RNA transcripts such as priming RNA strands for adenylylation to protect them from exonucleases or to mark them for further processing so they serve as substrates for downstream reactions performed by additional enzymes [31,32]. Therefore, we hypothesize that this operon may be uniquely upregulated by the phage in order to participate in the global degradation of RNAs during infection, which we observe in both the RNA-Seq and metabolomics data, by tagging transcripts for degradation by phage encoded enzymes. An alternative hypothesis to explain this dramatic up-regulation of PAK_4493–4499 relies on RtcB (PAK_4494), a predicted RNA ligase. Together, the RtcAB system has been shown to play a role in tRNA repair after stress-induced RNA damage (e.g. viral infection) in E. coli [33]. Also, it has been shown that phage T4 RNA 3'-phosphate cyclase (encoded by pseT) and RNA ligase (rli) are involved in overcoming resistance [34] by restrictive strains of E. coli producing phage induced tRNA anticodon nuclease (encoded by the prr locus) which causes abortive infection by preventing effective translation of phage genes [35]. Therefore, it is possible that PAK_P3 upregulates this operon to activate a host repair function to interfere with a yet uncharacterized host restriction system or a prr-like locus. However, deleting the RNA ligase rtcB gene (PAK_4494), appears to have no toxic effect on the host as well as no effect on the efficiency of plating (S1 Text).
The observation of a phage induced host RNA degradation is further supported by our metabolomics data. Indeed, the increased pyrimidine metabolism confirms that nucleotide turnover is a central viral need to achieve a successful infection cycle. Overall, we showed that upon PAK_P3 infection, the host metabolism is not shutdown but redirected to generate the required building blocks for viral replication and this redirection is not the result of a phage induced differential host gene expression, aside from the RNA processing operon. An explanation for this metabolic turnover relies on phage-encoded AMGs and phage early proteins. We propose that phage early proteins would interfere with host metabolic processes through interactions with bacterial proteins. Once the host machinery is disrupted, phage metabolic enzymes would take over and catalyze the reactions yielding the specific metabolites required for viral replication (Fig 8). For instance, we hypothesize that the observed global degradation of host mRNA eventually produces an excess of free ribonucleotides that are likely converted into deoxynucleotides by ribonucleotidases. Interestingly, PAK_P3 encodes a putative ribonucleotide-diphosphate reductase (alpha and beta subunits, respectively gp67 and gp69) that could catalyze such a reaction. An alternative explanation for the observed increase of (deoxy)nucleotides-phosphates relies on the putative deoxyribonuclease (gp57) encoded by PAK_P3, which could be responsible for host genome degradation during middle and late infection stages, as observed for phage LUZ19 [36]. Supporting these hypotheses, these three phage-encoded AMGs are strongly expressed by PAK_P3 during late infection stage as they are respectively the 7th, 22nd and 19th most expressed genes over 86 late genes (S5 Table). It is noteworthy that a fourth predicted phage-encoded AMG, a CMP deaminase (gp155), is also involved in nucleotide metabolism and also expressed during late infection although less intensely than the other AMGs mentioned. Altogether, these predicted AMGs involved in nucleotide metabolism highlight the central need for nucleotides during PAK_P3 infection, in accordance with the short infection cycle span during which about 50 genomes of 88 kb have to be synthesized. Indeed, the advantage found in precisely manipulating nucleotide depolymerization and pathways to shut down host mechanisms, provide material for phage DNA synthesis, and prevent osmotic stress appears to be significant across phage clades. For example, Pseudomonas phage Lu11 contains ORFs predicted to be involved in nucleotide metabolism [37], E. coli phage T5 degrades host DNA before exporting it outside of the cell [38], and T4 even encodes for its own, nearly complete, parallel DNA precursor biosynthesis pathway [6].
Another example of phage-driven interference with host metabolic pathway is given by LPS biosynthesis pathway. The observed accumulation of cell wall precursors, which is not correlated with an altered expression of corresponding host genes, may be a direct consequence of peptidoglycan degradation and the subsequent release of its precursors triggered by the infection. Consistent with this hypothesis, PAK_P3 has a potential AMG (gp151) similar to a bacterial cell wall hydrolase, which could explain such cell-wall degradation.
Further investigations are now required to fully associate transcriptomics and metabolomics data to viral gene functions, a process which is currently hampered by the lack of versatile genetic tools to construct mutants of virulent bacteriophage. Such effort to deeply characterize one particular phage genus (i.e. KPP10-like) is also motivated by the great therapeutic potential of these phages as demonstrated in animal models and recently strengthened by the identification of such a phage in the ‘Intesti phage’ cocktail, a key commercial product of the Eliava Institute in Georgia [4,39–41]. Overall, the knowledge of phage biology provided by next-generation “-omics” approaches not only enlighten viral mechanisms of infection but can also open an array of biotechnological applications based on regulatory elements and proteins found in this new sub-family of phages.
P. aeruginosa strain PAK [42] was cultured in LB medium supplemented with 10mM MgCl2 at 37°C unless stated otherwise. For RNA-Seq experiments, cells were infected with bacteriophage PAK_P3 using a multiplicity of infection of 25 in order to ensure the synchronicity of the infection (95% of the bacterial population killed after 5 min phage-bacteria incubation).
Bacteriophage growth parameters were assessed as described previously [43]. Briefly, a culture of strain PAK was infected at low MOI (0.1) and incubated 5 min at 37°C with agitation allowing bacteriophage particles to adsorb. Following a 1000-fold dilution, two 100 μL samples were collected every 2 min and either kept on ice until titration, or mixed with CHCl3. For each time point we thus determined the free bacteriophage count (samples with CHCl3) as well as the number of free bacteriophages and infective centres (samples without CHCl3) to calculate eclipse and latency periods respectively.
Experimental data were fitted with a logistical function:
f(x)=a1+e−k(x−xc)
(1)
a: ordinate corresponding to the asymptote when x→+∞, represents the maximal pfu count.
xc: abscissa of the inflection point, represents the mean duration of the infectious cycle.
k: slope of the tangent line to the exponential part of the curve.
Eclipse and latency periods were determined as the x value corresponding to f(x)>0.05a
The burst size was determined as:
Phage_titer(t=0,-CHCl3)-Phage_titer(t=0,+CHCl3)Phage_titer(t=30,±CHCl3)
(2)
Phage_titer(t = 0, +CHCl3) and Phage_titer(t = 0, -CHCl3): values of initial phage titers (t = 0 min) measured in samples treated or not with chloroform, respectively. The numerator represents the number of intracellular phages.
Phage_titer(t = 30, ±CHCl3): Mean of phage titers measured in samples treated and not treated with chloroform at t = 30 min.
Four independent adsorption assays were performed in the conditions described above with a lower MOI (10−3) and omitting the dilution step. Data could be approximated using an exponential function and adsorption time was defined as the time required to reach a threshold of 10% non-adsorbed bacteriophage particles.
Genomic DNA was isolated from P. aeruginosa strain PAK and pyrosequencing was performed on a Roche 454 FLX system with Titanium chemistry at the University of Texas Genomic Sequencing and Analysis Facility. The draft assembly of ~6.3 Mbp consists of 9 scaffolds, 490 large contigs, and 616 total contigs and was annotated at the University of Maryland Institute for Genomic Sciences using the IGS Prokaryotic Annotation Pipeline[44]. Scaffolds deposited in GenBank can be accessed via Bioproject accession no. PRJNA232360. More details are available in S1 Text.
RNA-Seq analysis was performed on an exponentially growing culture that was synchronously infected with PAK_P3. Three independent biological replicates were harvested at 0min, 3.5min and 13min to represent, respectively, a phage negative control and early and late transcription while one additional sample was collected at 6.5 minutes to assess the presence of an identifiable middle phase of transcription.
The preparation of cDNA libraries was performed as described in Blasdel et al. (in press) [45]. Briefly, samples were collected at three time points, representing early, middle, and late infection, from a synchronously infected culture, with <5% of bacteria remaining uninfected after 3.5 minutes, and halted by rapid cooling in 1/10 volume of ‘stop solution’ (10% phenol, 90% ethanol). Cells were then lyzed in TRIzol, total RNA was purified through a standard organic extraction and ethanol precipitation, and remaining genomic DNA was removed using TURBO DNase. DNA removal was confirmed with PCR before rRNA was depleted using the Ribo-Zero rRNA Removal Kit (Gram-Negative Bacteria). This rRNA depleted total RNA was then processed into cDNA libraries using Illumina’s TruSeq Stranded Total RNA Sample Prep Kit according to manufacturer’s instructions and sequenced using an Illumina NextSeq 500 desktop sequencer on the High 75 cycle. More than 11 million 75bp reads mapping to non-ribosomal regions were obtained from each library with the exception of one early sample and one late sample providing 1,221,867 and 941,631 mapped reads respectively due to incomplete rRNA removal. After trimming, sequencing reads were aligned separately to both the phage and host genomes with the CLC Genomics workbench v7.5.1. These alignments were then summarized into count tables of Unique Gene Reads that map to phage or host gene features respectively.
RNA-Seq data have been deposited in NCBI-GEO with accession no. GSE76513.
RNA-Seq coverage visualization is available through the COV2HTML software at [https://mmonot.eu/COV2HTML/visualisation.php?str_id=-32] for a comparison of the host (0 min / 13 min) and [https://mmonot.eu/COV2HTML/visualisation.php?str_id=-34] for a comparison of the phage (3.5 min/ 13 min)
RNA-Seq data of uninfected strain PAK were visualized using COV2HTML [7]. Reads mapping forward and reverse strands were manually scanned over the whole genome. Both coding regions and intergenic regions displaying an unexpected transcription profile were examined using CLC Genomics Workbench 7.5.1 and Blastp (default parameters) to annotate putative new coding sequences or RNA central (http://rnacentral.org/sequence-search/), Rfam search (http://rfam.xfam.org/) and RNAfold web server (http://rna.tbi.univie.ac.at/cgi-bin/RNAfold.cgi) with default parameters to predict putative small RNAs and riboswitches.
Each statistical comparison presented was performed using the DESeq2 [46] R/Bioconductor package to normalize samples to each other and then test for differential expression. Notably, we have chosen to normalize the population of reads that map to each genome independently of the other. In the context of a phage infected cell rapidly replacing host transcripts with phage transcripts, this has artificially enriched host reads and depleted phage reads progressively over the course of infection by normalizing away the biologically relevant shift in each organism’s proportion of the total reads in the cell. However, this has also allowed us to both show and test for differential expression of both phage and host gene features independently of the more global swing towards phage transcription.
P. aeruginosa strain PAK cells, grown in minimal medium (30 mM Na2HPO4, 14 mM KH2PO4, 20 mM (NH4)2SO4, 20 mM glucose, 1 mM MgSO4, 4 μM FeSO4), were infected with PAK_P3 at OD600 = 0.3 (approx. 1.25.108 CFU). At 0, 5, 10, 15, 20, and 25 minutes post infection, cells were collected by fast filtration [47]. The biomass quantity was adjusted to match the biomass of a 1 mL culture at OD600 = 1.0 (approx. 4.108 CFU) by following the OD600 and adjustment of the sampling volume. Four biological replicates were sampled and two technical repeats were made of each independent biological sample. The metabolic content was extracted as described by De Smet et al [3]. The samples were profiled using only negative mode flow injection-time-of-flight mass spectrometry and detected ions were annotated as previously reported [17]. Metabolite annotation and statistical analysis was performed using Matlab R2013b (Mathworks, Natick, MA, United States) according to the ion annotation protocol described by Fuhrer et al. [17]. With this method, 6006 ions were detected and 918 of them could be assigned to known P. aeruginosa metabolites. After removal of ion adducts, 377 ions were retained that were annotated as 518 metabolites (including mass isomers).
Differential analysis was performed for each time point versus time point zero using a t-test for two samples with unequal variances (Welch test). For metabolic pathway enrichment, lists of significantly changing metabolites for each time point were created based on the thresholds of │Log2(fold change)│ ≥ 0.5 and adjusted p-value < 0.1. In each list, metabolites were sorted by the adjusted p-value, and the pathway enrichment procedure was performed for each subset of size 1 to the size of the significant list using Fischer test as described in [48] and the smallest p-value for each pathway was reported. For differential analysis and pathway enrichment, p-values were adjusted for multiple hypotheses testing with the Benjamini-Hochberg procedure.
|
10.1371/journal.pgen.1004450 | A Loss of Function Screen of Identified Genome-Wide Association Study Loci Reveals New Genes Controlling Hematopoiesis | The formation of mature cells by blood stem cells is very well understood at the cellular level and we know many of the key transcription factors that control fate decisions. However, many upstream signalling and downstream effector processes are only partially understood. Genome wide association studies (GWAS) have been particularly useful in providing new directions to dissect these pathways. A GWAS meta-analysis identified 68 genetic loci controlling platelet size and number. Only a quarter of those genes, however, are known regulators of hematopoiesis. To determine function of the remaining genes we performed a medium-throughput genetic screen in zebrafish using antisense morpholino oligonucleotides (MOs) to knock down protein expression, followed by histological analysis of selected genes using a wide panel of different hematopoietic markers. The information generated by the initial knockdown was used to profile phenotypes and to position candidate genes hierarchically in hematopoiesis. Further analysis of brd3a revealed its essential role in differentiation but not maintenance and survival of thrombocytes. Using the from-GWAS-to-function strategy we have not only identified a series of genes that represent novel regulators of thrombopoiesis and hematopoiesis, but this work also represents, to our knowledge, the first example of a functional genetic screening strategy that is a critical step toward obtaining biologically relevant functional data from GWA study for blood cell traits.
| In this manuscript we report on a follow-up study of the GWAS loci associated with the platelet size and number. A GWAS meta-analysis identified 68 genetic loci controlling platelet size and number. Only a quarter of those genes, however, are known regulators of hematopoiesis. To determine function of the remaining genes we performed a medium-throughput genetic screen in zebrafish using morpholinos (MOs) to knock down selected candidate genes. Here, we report on two major findings. First we identified 15 genes (corresponding to 12 human genes) required for distinct stages of specification or differentiation of HSCs in zebrafish. A detailed review of databases and literature revealed limited knowledge about the functional role of Satb1, Rcor1 and Brd3 in hematopoiesis and for the remaining nine genes our work represents the first study on their putative role in hematopoiesis. And secondly, we demonstrate that brd3a is critical for establishing, but not maintaining thrombopoietic compartment. Importantly, our study introduces zebrafish as a model system for functional follow-up of GWAS loci and generates a valuable resource for prioritization of platelet size and number associated genes for future in-depth mechanistic analyses. Following this route of investigation new regulatory molecules of hematopoiesis will be added to critical pathways.
| Erythrocytes and platelets (thrombocytes in zebrafish) are the most abundant cells in blood. In an individual, the number and volume of both erythrocytes and platelets are highly heritable and tightly regulated within narrow ranges, but there is a wide variation of these parameters in the population [1], [2]. The 80% heritability of blood cell indices provided the foundation for our recently completed GWAS meta-analysis in ∼68,000 healthy individuals for both cell types. We identified 68 genetic loci that control the mass (volume x count) of platelets [3] and another 75 for red cell indices [4]. About a quarter of the genes proximal to the platelet GWAS association single nucleotide polymorphisms (SNPs) encode well studied and generally pivotal regulators of hematopoiesis but the function of the remaining ones is unknown, demonstrating the power of GWAS to identify novel regulators of hematopoiesis. We have recently reported functional validation of six genes in which the sentinel SNP was localized within a gene and silenced ak3, rnf145, arhgef3, tpm1, jmjd1c and ehd3 in zebrafish by MO injections. Profound effects on thrombopoiesis were observed for all but ehd3 [3]. Furthermore, our detailed studies of the arhgef3 gene, which encodes one of the ∼70 Rho guanine nucleotide exchange factors, showed its important role in iron uptake and transferrin receptor internalisation in erythrocytes [5]. Based on these preliminary data, we hypothesized that the majority of genes identified in our recent genomics efforts are important and rate-limiting regulators of hematopoiesis and therefore worthwhile of further investigation.
The zebrafish model has distinct advantages over other animal models for screening large numbers of genes. Zebrafish development occurs rapidly over the course of a few days with thrombocytes, erythroid- and myeloid- blood cells being fully formed and functional by 3 days post fertilisation (dpf). External fertilisation and transparency of zebrafish embryos allow easy visualisation of early blood-related phenotypes giving them the advantage over mice, where development occurs in utero. Importantly, transcriptional mechanisms and signalling pathways in hematopoiesis are well conserved between zebrafish and mammals [6].
Herein, we performed a MO injection screen of 15 genes identified in GWAS for platelet size and number to uncover novel pathways essential in thrombopoiesis and hematopoiesis in general. From this screen, we identified 12 new genes required for normal hematopoiesis and ordered them on a hematopoietic lineage tree based on their presumed function during hematopoiesis. Further analysis of the hematopoietic lineage tree revealed a distinct pattern of gene distribution suggesting two main gene clusters. One cluster of genes appears to work at the level of HSCs affecting all derived blood cell types and the second cluster appears to be limited to controlling the specification of the thrombocyte-erythroid progenitors. Additionally, we show that one of the novel candidate genes, brd3a, is essential in differentiation of thrombocytes from HSCs but is dispensable for their maintenance.
To interrogate the large number of novel hematopoietic genes identified in the GWAS for platelet size and number we developed an in vivo functional genomics screen in zebrafish (Figure S1). The first step was selection of the most suitable candidate genes and initially we selected a single gene, closest to the sentinel SNP, from each GWAS locus [3] (Table S1). Distance from the nearest gene was calculated as the absolute distance between SNP and transcription start site of the gene or 3′ end of last exon [3] (Table S1). We further eliminated genes with a known function in hematopoiesis and identified a putative zebrafish ortholog, with over 38% identity, on the protein level, with its human counterpart for 40 genes. Finally, we excluded all genes that would require the use of more than two MOs, resulting in a list of 33 genes. Nearly 80% of these genes had a sentinel SNPs localized within 10 kb. In the last step, we selected 19 genes of which 16 had a sentinel SNP within 10 kb and three genes had a sentinel SNP >10 kb. Five of the selected genes were duplicated in zebrafish, resulting in a total of 24 genes to be further investigated aiming to define their function in blood cell formation by a MO knockdown approach in zebrafish.
We designed splice-blocking MOs for each gene and validated their efficacy with RT-PCR and sequencing. Of the 24 MOs tested five MOs had no effect on the target RNA and these MOs were excluded from further analysis (Figure S2). We then assayed the remaining 19 functional MOs for their effect on overall development, morphology and hematopoiesis during the first 72 hours post fertilisation (hpf) and selected the optimal dose of MO to be injected (Figure S3). Of note, based on the information available at the ZFIN at the time and our in situ hybridization data (Figure S4) none of the selected genes had hematopoietic specific gene expression. For all genes, except grtp1a, the optimal dose of MO was selected that resulted in a specific phenotype but without gross lethality or defects in body shape or size, vasculature, heart and circulation. We found that grtp1a MO injected embryos died by 15 hpf even when injected with 0.8 ng of MO, thus we excluded grtp1a from further analysis.
Although morphological examination can detect defects with great sensitivity, some specific defects in hematopoiesis can be missed, and as a result information obtained from the initial analysis might be limited. Thus, we carried out a second level of analysis by performing in situ hybridisation with several hematopoietic markers, specifically, c-myb, ae1 globin, mpeg and rag1. These were complemented with the use of the Tg(cd41:EGFP) line and two histochemical stains, o-Dianisidine and Sudan black. The hematopoietic markers used for phenotyping were carefully chosen to distinguish between early and late stages of hematopoiesis as well as thrombocytes, erythrocytes, neutrophils, macrophages and lymphocytes (Figures S5).
We initiated screening by taking advantage of the Tg(cd41:EGFP) reporter line, which labels thrombocytes, to identify genes in zebrafish that, when disrupted, affect thrombocyte number. We found that knock down of 15 of the 18 genes resulted in 30–95% reduction in thrombocyte number (Figure 1, Figure S6). Importantly, where the functional second MO was available, the observed phenotype was comparable to the one observed with the first MO (Figure S7, S8, Table S2). Furthermore, concurrent knock down of p53 with gene specific MOs did not attenuate the thrombocyte phenotype induced by gene specific MOs, confirming that the observed decrease in the number of thrombocytes was not induced by p53 mediated off target effects of MO injection (Figure S9). For one of the “phenotypic” genes, brf1b, we have obtained the mutant through ZMP. To test if brf1b mutants have a decreased number of thrombocytes, the offspring were subjected to the “clotting time assay” at 5 dpf [7]. Clotting time in brf1b mutant larvae was significantly longer than in the wild type fish, suggesting a defect in thrombopoiesis (Figure S10 A). To further confirm the specificity of the phenotype observed in MO injected embryos and in the absence of available mutants we performed the rescue of rcor1 MO injected embryos using full-length zebrafish rcor1 RNA (Figure S10 B, C). We focused our further analysis on the 15 “phenotypic” genes and their paralogs.
To maintain the population of differentiated blood cells within normal ranges, HSCs need to continuously maintain the balance between self-renewal and differentiation. We reasoned that decreased number of thrombocytes in MO injected embryos could be the result of either reduced numbers of HSCs or altered HSC differentiation. To assess which stage of hematopoiesis was affected in each MO injected embryo, we performed in situ hybridization at 3 dpf and looked for alterations in expression of definitive hematopoiesis marker c-myb (Figure 2). Although more than half of the MOs tested had no effect on the number of HSCs, depletion of rcor1 resulted in an increased number of HSCs and depletion of kalrn (1 and 2), mfn2, pdia5, psmd13 and wasplb resulted in decreased numbers of HSCs in caudal hematopoietic tissue (CHT) at 3 dpf. This reduction in the number of HSCs was not evident at 30 hpf in kalrn1, mfn2, pdia5, psmd13 and wasplb MO injected embryos (Figure S11) suggesting that the number of HSCs at 3 dpf was most probably adversely affected by either their homing to or survival/proliferation in CHT. However, kalrn2 and rcor1 depleted embryos had a marked decrease in the number of HSCs at 30 hpf, implying the important role of these genes in specification of HSCs in the aorta-gonad-mesonephros (AGM) (Figure S11). Importantly, analysis of vascular development by injecting candidate gene MOs into Tg(fli1:EGFP) embryos, which express EGFP in endothelial cells, demonstrated no major abnormalities in vascular morphogenesis or remodeling that would preclude circulation, indicating that the hematopoietic defects were not secondary to a vascular phenotype (Figure S12). Thus, our screen effectively defined a set of nine genes required for differentiation of HSCs to thrombocytes and possibly other blood lineages.
Hematopoiesis is often depicted by a hierarchical differentiation tree, with HSCs at the root and the mature blood cells as the branches. One of the intermediate cellular states is the common myeloid progenitor (CMP), which can proliferate and differentiate into megakaryocyte-erythrocyte progenitors (MEP) and granulocyte-monocyte (GM) progenitors, which further give rise to megakaryocytes, erythrocytes, granulocytes, monocytes and other cell types. To investigate lineage-specific effects of the candidate genes, we assessed the status of definitive erythropoiesis in MO injected embryos at 4 dpf. As αe1-globin RNA was reported to be expressed in definitive erythrocytes at 4 dpf [8], a detailed analysis of the expression pattern of αe1-globin transcript was exploited to reveal the initiation of definitive erythropoiesis after gene silencing. Profound effects on definitive erythropoiesis were observed for all but brd3a, brf1b, waspla and wdr66 (Figure S13). However, silencing of brf1a and wasplb resulted in diminished definitive erythropoiesis, reflecting functional divergence of duplicated genes (brf1 and waspl) (Figure S13). Furthermore, an extensive analysis of hemoglobin levels in primitive erythroid cells at 2 dpf showed that, with the exception of brd3a, kalrn2 and kif1b, primitive erythropoiesis was largely unaffected following MO knock down of candidate genes (Figure S14). These results are consistent with the notion that the majority of candidate genes are dispensable for specification and differentiation of primitive erythrocytes and that fundamentally different molecular mechanisms regulate primitive and definitive erythropoiesis.
To establish the role of candidate genes in differentiation of the myeloid lineage, that is neutrophils and macrophages, we performed Sudan Black staining (for neutrophils) and in situ hybridization using mpeg riboprobe (for macrophages) in control and candidate gene depleted zebrafish embryos at 3 dpf. Out of 15 genes tested, depletion of nine resulted in reduced numbers of Sudan Black positive cells and two (kif1b, waspla) had an effect on the number of macrophages (Figures S15, S16, and S18). Reduction in the number of Sudan Black positive cells could reflect the absence of granules rather than neutrophils. We have, therefore, performed in situ hybridization using mpx riboprobe for all genes for which the knockdown resulted in a decrease in the number of Sudan Black positive cells. For all the tested genes the observed phenotype was comparable to the one we reported following Sudan Black staining (Figure S17).
Finally, we analyzed the impact of loss of candidate gene function on lymphoid development. Differentiated thymic T-cells are exclusively derived from definitive HSCs and can be readily identified by rag1 expression when examined at 4 dpf (Figure S19). Not surprisingly, a significant decrease in rag1 staining was mostly observed in the same set of genes in which we observed a decrease in c-myb staining (a marker for HSCs), namely kalrn1, kalrn2, mfn2, pdia5 and psmd13. In addition, injection of akap10 MO and rcor1 MO, which had no negative impact on the number of c-myb positive cells, resulted in a significant decrease in the number of T lymphocytes.
The large number of genes analyzed and the resultant volume of data acquired can present challenge in understanding and interpreting the results. Hence, we used the information gained from the initial MO knockdown screen to generate a heat-map of phenotype profiles (Figure 3) and cluster genes with similar phenotypic profiles. We then hierarchically positioned candidate genes on the hematopoietic lineage tree and assigned each of them a potential role during hematopoietic differentiation (Figure 4).
Our analysis of the hematopoietic lineage tree revealed a distinct pattern of gene distribution, suggesting two main gene clusters. The first cluster represents a set of genes, namely kalrn1 and -2, mfn2, pdia5, psmd13, rcor1 and wasplb, which upon depletion affect the number of HSCs. The second major cluster represents a set of genes, namely akap10, brf1a, kif1b, satb1 and wasplb, which appear to be essential further down the hematopoietic tree and affect differentiation of both definitive erythrocytes and thrombocytes. These genes have presumed role in HSC fate decisions prior to specification of the thrombocyte and erythrocyte progenitors.
Although the frequency of blood defects observed in our screen was high, the screen was not as well suited for the identification of knockdown phenotypes that result in subtle differences in myeloid lineage cell production or skewing of myeloid lineage differentiation. This is mainly because changes in the number of neutrophils and macrophages arising from HSCs may be undetectable using markers and the developmental time point outlined here. Previous studies have shown that erythroid-myeloid progenitor cells (EMPs) are capable of generating both macrophages and neutrophils and that these blood cells appear in mib zebrafish despite the absence of HSCs [9]. However, even with these caveats, we believe the screening procedure used has proven effective for extracting functional information from a GWAS dataset in a medium-throughput manner.
To gain an additional insight into mechanisms by which these newly discovered genes affect thrombopoiesis, we performed a more extensive analysis of the function of brd3a in hematopoiesis. The bromodomain and extra terminal domain (BET) family of proteins, including BRD2, BRD3, and BRD4, are evolutionally conserved and play a key role in many cellular processes by controlling the assembly of histone acetylation-dependent chromatin complexes [10]. To further confirm that the defects observed in the brd3a depleted embryos resulted from loss of brd3a, in vitro–transcribed RNA encoding human BRD3 (hBRD3) was injected into 1-cell stage embryos. Live confocal imaging of zebrafish embryos injected with hBRD3-GFP confirmed that hBRD3 binds to mitotic chromosomes (Figure 5A), a feature previously reported for BET family proteins i.e. BRD2, BRD3 and BRD4 [11]–[13]. Expression of hBRD3 in brd3a MO injected embryos resulted in partial but statistically significant rescue of the number of thrombocytes demonstrating that brd3a MO used in this study exerted a specific effect (Figure 5B, C).
Morpholinos do not allow gene-specific perturbation to be carried out with temporal resolution, which is a disadvantage when dissecting the precise role of a selected gene in hematopoiesis. A number of studies reported that compounds targeting BET proteins might be used to manipulate hematopoietic development for exploratory or therapeutic purposes [14]–[16]. The BET family inhibitor, thieno-triazolo-1,4-diazepine ((+)-JQ1 in short) is a potent, highly specific inhibitor which displaces BET proteins from chromatin by competitively binding to the acetyl-lysine recognition pocket of BET bromodomains [17], [18]. Thus, we evaluated the pharmacological impact of (+)-JQ1 on zebrafish development and thrombopoiesis. Exposure of zebrafish embryos to (+)-JQ1 disrupted the chromatin occupancy of hBRD3-GFP confirming the efficacy of the inhibitor (Figure S20 A–C).
We next incubated embryos from 6 hpf in various concentrations of (+)-JQ1 and (−)-JQ1 (stereoisomer which has no appreciable affinity to BET bromodomains) [17] as a control (Figure S21). Exposure of zebrafish embryos to 1 µg/ml (+)-JQ1 resulted in complete mortality by 24 hpf whereas the (−)-JQ1 enantiomer showed no observable effect on embryo development (Figure S21). This early embryonic death of zebrafish embryos was not surprising considering that knockout of Brd2 or Brd4 in mice results in embryonic lethality, indicating an important role of these two proteins in embryonic development [19], [20]. When treated, however, with (+)-JQ1 from 24 hpf, embryos exhibited overall normal development even at the higher concentration (1 µg/ml) of (+)-JQ1 (Figure S21). Although morphologically normal, thrombopoiesis was completely abolished in these embryos (Figure 6 A). Interestingly, the decrease in the number of thrombocytes appeared more prominent in the presence of (+)-JQ1 inhibitor compared to brd3a MO knock down. This opened the possibility that other members of the BET family might be contributing to the observed phenotype. To investigate this further, we performed MO knock down of zebrafish brd2a, brd2b and brd4 and assessed the number of thrombocytes at 3 dpf. Single MO knock down of all three genes resulted in a severe decrease in the number of thrombocytes. These data strongly suggested that, indeed, other members of BET family of proteins (i.e. brd2 and brd4) play an important role in thrombopoiesis (Figure S22 A–C).
Both MO knock down and treatment with (+)-JQ1 inhibitor from 24 hpf resulted in a severe reduction in the number of thrombocytes at 3 dpf, suggesting an essential role for brd3a in the differentiation of thrombocytes as opposed to a requirement for their maintenance and survival. To address this question we incubated Tg(cd41:EGFP) embryos with (+)-JQ1 inhibitor starting from 3 dpf, when there is already a considerable number of thrombocytes in CHT, and assessed their number 24 hours later, at 4 dpf. We found that in untreated and (−)-JQ1 treated embryos the number of thrombocytes markedly increased between 3 and 4 dpf. However, in (+)-JQ1 treated embryos we observed no change in the number of thrombocytes (Figure 6C). Moreover we found that (+)-JQ1 had no adverse effect on the number of HSCs during this 24 h period of treatment (Figure 6D). Taken together this strongly suggests that brd3a is important in differentiation of thrombocytes from HSCs, however, once differentiated, brd3a was dispensable for their maintenance and survival.
GWAS meta-analysis of platelet size and number has been successful in identifying SNPs associated with the mass (volume x count) of platelets. In contrast with the results of GWAS in common diseases, more than 80% of SNPs associated with hematological traits are localized within 10 Kb of genes providing a sound argument to infer biologically relevant candidate genes [3], [4]. Canonical pathway analyses detected a highly significant over-representation of “core genes” (the sentinel SNP is within the gene or within 10 kb from the gene) in relevant biological functions such as hematological disease, cancer and cell cycle [3]. However, three quarters of regions proximal to the platelet GWAS SNPs harbor unfamiliar genes or known genes not previously implicated in hematopoiesis that merit extensive follow-up analysis. Therefore, this study was set up to address the need for a medium-throughput method in zebrafish to dissect the functional roles of these assumed novel regulators of hematopoiesis.
In total, our screen identified 15 genes (corresponding to 12 human genes) required for distinct stages of specification or differentiation of HSCs in zebrafish. A detailed review of the content of databases and literature revealed limited knowledge about the functional role of Satb1, Rcor1 and Brd3 [21]–[24] in hematopoiesis and for the remaining nine genes our work represents the first study on their putative role in hematopoiesis. Importantly, our results are well in line with some of the findings reported by others. One example is RCOR1 - lineage-restricted deployment of RCOR1 and LSD1 cofactors, through interaction with Gfi proteins, controls hematopoietic differentiation [23]. Knock down of rcor1 in zebrafish resulted in completely blocked differentiation of erythroid, thrombocytic, myeloid and lymphoid lineages. These findings strongly support the hypothesis that the published platelet GWAS [3] enriched for functional regulators of the hematopoiesis and further support previous assumptions that a large proportion of the genes uncovered by the aforementioned GWAS also have a conserved role in zebrafish.
In this study, we followed a two step screening approach: in the first instance, we used the Tg(cd41:EGFP) line in conjunction with a panel of hematopoietic in situ hybridization probes and histochemical staining to create a heat map with distinct “phenotype signatures” of each gene knock-down. We then positioned the candidate genes on the hematopoietic cell lineage tree and assigned them a potential role in hematopoietic differentiation. Interestingly, our screen revealed that, although initially selected based on their effect on the platelet size and/or number, none of the candidates exerted a thrombocyte specific effect. These results should be interpreted within the context of several major differences between the effect of the SNPs and whole embryo MO knock down on hematological traits.
First, the majority of associated SNPs identified in platelet GWAS are in non-transcribed regions and it is likely that the underlying mechanism linking them to the phenotype is regulatory. Thus, the functional effects of SNPs are subtler compared to the knock down of transcripts achieved by MOs in our screening. Secondly, although GWAS provided a list of SNPs associated with the platelet size and number, there is no evidence about the biological processes that link the associated SNP to the phenotype. Indeed, it has been shown that in most cases the reported SNP is not the functional SNP itself but is in linkage disequilibrium with the SNP overlapping a functional region [25]. Experimental evidence shows that open chromatin profiles of megakaryocytes and erythroblasts differ and thus cell type-restricted regions of open chromatin could influence the penetrability of the functional SNP [26], [27] in a lineage specific manner. In contrast, MO knock down in zebrafish is not spatially restricted and thus offers the opportunity to determine the functional role of candidate genes in all blood lineages.
To further verify the hematopoietic role of genes identified in GWAS, we performed a more extensive evaluation of the effect of brd3a on thrombopoiesis. It has been shown that BRD3 interacts with acetylated GATA1 and stabilizes its chromatin occupancy [21]. A pharmacologic compound, JQ1, that occupies the acetyl-lysine binding pockets of Brd3 bromodomains disrupts the BRD3-GATA1 interaction, diminishes the chromatin occupancy of both proteins, and inhibits erythroid maturation [21]. Although GATA1 and BRD3 co-occupancy on GATA1 target genes was also observed in a megakaryocytic cell line [21], the biological relevance of this binding was never confirmed. Here we report on the important role of brd3a in thrombopoiesis. Indeed, knock-down of brd3a with two independent MOs as well as treatment of zebrafish embryos with the JQ1 inhibitor starting from 24 hpf severely reduced the number of thrombocytes in 3 days old embryos. Interestingly, incubation of embryos with JQ1 inhibitor between 3- and 4 dpf, that is after the onset of thrombopoiesis, did not have any effect on the already differentiated thrombocytes. However, the number of thrombocytes failed to increase compared to control embryos within this 24-hour period. These results strongly support the idea that brd3a is critical for establishing but not maintaining thrombopoietic compartment. Some previous studies suggest that BRD3, as well as some other mitotically retained factors, functions as a molecular “bookmark” by enabling post-mitotic transcription re-initiation of its target genes [13], [28]. It is plausible to assume that a similar mechanism is employed during thrombopoiesis. In that scenario, retention of Brd3 on chromatin during mitosis of thrombocyte precursors or erythroid-thrombocyte progenitor cells would contribute to the maintenance of transcription patterns necessary for establishment of thrombocyte identity. However, further work will be necessary to identify the precise molecular mechanisms by which brd3a exerts its effect on thrombopoiesis. Taken together, our study provides a paradigm of the usefulness of zebrafish for efficient translation of GWAS findings into relevant biological information in an objective and unbiased manner. GWAS have mapped many novel, convincingly associated loci in the proximity of genes where functional significance is expected. So far, functional validation of such genes has remained confined to single gene approaches. Here we utilized the powerful genetics and translucency of zebrafish larvae to undertake a medium-throughput screen of genes implicated in human hematopoietic variation. The results of this screen will help us to tentatively place novel genes in molecular pathways and thus close the ever-increasing knowledge gap on the biological function of gene candidates identified by genomic technologies.
The maintenance, embryo collection and staging of the wild type (Tubingen Long Fin) and transgenic zebrafish lines (Tg(cd41:GFP), Tg(fli1:GFP), Tg(c-myb:EGFP) were performed in accordance with EU regulations on laboratory animals, as previously described [29], [30].
Morpholinos (GeneTools, LLC) were re-suspended in sterile water and diluted to chosen concentration. Approximately 1 nl was injected into embryos at 1- to 2-cell stage. MOs used are summarized in Table S2.
Plasmid with full-length human hBRD3 cDNA was purchased from Source Bioscience (Nottingham, UK). hBRD3 was cloned into pCS2 expression vector using gene-specific primers: AATTACATCGATACCATGTCCACCGCCACGACAGT (forward) and CCCGAGTCTAGACTATTCTGAGTCACTGCTGTCAGA (reverse) and AAATTAGAATTCACCATGTCCACCGCCACGACAGT (forward) and ATGTTAACCGGTAGTTCTGAGTCACTGCTGTCAGA (reverse) for cloning into the pCS2-EGFP vector. Restriction enzyme sites (ClaI/XbaI and EcoRI/AgeI respectively) used for cloning are underlined. Zebrafish full-length rcor1 cDNA was cloned into pCS2 vector using gene-specific primers: GTTATAGAATTCATGCCCGCAATGTTAGAGAAG (forward) and AGGCGCCTCGAGTCAGGAAACCGAAGGGTTCTG (reverse). Restriction enzyme sites (EcoRI and XhoI, respectively) are underlined.
hBRD3, hBRD3-GFP and rcor1 mRNA was synthesized with mMESSAGE mMACHINE kit (Ambion), according to the manufacturer's protocol. In the rescue experiment, 100 pg of hBRD3 mRNA or 125 pg rcor1 RNA was injected into the one cell stage control and MO-injected Tg(cd41:EGFP) embryos. In the hBRD3 localization experiment, 300 pg of hBRD3-GFP mRNA was injected into Tubingen Long Fin embryos at 1-cell stage.
In order to verify the effectiveness of MOs in affecting their target transcripts, RT-PCR was performed. RNA was subjected to reverse transcription using Superscript II Reverse Transcriptase (Invitrogen). PCR was performed using gene-specific primers (listed in Table S2) and KOD Hot Start DNA Polymerase (Novagen).
In situ hybridization was performed with riboprobes specific for c-myb, αe1-globin, mpx, mpeg and rag1 as previously described [31], as well as for brd3a, brf1b, kalrn1, waspla and wdr66. Primers used for PCR amplification of candidate genes for probe synthesis are listed in Table S3. Photomicrographs were taken with a Zeiss camera AxioCam HRC attached to a LeicaMZ16 FA dissecting microscope (Leica Microsystems, Germany) using the AxioVision software.
O-dianisidine staining was performed as previously described [32].
Sudan Black staining was performed as previously described [33].
Clotting time assay was performed as previously described [7]. In short, 5 dpf larvae were anaesthetized in 0.02% tricaine solution in embryonic water and transferred onto a Petri dish in a small drop of liquid. Caudal veins of the larvae were wounded with the tip of a Microlance needle (0.4 mm×13 mm, Becton Dickinson) in the anal area. For each larva the time passing between inflicting the wound and the stop of bleeding was recorded.
Genomic DNA was isolated from 5 dpf larvae, which were individually loaded into wells of a 96-well plate. Fish were incubated in 25 µl of lysis buffer (25 mM NaOH with 0.2 mM EDTA) at 95°C for 30 min. Afterwards 25 µl of neutralization buffer (40 mM Tris-HCl) was added. Genotyping was carried out using the KASP genotyping assays (KBioscience). Each reaction consisted of 4 µl of genomic DNA and 5 µl of PCR mix, according to the manufacturer's protocol (KBioscience). PCR products were analyzed using PHERAstar plus (BMGlabtech) and KlusterCaller software (KBioscience).
Images were captured with the use of a Leica TCS SP5 confocal microscope with Leica LAS AF software (Leica Microsystems), using a 40× immersion lens or with Axio Zoom.V16 fluorescent microscope with AxioCam MRm camera using 260× magnification.
Selective inhibitor of human BET family of bromodomain-containing proteins, thieno-triazolo-1,4-diazepine, named JQ1, was kindly provided by Dr Chas Bountra, Structural Genomics Consortium, University of Oxford, Oxford, UK. Both the active inhibitor, (+)-JQ1, and its inactive stereoisomer, (−)-JQ1, were dissolved in dimethylsulfoxide (DMSO) to 10 mg/ml and stored in aliquots at −20°C. For zebrafish embryo treatment, (+)-JQ1 and (−)-JQ1 were diluted in egg water to desired concentration and added to the embryos at ∼6 hpf, 24 hpf or 3 dpf and afterwards replaced daily. Control embryos were incubated in the equal concentration of DMSO in egg water as inhibitor-treated embryos.
|
10.1371/journal.pgen.1002059 | Alkylation Base Damage Is Converted into Repairable Double-Strand Breaks and Complex Intermediates in G2 Cells Lacking AP Endonuclease | DNA double-strand breaks (DSBs) are potent sources of genome instability. While there is considerable genetic and molecular information about the disposition of direct DSBs and breaks that arise during replication, relatively little is known about DSBs derived during processing of single-strand lesions, especially for the case of single-strand breaks (SSBs) with 3′-blocked termini generated in vivo. Using our recently developed assay for detecting end-processing at random DSBs in budding yeast, we show that single-strand lesions produced by the alkylating agent methyl methanesulfonate (MMS) can generate DSBs in G2-arrested cells, i.e., S-phase independent. These derived DSBs were observed in apn1/2 endonuclease mutants and resulted from aborted base excision repair leading to 3′ blocked single-strand breaks following the creation of abasic (AP) sites. DSB formation was reduced by additional mutations that affect processing of AP sites including ntg1, ntg2, and, unexpectedly, ogg1, or by a lack of AP sites due to deletion of the MAG1 glycosylase gene. Similar to direct DSBs, the derived DSBs were subject to MRX (Mre11, Rad50, Xrs2)-determined resection and relied upon the recombinational repair genes RAD51, RAD52, as well as on the MCD1 cohesin gene, for repair. In addition, we identified a novel DNA intermediate, detected as slow-moving chromosomal DNA (SMD) in pulsed field electrophoresis gels shortly after MMS exposure in apn1/2 cells. The SMD requires nicked AP sites, but is independent of resection/recombination processes, suggesting that it is a novel structure generated during processing of 3′-blocked SSBs. Collectively, this study provides new insights into the potential consequences of alkylation base damage in vivo, including creation of novel structures as well as generation and repair of DSBs in nonreplicating cells.
| DNA double-strand breaks (DSBs) are an important source of genome instability that can lead to severe biological consequences including tumorigenesis and cell death. Although much is known about DSBs induced directly by ionizing radiation and radiomimetic cancer drugs, there is a relative dearth of information about the formation of derived DSBs that arise from processing of single-strand lesions. Since as many as 10,000–200,000 single-strand lesions have been estimated to occur each day in mammalian cells, conversion of even a small percentage of such lesions to DSBs could dramatically affect genome stability. Here we addressed the mechanism of formation and repair of derived DSBs in vivo during the processing of DNA methylation damage in yeast that are defective in base excision repair (BER) due to a lack of AP endonucleases. Armed with a technique developed in our lab that detects resection at DSBs, a first step in DSB repair, we demonstrated formation of DSBs in G2 cells and the role of recombinational repair in subsequent chromosome restitution. Furthermore, we have identified a novel repair intermediate that can be generated if abasic sites are nicked by AP lyases, providing additional insights into the processing of 3′-blocked groups at single-strand breaks.
| DNA double-strand breaks (DSBs) are important sources of genome instability, giving rise to chromosomal aberrations and severe biological consequences including tumorigenesis and cell death [1], [2]. We and others also showed that regions adjacent to DSBs are prone to mutagenesis through a variety of mechanisms [3]–[7]. DSBs can be induced directly by exposure to DNA-damaging agents such as ionizing radiation (IR) and radiomimetic chemicals. While there is a great deal of information about direct DSBs, little is known about the contribution of single-strand lesions to the production of DSBs, although single-strand lesions are generally accepted to be a source of DSBs via replication fork collapse in regions of single-strand DNA [8].
A common single-strand lesion that is generated during normal cell metabolism and repair is an apurinic/apyrimidic (AP) site, one of the most abundant DNA lesions in the cell [9], [10]. As many as 10,000–200,000 single-strand lesions appear each day in mammalian cells [11], [12]. Most of these are subject to base excision repair (BER), a highly coordinated process initiated by a lesion-specific glycosylase removing damaged bases and forming AP sites. Removal of AP sites by AP endonucleases or AP lyases involves the generation of single-strand breaks (SSBs) with blocking groups at their 3′ or 5′-ends that cannot be joined by DNA ligases [13]–[15]. Subsequent SSB end-processing involves a diverse set of enzymes/pathways to deal with the termini [16].
Single strand lesions which are produced by many mutagens are also potential sources of DSBs if they are processed to form closely-opposed SSBs. Closely-opposed SSBs could result in derived DSBs simply through loss of pairing of short DNA duplex regions bounded by the SSBs, as shown by in vitro analysis [17]–[19] and a limited number of in vivo studies [20], [21]. A DSB could also be generated if two more distant SSBs are processed to form closely-opposed SSBs. This second category of derived DSBs have been proposed following induction of methyl methanesulfonate (MMS) lesions and subsequent processing of AP sites to 5′-blocked SSB termini in rad27/FEN1 and pol32 mutants [20]. Removal of these 5′-blocked SSB ends involves DNA synthesis and strand displacement that can move distant SSBs closer [20], [22]. However, there is little information about in vivo generation of derived DSBs from nearby opposed SSBs with 3′-blocked termini. Such termini are a challenge to the repair machinery since they must be removed to enable repair synthesis at 3′-OH ends [23], [24]. Besides being formed directly from sugar damage, SSBs with 3′-blocked termini, the α,β-unsaturated aldehyde (3′-dRP), can be generated during incision at the 3′-side of AP sites by AP lyase [14].
In the budding yeast Saccharomyces cerevisiae, the AP endonucleases Apn1 and Apn2, which have 3′-phosphodiesterase activity, are responsible for removing most 3′-dRP ends as well as other blocking groups [25]–[27]. Previously, we found that deletion of both AP endonucleases appears to lead to accumulation of chromosome breaks in nongrowing G1 haploid yeast [28] and the number of chromosome breaks increased with time of liquid-holding in buffer. However, it was possible that the DSBs were not formed in vivo but actually appeared during subsequent pulsed-field gel electrophoresis (PFGE) processing. Therefore, while those findings highlighted the potential for single-strand damage to generate DSBs, they did not definitively show that the derived DSBs were generated in vivo or that they could be generated later in the cell cycle. Importantly, there was no evidence of repair of the DSBs, which is not surprising since the haploid cells were in the G1 phase of the cell cycle when there would be no recombinational partner. Our previous study also showed that non-homologues recombination (NHEJ) has little if any role in dealing with the derived “DSBs” caused by MMS in G1 cells based on deletion of yku70, especially for the apn1/2 mutant which accumulated DSBs even though it has the wild type NHEJ machinery [28].
Although there is abundant genetic evidence for homologous recombinational (HR) repair proteins dealing with MMS damage-induced lesions in a variety of systems, there has been no direct demonstration of MMS-induced DSBs being formed or subsequent repair in vivo. It is generally assumed, though not proven, that recombinational repair deals with DSBs generated during replication fork collapse following induction of MMS lesions. Here, we demonstrate MMS can generate derived DSBs within G2/M arrested cells and that these DSBs are processed and undergo repair. Utilizing our recently developed PFGE assay [29], we establish that MMS-derived DSB ends are subject to resection, one of the earliest steps in DSB repair. In addition we identify a novel repair intermediate detected as slow mobility chromosomal DNA during PFGE, providing additional insights into the processing of 3′-blocked groups in vivo.
We previously described a system using PFGE for analyzing in vivo repair of alkylation base damage caused by MMS [28] in yeast that is based on detection of chromosome breaks. Though MMS does not cause SSBs directly [28], [30], they can arise as repair intermediates during BER. If the SSBs are closely-spaced on complementary DNA strands, they are detected as “DSBs” with PFGE. Most of the closely-opposed single strand lesions were shown to be efficiently repaired in stationary G1 haploid wild type cells by BER [28], thereby preventing the formation of derived DSBs in vivo. We now extend this system to a characterization of derived DSBs in G2 cells where there is the opportunity for recombinational repair between sister chromatids.
Haploid yeast were grown to log phase in rich medium (YPDA), arrested in G2/M with the microtubule and mitotic spindle disrupter drug nocodazole and treated with 0.1% MMS (11.8 mM) for 15 min in PBS. They were subsequently incubated in YPDA+nocodazole to prevent G2 cells from progression into the next cell cycle stage. Changes in chromosomes at various times after treatment were determined using PFGE. The treatment of WT cells with MMS did not cause fragmentation of chromosomes (which range in size from ∼200 kb to ∼2.5 Mb; Figure 1), suggesting that there is efficient repair of single strand damage and, therefore, no apparent generation of DSBs. There was also no apparent reduction in survival (Figure 2). However, MMS treatment of apn1 apn2 (apn1/2) cells led to loss of all but the smaller chromosome bands (Chr I, 230 kb and Chr VI, 270 kb) as well as decreased survival (Figure 2). At later times, there is “restitution” of the broken chromosomes (i.e., formation of full size chromosomes) as shown in the ethidium bromide stained gel (indicated in Figure 1), and the survival is somewhat higher for MMS-treated apn1/2 after 8 hour G2/M holding. Surprisingly, there was also a rapid accumulation of slow-moving DNA (SMD) that appeared below the well. These results differ from those with MMS-treated stationary arrested G1 apn1/2 cells which did not give rise to SMD although chromosome breakage was detected by PFGE [28]. The amount of SMD decreased after 4 hours, at which time restituted chromosomes were detected. The mechanism(s) of induction and disappearance of DSBs and SMD, as well as possible relationship, is investigated below.
In yeast, the first step in the BER of MMS-induced lesions requires Mag1 glycosylase which removes damaged bases and forms abasic sites [31]. To confirm that the DSBs as well as SMD resulted from BER, the MAG1 gene was deleted in the apn1/2 background. As shown in Figure 3, the appearance of DSBs and SMD requires at least the first step in BER in G2 arrested cells since there was no apparent change in chromosomes and no SMD formation in the triple apn1/2 mag1 mutants following MMS treatment. Although methylated bases can be spontaneously depurinated to form AP sites, the impact of this process to DSB formation is limited as indicated by the limited appearance of DSBs in the apn1/2 mag1 mutant (Figure S3). The formation of AP sites and subsequent DSBs were also prevented in the triple mutant arrested as G1 stationary cells [28]. The absence of Mag1 also resulted in a considerable increase in toleration of MMS damage (Figure 2). We examined further the role of AP sites by including methoxyamine (MX) during the MMS treatment and subsequent incubation of the apn1/2. Methoxyamine covalently binds to AP sites, preventing subsequent BER processing [32]. The MX results shown in Figure 3 were similar to those observed with the mag1 apn1/2 mutant. Thus, the appearance of DSBs and SMD in the G2/M cells lacking the Apn1 and Apn2 endonucleases, along with increased MMS hypersensitivity, requires the generation of AP sites by BER and/or repair events downstream of AP sites.
To further address how DSBs are generated by MMS in the apn1/2 mutant and their repair in G2/M cells as well as the mechanism of SMD appearance and loss, we first investigated whether HR has a role in these processes. Deletions of key genes involved in HR including RAD50, -51, 52, -54 and MRE11 were generated in the apn1/2 background. While HR mutants are MMS sensitive even in an APN+ background [33], they do not affect the appearance of MMS-induced chromosomal damage or repair in G1 stationary cells as compared to wild type cells [28] because of efficient BER. The reports of MMS sensitivity of HR mutants are likely due to small number of lesions that remain unrepaired when cells pass into S-phase [34]. Nothing is known about the induction and repair of MMS derived DSBs in G2 cells where there are opportunities for recombinational repair between sister chromatids. Similar to results in G1 cells, there appeared to be little or no induction of derived DSBs when rad52 APN+ cells were treated in the G2/M phase of the cell cycle (Figure S1). However, efficient restitution of full-length chromosomes in the apn1/2 cells treated with MMS in G2/M does require components of the recombinational repair pathway as shown in Figure 4A (rad52 and rad51), Figure 5A (mre11 and rad50) and Figure 6A (mre11 and rad54). The small amount of chromosome restitution in some of the mutants might be due to some sort of microhomology mediated end-joining given the RAD51 independence and RAD52 dependence. Thus, in contrast to the situation in G1 cells, where there are no opportunities for recombination, derived DSBs created in G2 cells by MMS can undergo recombinational repair between sister chromatids (also addressed below using an mcd1 cohesin mutant).
The gain and loss of SMD in the apn1/2 mutant appeared to parallel the timing of chromosome breakage and restitution, as shown in Figure 4A. However, SMD formation is not related to recombination since additional deletion of RAD52, RAD51 (Figure 4A), RAD50, MRE11 ((Figure 5A), or RAD54 (Figure 6A) did not significantly alter the appearance of SMD. To compare the levels of SMD in response to MMS and between strains, we determined the ratios between amount of material in the SMD region to DNA in the small chromosomes that experienced little breakage. As shown at the bottom of panel A in each figure (“SMD/Chr I+VI”), the ratios were similar between the apn1/2 and the various triple mutants over the first hour of treatment. In all cases, there was considerable reduction in SMD by 4 hours. The decreased amount of SMD in rad52 or mre11 (in Figure 4A, Figure 5A) could be due to delays in S-phase arising from repair defects which would lead to an overall reduction in DNA entering into the gel since the mutants have a somewhat reduced growth rate. We also investigated the 5′ to 3′ exonuclease I (EXO1) since it can resect at random DSBs and appears to enlarge single-strand gaps during nucleotide excision repair [35] that could lead to reduced chromosomal DNA mobility during PFGE. As shown in the Figure S2, exonuclease 1 does not influence either the repair of the MMS induced DSBs or the appearance of SMD.
The role of HR in repair of the MMS-derived DSBs and the lack of contribution to the appearance and loss of SMD is investigated further in Figure 6B. Using a LEU2 probe for Chr II+III, it is clear that nearly all the DNA of Chr II appears as SMD within 30 minutes after treatment. (This probe, which identifies Chr II, also hybridizes with circular and broken Chr III molecules, as described in Ma et al. [28] and is discussed below.) The amount of SMD started to decrease at 2 hours after MMS exposure and a significant amount of SMD is lost by 4 hours, where at the same time there is a reappearance of full size Chr II. However, there is substantial restitution of Chr II starting at 4 hours in the apn1/2 mutant, but not in the mre11 and rad54 derivatives. Thus, while the appearance and loss of SMD are not influenced by HR, based on the Southern blotting results with the apn1/2 mre11 and apn1/2 rad54, the reappearance of full-size Chr II requires HR.
While a role for recombinational repair of MMS-associated DSBs has been proposed for S-phase cells and demonstrated above for G2/M cells, there has been no direct demonstration of DSB processing or generation of recombinants. This is due in part to the difficulties of characterizing events associated with random DSBs (discussed in [29]). In addition, opportunities to examine MMS-induced events in S-phase cells using PFGE are limited because of the structures created in the replicating DNA which results in most of the DNA being retained in the starting wells [36].
Recently, we described an assay involving PFGE and circular chromosomes for characterizing resection and recombination at random DSBs [29]. Since a single DSB in a circular chromosome results in a unit length linear molecule, the direct or derived induction of random DSBs can be followed by the appearance of the corresponding band with PFGE (as described in [28], [29], [37]). Importantly, resection at the DSB ends leading to the generation of single-strand tails could be detected by reduced mobility of the unit length molecules (i.e., “PFGE-shift”). Previously, we showed that MMS treatment of apn1/2 stationary G1 cells led to the linearization of the circular Chr III. However these “DSBs” could have arisen from molecules with closely spaced-SSBs during preparation of chromosomal material for PFGE analysis. The resection in the G2 cells, as well as subsequent repair, establishes that MMS-induced DSBs actually occur in vivo in G2 cells.
Similar to the results with the G1 stationary cells, MMS treatment of apn1/2 cells resulted in the rapid appearance of linear Chr III molecules in Southern blots using a probe specific to this chromosome (Figure 4B and Figure 5B). However, unlike observations with apn1/2 cells treated in G1 [28], the linearized molecules from the G2/M cells exhibited the PFGE-shift similar to that found previously for direct DSBs induced by IR [29], suggesting that the MMS derived DSBs are subject to resection. The PFGE shift which appeared by 1 hour after treatment was also found in the apn1/2 rad52 and rad51 triple mutants (Figure 4B). The bulk of PFGE-shift required MRX (Mre11, Rad50, Xrs2) since there was no apparent resection with the apn1/2 rad50 or mre11 mutants (Figure 5 and Figure 6). The PFGE profiles of DSBs induced by MMS in the triple mutants of apn1/2 combined with deletions of the recombinational repair genes were similar to patterns found for rad52, rad51, rad50 and mre11 single mutants exposed to IR [29]. Thus, the MMS derived DSBs are subject to resection; MRX plays an important role, presumably through initiation of 5′ to 3′ resection. The processing of DSBs induced by IR appears different from that found for MMS-derived DSBs in that the PFGE shifted band is smeared as compared to a narrow shifted band for IR damage (e.g., the rad52 mutant; [29]). The smearing of the band after MMS treatment might be due to the formation of single-stranded tails with variable lengths or the timing of DSB formation and resection (Figure 4B).
Similar to our findings with radiation-induced direct DSBs [29], the MMS treatment of apn1, 2 cells leads to the appearance of linear Chr III molecules at 2 to 4 hours post-treatment that are twice the size of the broken Chr III (Southern blots in Figure 4B and Figure 5B). Based on our previous studies [29], these dimers are likely the product of recombination between full size Chr III sister chromatids (discussed in [29]). The role of recombination is supported by the observed dependence on Rad51 and Rad52 (Figure 4A). Overall these results provide the first direct physical evidence of i) MMS lesions being processed to DSBs in G2 cells, ii) resection of the ends, and iii) MMS generation of recombinant molecules.
In the absence of the Apn1 and Apn 2 endonucleases, AP sites can be nicked at the 3′ side by the bifunctional DNA N-glycosylases/AP lyases Ntg1 and Ntg2 that convert AP sites into 3′-blocked SSBs [13]. Additionally, the bifunctional 8-oxyguanine glycosylase Ogg1 can nick an AP site that is opposite a cytosine [38]. Mutants of these genes were created in the apn1/2 background to identify possible contributors to the appearance of DSBs and SMD. As shown in Figure 7, the apn1/2 ntg1/2 quadruple mutant also exhibited appearance and loss of SMD. However, there was less chromosome breakage and nearly full chromosome restitution by 4 hours, as compared to the incomplete chromosome restitution in the apn1/2, even after 8 hours. The additional ntg1/2 mutations increased the time required for maximal appearance of SMD and there is less material lost in the full-size chromosome bands, consistent with a decreased likelihood of single-strand break generation. The increased resistance of apn1/2 ntg1/2 compared to apn1/2 cells suggests that DSBs rather than SMD are the major contributor to loss of survival with or without arrest in G2 following MMS treatment (Figure 2; the dose-modifying factor is ∼3). The combination of apn1/2 ntg1/2 does not totally block formation of DSBs. While there appears to be less processing of ends, recombinants can still be formed based on the formation of Chr III dimers (Figure 7B). A similar finding of dimer generation, independent of resection, was reported for MRX mutants following IR [29].
Further deletion of OGG1 (i.e., apn1/2 ntg1/2 ogg1) greatly reduced the appearance of SMD formation and decreased DSB induction (Figure 7). Survival was also improved compared to the apn1/2 ntg1/2 mutant (Figure 2). Thus, SSBs generated at AP sites are a likely source of SMD consistent with the above findings with the apn1/2 mag1 mutants as well as the effect of MX (Figure 3) where AP sites are prevented or blocked. Furthermore, these results demonstrate a role for OGG1 in the general processing of methylated base damage or imply that MMS causes additional types of damage that are substrates for Ogg1. The reduced amount of overall chromosome breakage observed with the ethidium bromide stained gel (Figure 7A) is expected if there are less SSBs to generate derived DSBs. Surprisingly, some linearized Chr III molecules and dimers were generated, based on Southern blotting with a probe specific to Chr III, even though there is no resection. Possibly they were formed through additional mechanisms for processing abasic sites and/or in vitro during PFGE of DNAs with opposed SSBs that are sufficiently close.
While the results with the various RAD mutants demonstrate that neither the appearance nor the disappearance of SMD is dependent on HR, there is still the possibility of SMD arising through unknown interactions with sister chromatids. To address this, a temperature-sensitive cohesin mutant mcd1-1 was generated in both a WT and apn1/2 background. Cohesin is required to hold sister chromatids together and is essential for efficient repair of radiation induced DSBs [39]–[41]. The mcd1-1 single mutant grows well at permissive temperature of 23°C but is not viable at 37°C [42], consistent with our observation of no growth of the apn1/2 mcd1-1 triple mutant at the elevated temperature. In preliminary experiments we found that apn1/2 cells exhibited less repair of MMS lesions at 37°C. This may be due to more closely-opposed lesions being converted into DSBs, since methylated bases are heat-labile. Therefore, cells were incubated at 37°C for 3 hours to inactivate the temperature-sensitive cohesin; during this period cells were arrested in G2/M by nocodazole. Cells were then shifted to the semi-permissive temperature of 30°C for MMS treatment as well as post-MMS incubation in YPDA with nocodazole.
As shown in Figure 8, MMS did not lead to the appearance of DSBs or loss of chromosomes in the mcd1-1 single mutant, similar to results with wild type cells (Figure 1). As expected, derived DSBs were detected in the apn1/2 mcd1-1 triple mutant. However, unlike observations with the apn1/2 mutant (Figure 1) there was much less overall restitution of chromosomes at 4 and 8 hours, more similar to the apn1/2 rad51 mutant (Figure 4). Yet, the formation and disappearance of SMD was not affected. Most chromosome bands, especially the larger chromosomes, were lost from the PFGE gels in the apn1/2 mcd1-1 triple mutant but not in the mcd1-1 single mutant (Figure 8), which is comparable to results with other HR mutants. These findings suggest that cohesin is not required during SMD formation and that SMD is not a consequence of sister chromatid interaction.
As described above, the generation of SSBs at abasic sites is required for the generation of SMD. However, the SMD molecules do not rely on sister chromatid interactions, recombination or DSB ends, based on a lack of SMD at later times in the HR mutants (Figure 4, Figure 5, and Figure 6). Examination of several of the ethidium bromide gels, reveal that SMD is dependent on chromosome size (see, for example, Figure 1 and Figure 4A). In several of the experiments there is little disappearance of the smallest chromosomes (I and VI). This is confirmed in Figure 6 where Chr II (800 kb) and linearized Chr III can be detected with a common probe. Nearly all the larger Chr II molecules appear in SMD, while there is little change in the amount of the smaller Chr III. This is further substantiated in Figure 4 and Figure 5, where only Chr III is probed and there is little, if any SMD. With a decrease in SSBs, there is less SMD with only the large chromosomes being affected (Figure 7). Thus, the SSB related lesion(s) or combination of lesions leading to SMD appear to be less than an average of 1 per few hundred kb.
Given the retardation of much of the chromosomal DNA in PFGE, we investigated various enzymes that recognize structural changes in DNA to help discern the nature of SMD. (We also considered proteins bound to DNA that could give rise to SMD; however, we found that extending the proteinase K treatment beyond that normally used in preparation of plugs for PFGE did not change the SMD as noted in “Material and Methods”.) Mung bean nuclease had been used to demonstrate resected DNA at radiation induced DSB ends [29]. However, for the DNA obtained after MMS treatment, there was general degradation by this nuclease of the chromosomal DNA treated in plugs before PFGE. This extensive digestion is likely due to this nuclease acting at SSBs and possibly gap-like structures. We also investigated bacteriophage T7 endonuclease I because of its ability to recognize and cleave at a variety of structures including DNA mismatches, nicks as well as branch molecules such as Holliday type junctions [43]–[45]. As shown in Figure 9, endonuclease I treatment of the chromosomal DNAs following MMS treatment of apn1/2 cells eliminated much of the SMD leading to the appearance of small DNA fragments (∼50 to 100 kb). This enzyme specifically acted on SMD since there was little effect on the chromosomal DNA from untreated cells or the chromosomal DNA following 4 hour to 8 hour of repair. The cutting of SMD into small molecules by T7 endonuclease I (i.e., smaller than the Chr I and Chr VI which exhibit little SMD) suggests the presence of few endonuclease responsive substrate structures in the smaller chromosomes. Interestingly, the use of T7 endonuclease I to remove the SMD enabled us to establish further that repair of derived DSBs does occur between 2 and 4 hours (as previously suggested in Figure 4, Figure 5, Figure 6). The SMD structures sensitive to T7 endonuclease I are unlikely to be due to recombinational intermediates because SMD was observed in the various HR mutants as described above. While they could be related to branched molecules produced from persisting nicks, other possibilities exist given the various types of structures susceptible to this endonuclease.
BER is critical for dealing with a variety of single strand lesions. Many enzymes in this pathway are conserved from microorganisms to humans and serve as antimutators, especially in terms of tumor suppression and preventing hereditary neurodegenerative disease [46], [47]. Aberrant BER processes might result in the eventual appearance of DSBs, which are a major source of genome instability. MMS-induced lesions are considered a source of DSBs as a result of collapsed replication forks at the lesions or processed intermediates. Based on genetic evidence, these replication-associated DSBs have been considered to be repaired by HR mechanisms.
Previously, we showed that MMS-induced single-strand damage in G1 arrested cells had the potential for generating derived DSBs and highlighted the role that Rad27 and Pol32 play in preventing such breaks [20]. We had concluded that closely-spaced opposing lesions could be a source of the derived DSBs and that well-coordinated BER assures prevention of these downstream DSBs. The present study using G2/M cells is the first to characterize directly the generation, processing and repair of derived DSBs following treatment by an alkylating agent. While two closely-opposed SSBs with 5′-blocked termini could be “moved closer” to form a DSB during repair-associated DNA synthesis and strand displacement [17], [18], [19], [20], [21], this is not expected to be the case for SSBs with 3′-blocked termini since they would not support DNA synthesis directly. The present results are consistent with derived DSBs resulting from generation of SSBs at closely opposed lesions. It is also possible that derived DSBs could arise through processing of more distant SSBs with 3′-blocked ends in cells lacking AP endonucleases, in essence the breaks are “moved closer,” as discussed below. Furthermore, we have described a novel repair intermediate, SMD, which can be generated if abasic sites are nicked by AP lyases.
While the APE1 gene, which codes for the major mammalian AP endonuclease (the APE1 homologue APE2 has only weak endonuclease activity, and its role in human BER is not clear), is essential for human cell survival and results in embryonic lethality when knocked out in mouse [48]–[51], yeast can survive the deletion of both AP endonucleases with almost no growth defect. It is, therefore, possible to study alternative mechanisms/pathways that deal with AP sites and 3′-blocked SSBs in vivo and their role in generating DSBs. Earlier studies had demonstrated that MMS does not cause DSBs directly [28], [30]. With an assay that can specifically monitor the processing of closely-opposed single strand lesions, our previous study showed that PFGE-detected DSBs were accumulated in G1 apn1/2 haploid yeast after MMS damage. However, since closely-opposed SSBs might lead to chromosome DNA breakage during in vitro handling, the extent to which the PFGE-detectable DSBs were actually formed in vivo remained a question. Here, we confirm that DSBs do appear after MMS treatment of G2 cells lacking AP endonucleases, as demonstrated by i) resection, ii) a requirement for HR components to reconstitute chromosomes, and by iii) the formation of Chr III dimers. While resection is generally considered essential for DSB repair mechanisms [52], [53], we have demonstrated that it also occurs at the MMS-derived DSBs and like radiation-induced DSBs they are subject to MRX control. As in the case of randomly generated radiation-induced DSBs [29], we aimed to determine other factors affecting resection at the MMS-derived DSBs, especially factors that may lead to increased resection. We have recently shown that UV as well as MMS damage to single-strand DNA formed at site-specific DSBs cause high level of mutagenesis [4], [5]. Increasing resection at MMS derived breaks could further enhance its mutagenic potential.
The current results further confirm that DSBs can be derived from AP sites arising during BER, since the appearance of DSBs could be blocked either at the step in which methylated bases are removed or if cleavage of AP sites is prevented by MX (Figure 3). It is clear that DSBs were generated by the bifunctional glycosylases because deletion of NTG1, NTG2 and OGG1 along with APN1 and APN2 blocked the formation of DSBs as well as resection. The targets of these enzymes are limited to AP sites instead of methylated bases based on efficient DSB inhibition following deletion of MAG1 (Figure 3). Though OGG1 is known to deal primarily with oxidative damage [14], [54], we have shown that this bifunctional glycosylase provides a backup for cleavage at AP sites following induction of MMS damage since derived DSBs that appeared in the apn1/2 ntg1/2 mutant were prevented by a further ogg1 mutation (Figure 7). This is the first direct demonstration for the Ogg1 glycosylase dealing with lesions other than oxidative damage in vivo, suggesting a potentially more general role for this gene in repair. Considering that the predominant lesions induced by MMS are N7-methylguanine and N3-methyladenine [55], the function of Ogg1 in the development of DSBs and SMD is likely due to its action on AP sites derived from N7-methylguanine. There was still a small amount of DSBs after removal of all the bifunctional glycosylases/lyases (Figure 7) which might be due to NER or some other enzymes. It was shown that DNA Topoisomerase I (Top1) forms DNA-protein adducts with nicked and gapped DNA structures [56], [57]. Possibly the AP sites could also be processed by yeast topoisomerases to generate DSBs. This might explain the small amount of SMD presented in apn1/2 ntg1/2 ogg1 mutants (Figure 7).
As summarized in Figure 10, the generation of derived DSBs would require that opposed AP sites either be sufficiently close (left side of figure) so that DSBs are created directly in vivo or there is a nick-processing mechanism that “moves” the relatively distant opposing-nicks closer (central part of figure) to form a DSB. Considering that MMS is an SN2 type of alkylating agent that methylates DNA bases in a random manner with a limited ability to produce closely-spaced lesions under the conditions used in this study (in contrast to ionizing radiation [58]), many of the MMS-derived DSBs might be generated from distant single-strand breaks during processing/repair of the end-blocking groups as also suggested from our previous study with rad27 and pol32 mutants [20]. Since AP lyases generate blocked 3′-ends (3′-dRP) while repair of either SSBs or DSBs requires an unblocked 3′-OH end for repair synthesis or ligation, we suggest that both the formation of DSBs and SMD are related to the processing of 3′-blocking groups. Either or both might be generated through development of 3′-flaps, possibly by helicases or nucleases. For example, opposing SSBs could be “moved” together to form a DSB if 3′-flaps are generated toward each other. Possibly it is the generation of multiple flaps that leads to the reduced mobility of large DNAs on PFGE, and the SMD molecules; however, the reduction in mobility is not as great as observed with replicating chromosomes, which remain in the well during PFGE. Although exonuclease 1 generated gaps at UV-damage sites can lead to reduced mobility, they are unlikely to be the source of SMD in the present experiments, While we have shown that the DSBs and SMD arise from a common BER intermediate, their subsequent appearance and disappearance are genetically separable. Importantly, we have established that SMD does not involve the HR pathway. The derived DSBs are subject to processing by MRX and the subsequent DSB repair as well as the appearance of dimer recombination products requires HR.
Regardless, there are limitations on the appearance of SMD. The loss of chromosomal DNA along with the appearance of the wide band of SMD following MMS treatment of apn1/2 cells is dependent on the size of the chromosome. SMD was substantially greater for larger chromosomes than smaller ones (Figure 6). This is clearly shown in a Southern blot comparison of linearized Chr III with Chr II and in comparisons of the 230 kb Chr I and 270 kb VI with the larger chromosomes where there was little if any loss of the smaller chromosome bands (Figure 6). Based on our previous results [28], we anticipate ∼0.4 SSBs/kb which would lead to considerable damage in even the smallest chromosomes (∼100 SSBs/Chr I). Thus, while SMD requires the generation of SSBs, other factors determine its appearance. Possibly, the appearance of SMD depends simply on the likelihood of producing some minimum amount of lesions or certain types of structures (i.e., sensitive to T7-endonuclease) that are stable in vitro.
The requirement for generation of a 3′-flap to remove 3′-blocked termini had been proposed previously to explain the synthetic lethality between apn1/apn2 and rad1 or rad10 [25], [59]. Although in vitro studies demonstrated that a 3′-flap can be removed by Rad1/Rad10 proteins [60], direct evidence for flap removal in vivo has been lacking. The observation of SMD in our current study fits well with this hypothesis though the actual mechanism for its formation and release might be more complex than previously proposed. It is interesting that while we have eliminated SMD as a recombination product, it is sensitive to the T7 endonuclease I which can cleave structures that might arise during recombination as well as branched molecules containing single strand regions (possibly as a result of 3′-flap formation as proposed in Figure 10).
In conclusion, our study identifies new mechanisms for processing abasic sites and provides the first direct demonstration in nonreplicating G2/M cells of MMS-derived DSBs and that the DSBs are subject to recombinational repair. In addition, we identify and characterize the generation of SMD. While not previously described, possibly because of the techniques used to assess DNA damage and repair, SMD might be a general repair intermediate for various types of DNA damage, a view that we are currently pursuing. Interestingly, there has been an indication, though not directly addressed, of SMD-like material in exonuclease 1 defective yeast cells during excision repair of UV damage [35]. The combination of genetics and systems for detection of novel structures has provided a unique opportunity to address processed events at intermediates in repair of DNA lesions. While the derivation and repair of derived DSBs has been addressed as well as the generation of SMD, it will be interesting to determine the specific nature of the actual DNA changes that lead to SMD and the eventual resolution including the genetic controls. To our knowledge, this is the first report of a novel branched repair intermediate being generated during the processing of 3′-blocked termini. These findings are expected to expand our understanding of mechanism for repair of 3′-blocked ends as well as their impact on genome stability.
All strains are haploid derivatives of two isogenic haploid yeast strains MWJ49 and MWJ50 (MATα leu2-3,112 ade5-1 his7-2 ura3D trp1-289) which contain a circularized chromosome III and has the construct lys2::Alu-DIR-LEU2-lys2D5′ on Chr II [28]. The construction of strains with circular Chr III was described in [28]. Deletion strains of apn1, apn2, rad50, rad51, rad52, mre11, mag1, ntg1, ntg2, ogg1 and derived multiple mutants were created by replacement of the relevant open reading frame with selectable markers by PCR [61]. Temperature-sensitive mutants of mcd1-1 in wild type or apn1/2 background were generated using plasmid pVG257 [41]. Experiments were done at 30°C, unless specifically stated at a different temperatue.
The generation of G2 arrested cells was described in [62]. Briefly, logarithmically growing cells in YPDA medium (1% yeast extract, 2% Bacto-Peptone, 2% dextrose, 60 mg/ml adenine sulfate) were incubated with nocodazole at a final concentration of 15 µg/mL. After 3 hours, most cells are arrested in G2/M as determined microscopically by the presence of large budded cells and verification using flow cytometry. Cells were then harvested by centrifugation, washed and resuspended in phosphate-buffered saline (PBS, 10 mM phosphate, 0.138 M NaCl; 0.0027 M KCl, pH 7.4). MMS treatment was performed as described in [28] with modification. Cells in PBS were incubated with 11.8 mM (0.1%) MMS for 15 or 30 min at 30°C with vigorous shaking, and then neutralized by mixing 1∶1 (v/v) ratio with 10% Na2S2O3. After washing with dH20, a portion of the MMS-treated cells was immediately resuspended in ice-cold cell suspension buffer (10 mM Tris (pH 8.0), 100 mM EDTA) to prepare DNA-agarose plug for pulsed-field gel electrophoresis (PFGE) as described below. Other portions of the MMS-treated and control cells were resuspended in YPDA media containing nocodazole and incubated at 30°C with constant shaking. Cells were collected up to 16 hours after MMS treatment, centrifuged, wash with dH20 and resuspended in cell suspension buffer for PFGE DNA-agarose plug preparation.
Nocodazole-arrested G2 cells were first incubated with MX (final concentration 100 mM) in YPDA for 15–30 min to first allow MX diffuse into cells. Then MMS treatment and post-treatment incubation were as described above with MX (final concentration 100 mM) present during the whole procedure. Cells were then collected at various times for plug preparation and PFGE analysis.
Detection of DSBs and repair intermediates (such as resected DNA molecule) were based on PFGE analysis as described [28]. PFGE was performed using a Bio-Rad CHEF-Mapper XA system (Bio-Rad, Hercules, CA). Preparation of agarose-embedded DNA (DNA plug) was described in [28]. Briefly, control and MMS-treated cells collected at different times following MMS treatment were embedded in 0.6% agarose with 1 mg/ml Zymolyase (100 U/mg, MP Biochemicals, Solon, OH). The plug was incubated for 1 h at 30°C in a “spheroplasting” solution (1 M sorbitol, 20 mM EDTA, 10 mM Tris pH 7.5) to remove the cell wall. This was followed by digestion with proteinase K (10 mM Tris, pH 8.0, 100 mM EDTA, 1.0% N-lauroylsarcosine, 0.2% sodium deoxycholate, 1 mg/ml proteinase K) for 24 hours at 30°C. Increasing the time of proteinase treatment did not influence the DNA mobility characteristics on PFGE. The parameters for CHEF gel separation of yeast chromosomes in a 1% agarose gel were 6 V/cm for 24 hours with a 10–90 sec switch time ramp and 120° switch angle (running buffer at the 14°C). Subsequently, the DNA was analyzed by Southern blotting as described in [28]. Hybridization was carried out with a probe for the CHAI gene to detect specifically Chr III material or a probe to the LEU2 gene that marked both Chr III and Chr II. Autoradiographs were digitized and densitometric analysis was performed using Kodak MI software (version 5.0).
DNA was digested in agarose plugs with T7 endonuclease I (New England Biolabs, Beverly, MA). A 50 µl plug slice was equilibrated 3 times for 20 minutes at room temperature in 150 µl of TE (10 mM Tris, pH 7.4, 1 mM EDTA), followed by 30 minute incubation at room temperature with 30 units T7 endonuclease in 150 µL reaction buffer, and stopped by washing 3 times with ice-cold Tris-EDTA (10 mM Tris, 50 mM EDTA, pH 8.0). PFGE analysis was performed as described above.
|
10.1371/journal.pgen.1007496 | Specification of Drosophila neuropeptidergic neurons by the splicing component brr2 | During embryonic development, a number of genetic cues act to generate neuronal diversity. While intrinsic transcriptional cascades are well-known to control neuronal sub-type cell fate, the target cells can also provide critical input to specific neuronal cell fates. Such signals, denoted retrograde signals, are known to provide critical survival cues for neurons, but have also been found to trigger terminal differentiation of neurons. One salient example of such target-derived instructive signals pertains to the specification of the Drosophila FMRFamide neuropeptide neurons, the Tv4 neurons of the ventral nerve cord. Tv4 neurons receive a BMP signal from their target cells, which acts as the final trigger to activate the FMRFa gene. A recent FMRFa-eGFP genetic screen identified several genes involved in Tv4 specification, two of which encode components of the U5 subunit of the spliceosome: brr2 (l(3)72Ab) and Prp8. In this study, we focus on the role of RNA processing during target-derived signaling. We found that brr2 and Prp8 play crucial roles in controlling the expression of the FMRFa neuropeptide specifically in six neurons of the VNC (Tv4 neurons). Detailed analysis of brr2 revealed that this control is executed by two independent mechanisms, both of which are required for the activation of the BMP retrograde signaling pathway in Tv4 neurons: (1) Proper axonal pathfinding to the target tissue in order to receive the BMP ligand. (2) Proper RNA splicing of two genes in the BMP pathway: the thickveins (tkv) gene, encoding a BMP receptor subunit, and the Medea gene, encoding a co-Smad. These results reveal involvement of specific RNA processing in diversifying neuronal identity within the central nervous system.
| The nervous system displays daunting cellular diversity, largely generated through complex regulatory input operating on stem cells and their neural lineages during development. Most of the reported mechanisms acting to generate neural diversity pertain to transcriptional regulation. In contrast, little is known regarding the post-transcriptional mechanisms involved. Here, we use a specific group of neurons, Apterous neurons, in the ventral nerve cord of Drosophila melanogaster as our model, to analyze the function of two essential components of the spliceosome; Brr2 and Prp8. Apterous neurons require a BMP retrograde signal for terminal differentiation, and we find that brr2 and Prp8 play crucial roles during this process. brr2 is critical for two independent events; axon pathfinding and BMP signaling, both of which are required for the activation of the retrograde signaling pathway necessary for Apterous neurons. These results identify a post-transcriptional mechanism as key for specifying neuronal identity, by ensuring the execution of a retrograde signal.
| While transcriptional networks and signalling pathways have been a primary focus during the neural specification processes, relatively little is known about how post-transcriptional modulations control the identity of individual cells. Splicing mechanisms have been widely studied and are known to play an essential role, not only during general mRNA processing, but also as a a regulatory mechanism for generating cellular diversity during development [1–4]. Alternative splicing plays a major part in the generation of protein variation, through mRNA isoform generation, and can also act as a regulatory element, by establishing splicing patterns of batteries of genes specific to certain cell types, tissues or even organisms[5]. Salient examples of alternate splicing patterns stem from Drosophila, where sex determination is regulated by dimorphic splicing of pre-mRNAs for Sex lethal, transformer, doublesex and fruitless [6]. In addition, splicing factors can specify identity in a tissue, such as the splicing factor embryonic lethal abnormal vision (elav), which promotes the production of the Ngr180 neuroglia-specific isoform in the nervous system, as opposed to the ubiquitous Ngr167 isoform [7]. However, to what extent splicing determines specific cell identities is not well understood.
The Drosophila Ap cluster is an excellent model for studying mechanisms involved in neural specification in the central nervous system (CNS). The Ap cluster is a well-studied subset of four neurons located laterally in each of the thoracic hemisegments (T1-T3), readily identifiable by the expression of the factors Ap and Eyes absent (Eya). All four Ap neurons are part of the last-born cells of the NB5-6T neuroblast, within which extensive regulatory cascades have been shown to determine the specific fate of each of the neurons of this cluster. These include the Tv1 and Tv4 neuropeptidergic neurons (expressing Nplp1 and FMRFamide, respectively), and also the Tv2 and Tv3 neurons [8–19]. Intriguingly, the Tv4 neuron requires a target-derived retrograde signal, mediated by BMP ligand Glass bottom boat (Gbb) activation of BMP signaling, to terminally differentiate and activate the FMRFa gene [11, 20]. This makes FMRFa expression in Tv4 neurons an excellent model for addressing the molecular genetic aspects of retrograde instructive signals during nervous system development.
A recent genetic screen looking for genes involved in Tv4 specification, using an FMRFa-eGFP transgene, identified the genes l(3)72Ab (FlyBase denomination; herein referred to as brr2) and Prp8 [21]. brr2 encodes a small ribonucleoprotein (snRNP) type DExD/H Box ATPase with helicase activity, while Prp8 encodes a PROCT domain protein, both of which associate with the snRNAs U5 complex and are principal components of the spliceosome [22]. Apart from their well-established role in splicing, most notably in yeast, very little is known about Brr2 and Prp8 function. The human ortholog of brr2, BRR2 has been described to be involved in retinitis pigmentosa disease (RP) [23, 24], but no role in cell specification.
Here, we address the function of brr2 and Prp8, focusing on brr2 and its putative role in RNA processing in the acquisition of Tv4 neuronal fate. Our results conclude that brr2 plays a key and highly specific role controlling the specification of the Tv4 neuron identity. This control is fulfilled by two mechanisms that lead to the activation of the BMP retrograde signaling pathway: (1) correct axonal pathfinding to the target tissue, essential to receive the BMP ligand and, (2) specific RNA processing of the transcripts encoding the BMP receptor subunit Thickveins (Tkv) and Medea, part of the Smad complex, both essential for TGF-β receptor signaling pathway activity. These data demonstrate that specific RNA processing plays a key role in ensuring proper TGFβ/BMP target-derived signalling, and extends our knowledge of the post-transcriptional mechanisms involved in diversifying neuronal identity.
Previously, an EMS forward genetic screen using an FMRFa-eGFP transgene, a marker for the restricted expression of proFMRFa in the Tv4 neuron, was performed to find key genes involved in the NB5-6T lineage specification. This screen resulted in the identification of several mutants displaying specification defects in the NB5-6T lineage, including two nonsense alleles for the brr2 gene (FlyBase l(3)72Ab), brr209C117 and brr213A036, and also one allele of Prp8 [21]. Here, we focus on the role of brr2 in the NB5-6T Apterous cluster (Ap cluster) specification at Drosophila embryonic stage AFT (Air-Filled Trachea, i.e. 18h after egg laying).
Initially, we analyzed expression of the terminal markers, proFMRFa and Nplp1 neuropeptides, in brr2 mutants. These neuropeptides show a highly restricted expression pattern in the VNC. ProFMRFa is expressed in the Tv4 neuron of the Ap cluster (T1-T3) and in two SE2 neurons, while Nplp1 is expressed in the Tv1 neuron of the thoracic Ap cluster (T1-T3) and in the dorsal medial row Ap neurons (dAp neurons, T1-A10; summarized in Fig 1A). Antibody staining of proFMRFa in brr2 mutants revealed a complete lack of expression of proFMRFa in the Tv4 neurons of the Ap clusters (Fig 1B–1D). Expression of proFMRFa was not affected in the SE2 cells, revealing a restricted function of brr2 to the Ap cluster. Similarly, antibody staining of Nplp1 in brr2 mutants revealed the loss of Nplp1 expression in the Tv1 Ap cluster cells, but not in dAp cells (Fig 1C and 1D). The complete absence of both terminal markers and the restriction of the phenotype to the Ap cluster cells in the VNC led us to postulate that Ap cluster cells are not generated, perhaps due to premature cell death or early defects in the NB5-6T lineage. To test this, we analyzed the Ap cluster cells by immunostaining for Eyes absent (Eya), which is present in the four Ap cluster cells at St16 [12]. Eya immunostaining revealed that all Ap cells were present in brr2 mutants at this stage (Fig 2A–2J).
We conclude that brr2 mutants generate Ap cluster neurons, but expression of the terminal differentiation markers proFMRFa and Nplp1 is specifically affected. This phenotype is not observed to other proFMRFa- or Nplp1-expressing neurons (SE2 or dAp cells), which suggests a specific role for brr2 in the determination of neuropeptidergic identity in Ap cluster neurons.
The identity of each cell of the Ap cluster is achieved through an orchestrated expression of transcription factors [8–19]. In order to study the onset of these complex genetic cascades in brr2 mutants, we performed immunostaining at St16 for key transcription factors, Cas, Col, Grh, Sqz, and Nab. At this stage, Ap cluster cells have not acquired their final identity but the combinatorial expression of these factors allow for the identification of Tv1-Tv4 cells during their specification and early differentiation [8, 13]. We found no apparent alteration in the expression of Cas, Col, Grh, Sqz, or Nab in the Ap cluster cells in brr2 mutants (Fig 2A–2J). These results revealed that the NB 5-6T lineage progresses normally and correctly generates each of the Ap cluster cells.
Next, to further resolve the brr2 phenotype, we analyzed the expression of several transcription factors that act during late stages of Ap neuron differentiation, and are necessary for Tv1 and Tv4 neuropeptidergic identities [13]. In the Tv1 neuron, Ap and Eya activate the expression of Dimmed (Dimm), and all three genes, together with Col, trigger Nplp1 expression [13]. In the Tv4 neuron, Ap and Eya also activate Dimm that, together with other two factors, the phosphorylated form of Mothers against dpp (pMad) and Dachshund (Dac), trigger the expression of proFMRFa [11–13]. Therefore, we analyzed the expression of these factors in the Ap cluster cells (Fig 3A–3H).
We did not find differences in expression of either Ap, Eya, Col, and Dac between control and brr2 mutant embryos (Fig 3A–3H). In contrast, the expression of Dimm was lost in one of the two Dimm-positive cells (Tv1 and Tv4) in all Ap clusters (Fig 3D). Taking advantage of Col immunostaining, which is restricted to Tv1 neurons within the Ap cluster at the AFT stage, we found that Dimm was selectively lost in 100% of Tv4 neurons, in brr2 mutants (n = 24 Ap clusters) (Fig 3I and 3J).
We previously reported that Svp represses Dimm expression in Tv2 and Tv3 neurons [25], and the timed reduction of Svp expression in Tv1 and Tv4 de-represses Dimm and neuropeptide expression in those neurons by Stg17 [25,26]. Hence, mis-regulation of Svp may explain the absence of Dimm in the Tv4 neuron. However, immunostaining did not reveal differences in Svp expression (Fig 3K and 3L).
In addition to the partial loss of Dimm, pMad expression was almost completely abolished in brr2 mutants (97% n = 49; p<0,001) (Fig 3F). Segments T1, T2 and T3 were similarly affected. However, pMad staining could be still detected in other cells outside the Ap cluster (S1A–S1F Fig), demonstrating that there was no general defect on the phosphorylation of pMad. To further address if the BMP signaling defect observed in brr2 mutants was limited to the Ap cluster we analyzed another peptidergic neuronal subtype, the Insulin-like peptide 7 (Ilp7) neurons, which also require BMP signalling for the proper Ilp7 expression [27]. Surprisingly, both Ilp7 and pMad expression was completely unaffected in these neurons in brr2 mutants (S2B and S2D Fig). Additionally, we quantified the total number of pMad-positive cells in thoracic and abdominal segments in brr2 mutants and did not find significant differences when compared to control (S1A–S1F Fig).
Our results demonstrate that brr2 is necessary for the expression of pMad, Dimm and proFMRFa in Tv4 neurons, and for Nplp1 in Tv1 neurons.
Tv4 neurons project their axons towards the midline and exit the VNC at the dorsal midline, to innervate a peripheral secretory gland; the dorsal neurohemal organ (DNH). When the axon reaches the DNH, it receives the TGFβ/BMP ligand Glass bottom boat (Gbb), which results in the phosphorylation of Mad (pMad), triggering the expression of the FMRFa neuropeptide gene[11–20]. Therefore, the absence of pMad in brr2 mutants could be due to defects at different points of this process, including problems with BMP signaling, as well as axon pathfinding problems.
Because proFMRFa is absent in brr2 mutants, the lack of markers prevented us from identifying the Tv4 neuron and follow its axon trajectory. Thus, in order to address possible defects in the axonal pathfinding of Tv4 neurons during its innervation of the DNH, we monitored the axon pathfinding of the Ap cluster through the expression of CD8::GFP under the promotor of ap gene, using the Gal4-UAS system [28] in combination with the reporter buttonless-lacZ (btn-LacZ), which identifies the DNH. 3D analysis of this staining revealed that while Tv4 always reached the DNH in controls it failed to do so in brr2 mutants (Fig 4A and 4B and S2 Video). These results indicate that in brr2 mutants, Tv4 axons fail to receive the Gbb ligand and therefore fail to phosphorylate Mad.
Previous studies described the failure of Mad phosphorylation upon a failure of axonal pathfinding to the DNH, in sqz mutants and upon misexpression of robo in Ap-neurons [11]. In both cases, ectopic expression of gbb in Tv4 rescued Mad phosphorylation and proFMRFa expression. In order to elucidate if BMP signaling is not triggered in brr2 mutants merely due to defective axonal pathfinding, we ectopically expressed gbb in the Ap cluster of brr2 mutants using the ap-Gal4 driver (ap>gbb), and performed immunostaining of pMad and proFMRFa (Fig 4I–4P). Surprisingly, we found that expression of Gbb in the Ap neurons failed to restore pMad or proFMRFa expression in brr2 mutants. This suggested that expression and/or processing of other components in the BMP signaling pathway may be affected.
The BMP signaling pathway is triggered by binding of Gbb to a hetero-tetrameric receptor complex, formed by Wishful thinking (Wit), Saxophone (Sax) and Thickveins (Tkv), and Tkv/Sax-kinase activity to phosphorylate Mad, which is transported to the nucleus to act as a transcription factor, in a complex with the co-Smad, Medea [29]. To address if BMP receptor expression/assembly was affected in brr2 mutants, we ectopically expressed constitutively active forms of Sax (UAS-saxA) and Tkv (UAS-tkvA) [30], previously shown to restore proFMRFa expression in mutants of BMP signaling pathway receptors [11,31]. To this end, we used two different drivers: elav-Gal4 (elav>saxA, tkvA; brr2) and ap-Gal4 (ap-Gal4> saxA, tkvA; brr2). We observed a partial rescue of brr2 mutants in both cases (29.8% n = 84 p<0.001 and 29.8% n = 47 P>0.001, respectively) (Fig 5A–5D). This suggested that defects in receptor expression/assembly may in part underlie the brr2 phenotype. Nevertheless, the rescue was less robust than anticipated. This could potentially explained by FMRFa mRNA processing defects, partial rescue of Mad phosphorylation or perhaps Dimm expression.
To confirm that brr2 affected transcription of the FMRFa gene, as opposed to the processing and translation of the FMRFa mRNA, we analyzed the expression of the FMRFa-eGFP reporter gene; a construct with the previously identified FMRFa Tv enhancer of 446 bp inserted upstream of eGFP [21, 32]. We found that expression of eGPF was lost hand-in-hand with the proFMRFa immunostaining (Fig 5B).
We postulated that the lack of Dimm, observed in Tv4 neurons in brr2 mutants, may underlie the partiality of the rescue. However, further analysis of the rescue did not reveal any correlation between the presence of Dimm and the rescue of proFMRFa (S4D and S4E Fig). Indeed, we observed partial rescue of Dimm expression in ap-Gal4> saxA, tkvA; brr2 (63% n = 82 clusters p>0,001, S4F Fig). With regards pMad, we found that pMad was rescued in the majority of Tv4 neurons, in ap-Gal4> saxA, tkvA; brr2 embryos (89% n = 47 P> 0,001, S4I Fig). Thus, a lack of pMad rescue did not fully account for the partial proFMRFa rescue.
The rescue of proFMRFa expression in a subset of Tv4 neurons, in ap-Gal4>saxA, tkvA; brr2 embryos, allowed us to track those axons in brr2 mutants. We found that the Tv4 axon initially projects normally towards the midline, but subsequently fails to project dorsally into the DNH. Instead, it follows the same trajectory as the rest of the Ap cluster neurons, projecting ipsilaterally and anteriorly (Fig 5F–5K, S3 and S4 Videos).
Our results reveal that brr2 has a crucial role in the specification of Tv4 identity through both axonal pathfinding and BMP signaling, both of which are essential for proFMRFa neuropeptide expression.
Brr2 is an important component of the U5 subunit of the spliceosome. To address whether the defects observed in brr2 mutants were indeed related to a defect in splicing activity, we analyzed Prp8, which encodes another component of the U5 subunit of the spliceosome and has been described as a regulator of Brr2 in mRNA processing [22, 33–35]. Prp8 was also identified in the FMRFa-eGFP genetic screen as a gene showing loss of eGFP expression [21]. If the phenotype of brr2 in Tv4 neurons is due to its splicing activity, we expected to find a phenocopy in Prp8 mutants. To this end, we performed immunostaining of proFMRFa, Dimm and pMad in Prp8 mutants at AFT. Our results revealed an absence of proFMRFa specifically in Ap cluster (S6A–S6C Table). The loss of Dimm and pMad expression in Prp8 mutants resembled that found in brr2 mutants (S5 Fig).
The phenocopy of brr2 and Prp8 mutants suggested that Tv4 cell differentiation defects result from their roles as components of the spliceosome. The selective effect of brr2 and Prp8 on Ap neurons stands in contrast to their described roles as essential and ubiquitous components of the spliceosome [36, 37] (Figs 1C and 5). This prompted us to attempt to unravel which specific genes were responsible of the defective differentiation of the Ap cluster neurons. To this end we performed RNA-Seq analysis of RNA extracted from control, brr2 and Prp8 St15-AFT embryos. The RNA-Seq data were analyzed both for splicing aberrations and for gene expression changes. Focusing first on splicing aberrations, our results (S1–S5 Tables) revealed 1338 significantly altered splicing events (|ΔΨ| > 0.03 and q < 0.25) in brr2 and 855 Prp8, with 512 being common to both mutants (Fig 6D and 6E; S1 Table). Among the common genes identified, only the BMP receptor tkv is known to play a role in Ap cluster specification. Specifically, we found that tkv shows defective splicing involving an alternative first exon (AFE) in both brr2 and Prp8 mutants (|ΔΨ| = 0.03163; qvalue = 0.1448; pvalue = 0.0234). In the case of brr2, the AFE path E5-E6 is favored, which is predicted to generate the shorter protein isoform tkv-PD (509 aa). The PD isoform generates a protein with unaltered Tkv transmemebrane and kinase domains but lacking the signal peptide (Fig 6F). This protein isoform is less efficient in its ability to bind ligand and may be involved in dosage-dependent tkv-ligand interactions [38]. The RNA-seq results showed that isoform tkv-PA was still present in brr2 mutants, but since the RNA was extracted from whole embryo it remains unclear whether or not Tv4 neurons express both isoforms.
In addition, in brr2 mutants, but not in Prp8, we also found defective splicing of Medea. Medea encodes an ortholog of the co-smad Smad4, which complexes with phosphorylated Receptor Smads (Smad1,5,9 in vertebrates and Mad in Drosophila), to form the Smad complex which translocates to the nucleus to act as transcriptional regulator. In the case of brr2 mutants, the Exon Skipped (ES) E4 is favored, which generates the Med-RB isoform where the fourth exon is absent. In vitro studies have shown that the presence or absence of the alternative fourth exon does not affect Medea binding activity [39]. Nevertheless, the behavior of Medea isoforms in vivo is unknown. Medea amorphic alleles that affect all isoforms eliminate proFMRFa expression in Tv4 neurons but roles for specific isoforms have not been rigorously discriminated (personal communication Anthony Berndt and Douglas Allan).
In order to determine if pMad is properly transported into the nucleus, we analyzed the co-localization of pMad antibody and the nuclear marker DAPI in ap-Gal4>saxA, tkvA; brr2 genetic background. We found an apparently unaltered nuclear localization of pMad (S4G and S4H Fig). Nevertheless, we cannot rule out that the mis-spliced Medea variant affects DNA-binding efficiency, and hence, defective splicing of Medea could be contributing to the lack of proFMRFa found in brr2 mutants. Indeed, the limited efficiency of the genetic rescue suggests a contribution of Medea to the brr2 phenotype.
We performed Gene Ontology (GO) enrichment analysis in order to elucidate if brr2 mutant splicing defects are more prevalent in neuronal specification due to the observed effect upon proFMRFa. However, our results revealed that brr2 mutants were affected in many processes and molecular functions, albeit with some prevalence for nuclear components, cytoplasm and plasma membrane components (S6 Fig). Regarding axonal guidance related genes (GO 0007411), 57 genes showed significant splicing events (S7B Fig, S7 Table), which may explain the brr2 Tv4 axon pathfinding defects.
In addition to the splicing events detected, transcriptome analysis by RNA-Seq revealed 3,780 genes with 2-fold changed expression (up- or down-regulated) in brr2 mutants (Fig 6D, S7 Table). From the genes related to BMP signaling pathway, only screw was affected. Regarding axonal guidance, 8 genes were up or downregulated in brr2 mutants (S7B Fig, S7 Table).
Summarizing, the splicing defects observed in brr2 mutants lead to an increase of the tkv-PD spliced form, which is less efficient with respect to the binding of the ligand. This would hamper receptor assembly, the phosphorylation of Mad and therefore the specification of Tv4 neuron. This result agrees with our previous observations, in particular the loss of pMad staining and the partial rescue of brr2 mutants by expression of saxA and tkvA but not with gbb.
In this study, we describe a cell-specific role of Brr2 in controlling the specification of the neuropeptidergic identity of the Tv4 neuron in the embryonic Drosophila Central Nervous System (CNS). This control is achieved through its role in three different steps necessary for the Tv4 neuron specification: (1) the axon pathfinding; (2) Dimm expression; and (3) activation of the BMP signalling. The role of Brr2 in the spliceosome complex has been widely studied and it is known to be essential for proper RNA processing[22, 40]. Brr2 has also been identified as a fundamental component of the alternative splicing process in Drosophila [17], and in addition, brr2 expression is ubiquitous in Drosophila [36, 37]. These findings prompted us to anticipate a widely defective splicing pattern. However, our RNA-Seq analysis revealed that a limited number of genes were affected. Furthermore, our genetic analysis show that while brr2 is necessary for proFMRFa and Nplp1 expression in the Ap cluster, expression of these neuropeptides in other cells in the VNC (SE2 cells for proFMRFa and dAp for Nplp1) was not affected. In addition, after a detailed analysis of pMad expression along the VNC of brr2 mutants, we are not able to find differences with respect to the wild type pMad expression. This set of results demonstrates that the roles of ubiquitous components of the spliceosome can be highly cell-type selective.
Why Brr2 activity is restricted to promoting BMP signalling only in the Tv4 neuron of the Ap cluster, but not in other areas of the VNC, remains unclear. tkv and Medea are key components of the BMP signalling pathway and act in a number of neurons projecting out of the CNS. However, the number of pMad-positive neurons is not affected in brr2 mutants. Protein-protein interaction has been described between Eya and Brr2 in a co-immunoprecipitation analysis in Drosophila [41]. The same study also described a similar interaction between Prp8 and Ap. It is tempting to speculate that these interactions could allow specific recruitment/activity of the spliceosome complex in Ap cluster cells, granting efficiency and specificity to this component of the spliceosome. The association between transcription and splicing is frequent and increases expression efficiency [42, 43]. Examples of these phenomena can be found in Drosophila Mef2 transcription factor family, where integration of both processes is involved in muscle fibre specification [2]. Moreover, cell-specific splicing events have been previously reported in the Drosophila nervous system, for instance with regards to splicing of the Dscam gene [44–46].
Although the role of brr2 in the Tv4 neuron has been delimitated with high resolution in the case of the post-transcriptional effects upon tkv and Medea, the underlying mechanisms in the axonal pathfinding defects and the lack of Dimm expression observed in Tv4 neurons remains unresolved. Regarding the pathfinding phenotype, the RNA-Seq analysis in this study revealed significant changes in 8 genes and splicing defects in 57 genes related to axon guidance. The specific role, if any, of each of these genes in the Ap cluster is unknown. Nevertheless, our study has identified clear candidate genes which, together with the well-described axonal pathfinding of Tv4, provide an excellent starting point for studying axonal guidance. Using specific mutant backgrounds and co-expression of different splicing variants identified in this study, it would be possible to shed light into the function of those genes, not only at the expression level but with regards to different isoforms, as have been previously described for the axon pathfinding gene lola [47, 48]. With respect to Dimm, our RNA-Seq analysis did not identify any defects in known genes which could explain the lack of its expression, which suggest that further analysis is required to identify more genes related to Nplp1 and proFMRFa neuropeptide expression.
The low efficiency of the rescue mediated via expression of active forms of tkv and sax (apterous-Gal4>saxA, tkvA) suggests that Brr2 could be regulating additional components downstream of pMad. The nuclear localization of pMad in a subset of cells where proFMRFa was not restored suggests that the splicing defects in Medea are not interfering with its function, at least with respect to its translocation to the nucleus. Indeed, these results match those from in vitro studies wherein this mRNA isoform encodes a Medea protein with apparently normal DNA-binding activity [39]. Nevertheless, its behavior in vivo is unknown and we cannot rule out that changes in the Medea splicing may affect the function of the Med/pMad complex in other ways, maybe related to cofactor binding as well as DNA-binding.
The specification of Tv4 neurons requires an elaborate gene cascade, which converges to activate the expression of the FMRFa neuropeptide gene [8, 10–14, 32]. One of the necessary components in this cascade is retrograde BMP signalling, and a number of events are necessary to activate the BMP signalling in the Tv4 neuron. Here, we demonstrate how the brr2 gene regulates this process at three different points: Control of pathfinding of the Tv4 axon, which is necessary for BMP ligand reception and regulation of proper splicing of two crucial players of the BMP signalling pathway, tkv and Medea.
In summary, here we report post-transcriptional mechanisms that sculpt neural architecture at a cell-specific level, controlling precise aspects of cell development that together define precise and discrete cell identities. This increases our understanding of how alternative splicing is utilized during neural development. Future studies will be needed to determine if similar cell-specific splicing mechanisms is utilized in other lineages contributing to neural diversity, and emerging new technologies and bioinformatics for RNA studies will facilitate this work.
UAS-myr-mRFP, FMRFa-eGFP, brr209c117; UAS-myr-mRFP, FMRFa-eGFP, brr13A036; UAS-myr-mRFP, FMRFa-eGFP, Prp804p024 [21]. From Bloomington Drosophila Stock Center: apmd544 (referred to as apGal4, BL#3041), elav-GAL4 2 (referred to as elav-Gal4, BL#8765), UAS-mCD8GFP (BL#8746). Other stocks used: UAS-gbb, UAS-saxa, UAS-tkva, btn-lacZ [11].
Mutants were maintained over GFP- or YFP-marked balancer chromosomes. As wild type, OregonR or w1118 was used. Staging of embryos was performed according to Campos-Ortega and Hartenstein [49].
Immunohistochemistry was performed as previously described [8]. Antibodies used were: guinea pig anti-Col (1:1000); guinea pig anti-Dimm (1:1000); chicken anti-proNplp1 (1:1000); rabbit anti-proFMRFa (1:1000) [13]. Mouse anti-Seven up (1:50) (Y. Hiromi, National Institute of Genetics, Mishima, Japan). Rabbit anti-pMad (1:500) (41D10, Cell Signaling Technology). Mouse mAb Eya10H6 (1:250) (from Developmental Studies Hybridoma Bank). Rabbit α-Cas (1:250) [50]. Rat α-Sqz (1:750) [51]. Rat α-Grh (1:1000) [8]. Rabbit α-Nab (1:1000) [52]. Rabbit α-Ilp7 (1:1000) (Miguel-Aliaga, Irene, London institute of Medical Sciences, Imperial College London, UK). All polyclonal sera were pre-absorbed against pools of early embryos. Secondary antibodies were conjugated with FITC, Rhodamine-RedX or Cy5 and used at 1:500 (Jackson ImmunoResearch). Embryos were dissected in PBS, fixed for 25 minutes in 4% paraformaldehyde, blocked and processed with antibodies in PBS with 0.2% Triton X-100 and 4% donkey serum. Slides were mounted with Vectashield (Vector Labs). Wild type and mutant embryos were stained and analyzed on the same slide.
Zeiss LSM700 or Zeiss LSM800 confocal microscopes were used for fluorescent images; confocal stacks were merged using LSM software or Fiji software [53]. Images and graphs were compiled in Adobe Illustrator.
Two biological replicates of frozen collections of St15-AFT embryos control (Orizo2), brr209C117 and Prp804P024 (50 mg) were used for RNA extraction, using RNeasy Mini Kit (Qiagen, Hilden, Germany). Sample RNA yield was measured with a NanoDrop, precipitated in ethanol, and then sent to GeneWiz (GeneWiz, New Jersey, NJ) for library preparation and sequencing. Yield was checked, upon receipt of each sample, by use of NanoDrop, Qubit RNA Assay and Agilent Bioanalyzer. The samples were fragmented after RNA QC, reverse transcribed with random primers, and barcode tagged. Sequencing was performed by on the Illumina Hiseq 2500, in 1x50 bp single-read sequencing configuration (the output was stored as FASTQ-files,), which yielded 31–38 million reads/sample. The FASTQ-files were aligned against the dmel reference genome (Release 6, RefSeq GCF_000001215.4) and raw reads were normalized as reads per kilobase-length of gene per million mapped sequence reads (RPKM). Differential alternative splicing (DAS) analysis was performed as described [54–57]. Raw RNA-Seq reads were aligned to Drosophila melanogaster (dm3) genome using STAR (version 2.5.1b) [58] with default settings, and only uniquely mapped reads were retained to compute the number of reads for exons and exon-exon junctions in each sample using the Python package HTSeq [59], with the annotation of the UCSC Ensembl gene annotation (dm3_ensGene)[60]. Dirichlet-multinomial distribution was used to formulate the counts of the reads that were aligned to each isoform of each event [61], and the likelihood ratio test was used to test the significance of the changes in alternative splicing between brr2 or Prp8 mutants and controls [62]. The Benjamini-Hochberg procedure was applied to calculate the adjusted q-values from the p-values in the likelihood ratio test [63]. Seven types of differential alternative splicing events are tested, including Exon skipping (ES), Alternative 5' splice sites (A5SS), Alternative 3' splice sites (A3SS), Mutually exclusive exons (ME), Intron retention (IR), Alternative first exons (AFE) and Alternative last exons (ALE) [64]. In addition, Percent Spliced In (PSI, Ψ) was first computed to examine the percentage of the inclusion of variable exons in the exon-skipping events compared to the both isoforms [65], which is extended to evaluate splicing changes of all the seven splicing types in our DAS analysis. Particularly, the PSI was calculated as the percentage usage of the longer isoform compared to total mRNAs for ES, A5SS, A3SS, ME and IR. However, the PSI was calculated as the percentage of the usage of the proximal isoform, where the isoform with the variable exon is closer to the constitutive exon, for AFE and ALE. The differential alternative splicing events were identified under |ΔΨ| > 0.03 and q < 0.25.
Gene Ontology enrichment analysis was performed using GeneCodis3 for GO Biological process, GO Molecular function, and Go cellular component; lowest annotation levels, Hypergeometric statistical test, FDR P-value correction [66–68].
Pearson’s chi-squared test, Monte Carlo, Student’s t-test, were performed using SPSS v.23 (IBM; for specific statistical test used, see text and figures). Microsoft Excel 2010 was used for data compilation and graphical representation.
|
10.1371/journal.pntd.0006657 | Stratified sero-prevalence revealed overall high disease burden of dengue but suboptimal immunity in younger age groups in Pune, India | In India, dengue disease is emerging as the most important vector borne public health problem due to rapid and unplanned urbanization, high human density and week management of the disease. Clinical cases are grossly underreported and not much information is available on prevalence and incidence of the disease.
A cross sectional, stratified, facility based, multistage cluster sampling was conducted between May 4 and June 27, 2017 in Pune city. A total of 1,434 participants were enrolled. The serum samples were tested for detection of historical dengue IgG antibodies by ELISA using the commercial Panbio Dengue IgG Indirect ELISA kit. Anti-dengue IgG-capture Panbio ELISA was used for detection of high titered antibodies to detect recent secondary infection. We used this data to estimate key transmission parameters like force of infection and basic reproductive number. A subset of 120 indirect ELISA positive samples was also tested for Plaque Reduction Neutralizing Antibodies for determining serotype-specific prevalence.
Overall, 81% participants were infected with dengue virus (DENV) at least once if not more. The positivity was significantly different in different age groups. All the adults above 70 years were positive for DENV antibodies. Over 69% participants were positive for neutralizing antibodies against all 4 serotypes suggesting intense transmission of all DENV serotypes in Pune. Age-specific seroprevalence was consistent with long-term, endemic circulation of DENV. There was an increasing trend with age, from 21.6% among <36 months to 59.4% in age group 10–12 years. We estimate that 8.68% of the susceptible population gets infected by DENV each year resulting into more than 3,00,000 infections and about 47,000 to 59,000 cases per year. This transmission intensity is similar to that reported from other known hyper-endemic settings in Southeast Asia and the Americas but significantly lower than report from Chennai.
Our study suggests that Pune city has high disease burden, all 4 serotypes are circulating, significant spatial heterogeneity in seroprevalence and suboptimal immunity in younger age groups. This would allow informed decisions to be made on management of dengue and introduction of upcoming dengue vaccines in the city.
| Dengue disease, transmitted through the bite of DENV infected mosquitoes, is an increasing health problem in the Asian subcontinent, including India. Dengue ranges from mild undifferentiated fever to circulatory shock and potentially death. Clinical disease gives an incomplete picture of the magnitude of dengue, because many infections are asymptomatic. Presence of antibodies to DENV provides evidence of past infection. This study provides the first estimate of the prevalence and incidence of dengue, based on the data collected from a well-designed, comprehensive serosurvey. By studying age–wise antibody prevalence, we estimated the force of DENV infection by applying a catalytic model to our serosurvey data. Over 81% individuals were positive for DENV antibodies suggesting intense DENV transmission in Pune city. We estimate that 8.68% of the susceptible population gets infected by DENV each year resulting into more than 3,00,000 infections and about 47,000 to 59,000 cases per year. The estimated seroprevalence at 9 years age (SP9), taken as benchmark for introduction of Dengvaxia vaccine by WHO, was 54.17% suggesting moderate transmission intensity of dengue, making introduction of the vaccine unsuitable in younger children.
| Dengue disease is an important emerging public health problem in countries of tropical and subtropical regions.[1–3] Estimated annual global burden of disease is approximately 390 million infections, 96 million clinical cases, and 20 thousand deaths, with almost 34% of total dengue cases occurring in India.[4] According to recent estimates, 2·9 million dengue episodes and 5906 deaths, with an economic burden of $950 million occur annually in Southeast Asia (SEA) alone.[5] It is known that disease intensity and disease burden is highly variable between different places within a country or region.[6]
In India, dengue is a reportable disease and all confirmed cases are expected to be reported to government of India through NVDCP, Delhi.[7] Recent studies using various models have suggested gross underreporting of dengue cases. It is estimated that each case reported may be multiplied by 200 to get fair estimate.[8,9]
There are 4 antigenically distinct DENV serotypes (DENV 1–4). Dengue can result from infection with any one of four viral serotypes. Infection with one serotype provides long-term protection to that serotype, but not to others. Thus, DENV seropositive individuals could be monotypic due to primary infection or multitypic due to secondary infections. Presence of certain serotypes, including primary infection with DENV-3 from the SEA region and secondary infection with DENV-2, DENV-3, and DENV-4 also from the SEA region, as well as DENV-2 and DENV-3 from non-SEA regions, increased the risk of severe dengue infections.[10] Thus, age specific distribution for different serotypes and their contributions in monotypic and multitypic cases are worthy of special consideration.
Dengue infection results into subclinical disease in majority of the cases and clinical disease in about 25% cases. Proportions of asymptomatic, mild cases and severe cases are highly variable in different areas. Differential diagnosis between clinically similar diseases caused by DENV, Chikungunya virus and other febrile illnesses is almost impossible in resource limited countries like India. Therefore clinical surveillance data which already suffers with tremendous reporting bias is inadequate to estimate true burden of disease. In such situations, properly designed seroprevalence studies may adequately quantify and characterize the extent of transmission.
Currently there is no effective drug for treatment of dengue. Sustained effective vector control has become impractical in developing countries. Therefore vaccination has become focus of attention in management of dengue. Several vaccines are in different phases of developments and clinical trials. The first live attenuated (recombinant) tetravalent dengue vaccine, Dengvaxia, produced by Sanofi Pasteur, has been licensed for use in some countries in Asia and Latin America. World Health Organization (WHO) Strategic Advisory Group of Experts (SAGE) recommends that countries consider introduction of this dengue vaccine only in populations where epidemiological data indicate a high burden of disease. In order to maximize public health impact and cost effectiveness, the populations to be targeted for vaccination, as measured by seroprevalence, should be approximately 70% or greater in the age group targeted for vaccination.[11] Seroprevalence typically increases with age, and countries may choose to target vaccination to the youngest age (9 years or older) for which seroprevalence exceeds the recommended 70% threshold.[12] Since such data is not available for most of the endemic places in India, well designed serosurveys are recommended to support decision making for vaccine introduction for public health as well as for conducting clinical trials with dengue vaccines.
In view of these concerns, a stratified serosurvey was conducted in Pune city, Maharashtra, India. Pune is fast growing city, chosen under Smart Cities Mission scheme of the Prime Minister of India for speedy and orderly infrastructure development. The city has been experiencing seasonal, annual dengue outbreaks. It is pertinent to generate data on epidemiological determinants including disease burden estimates for proper planning of dengue management.
Pune, the second largest city in the state of Maharashtra after Mumbai and the seventh most populous city in the country is situated 560 meters above sea level on the Deccan plateau. Pune is the administrative headquarters of Pune district and is one of the fastest growing cities in the Asia-Pacific region.
It lies between 18° 32" North latitude and 73° 51" East longitude. Pune is 149 kilometers, southeast of Mumbai by road. Average temperatures ranges between 19 to 33°C. Pune experiences three seasons: summer, monsoon, and winter. Typical summer months are from mid-March to June with maximum temperatures sometimes reaching 42°C. The monsoon lasts from June to October, with moderate rainfall and temperatures ranging from 22 to 28°C. Most of the 722 mm of annual rainfall in the city falls between June and September, and July is the wettest month of the year. In winter, the daytime temperature hovers around 26°C while night temperature is around 10–14°C, sometimes dropping to 5 to 6°C.
The population of the Pune city is 3,124,458 and Pune Urban Agglomeration is 5,057,709 as of the 2011 census.[13] Annual exponential growth rate of population was 2.08 per year (for 2001–2011), with birth rate of 19.3 live births per thousand of population per year.[13,14] In 2017, the estimated population of Pune is 3.99 million.[15] Pune city is divided into 5 administrative zones, having 15 administrative units called wards. Each ward has one or 2 clinics managed by Pune Municipal Corporation, many private clinics managed by General Practitioners, and some tertiary care hospitals.
A cross sectional, stratified, facility based, multistage cluster sampling was conducted in Pune city between May 4 and June 27, 2017, following the principles of WHO guidelines.[12] The dengue season in this area is typically from July to December. The present survey was planned to capture activity of dengue from the previous 2016 dengue season. Medical clinics are the first contact point between febrile cases and health seeking facilities. In all 15 wards, a corporation clinic was chosen as first point for sampling. Additional 3 clinics of general practitioners were chosen in such a manner to provide fair representation to the ward. This ratio was based on assumption that about 25% of the primary healthcare in the city is provided by the corporation clinics and the rest by the private practitioners. Fig 1 shows approximate locations of the collection sites (health facility).
The data on dengue prevalence in Pune city were not available. However, dengue prevalence of 59% was reported from an urbanized village near Pune city.[16] Assuming that prevalence in Pune city will be higher than the adjoining urbanized village, for the purpose of sample size calculations we assumed 65% prevalence in Pune city. The minimum sample size of 1,396 participants was calculated under the assumption of 65% prevalence for dengue infection, α ± 5% error, Confidence level 95%. Accounting for the multistage sampling, the sample size considered a design effect of 4.0. Sample allocation to each ward and age groups was in proportion to the population of the ward and age group with respect to the Pune population. Allowing 5% additional samples to meet contingencies like insufficient sample, leakage and spoilage we targeted 1,465 samples. A team visited each health facility. Each non-febrile patient and/or the person accompanying them visiting the facility and resident of the same ward were invited to participate in the study. The willing persons were enrolled until the target sample collection was achieved for that site. Each enrolled person was requested to provide a blood sample following administration of ethical consent/assent approved by the Institutional Ethics Committee of the University.
We collected blood samples from a total of 1,434 participants, 31 less than the original 1,465 sample target. About 5 mL blood was collected from each participants in anti-coagulant free vacutainer tubes (BD Bioscience) by trained phlebotomists and kept overnight at 4°C. Serum samples were separated by centrifugation at 3,000 rpm for 10 minutes and stored at -80°C.
Each serum sample was tested for dengue IgG antibodies by ELISA using the commercial Panbio Dengue IgG Indirect ELISA kit (Panbio Diagnostics, Brisbane, Australia, Cat no. 01PE30) according to manufacturer’s instructions. The presence of detectable IgG antibodies indicates past exposure to dengue infection. Panbio units were calculated by dividing the sample absorbance by the cut-off value and then multiplying this value by 10. Samples were considered positive if Panbio units were >11, <9 Panbio units were considered negative and if Panbio units were between 9 to 11, samples were considered equivocal and retested to confirm the result.
An anti-dengue IgG-capture ELISA (Panbio Diagnostics, Brisbane, Australia, Cat no. 01PE10) was performed according to the manufacturer’s instructions. Anti-dengue IgG Panbio units were calculated by dividing the sample absorbance by the cut-off value and then multiplying this value by 10. Using this criteria, a value of >22 Panbio units was used to identify secondary infection. <18 Panbio units were considered negative for secondary infection and if Panbio units were between 18 to 22, samples were considered equivocal and retested to confirm the result for secondary infection.[17] High Panbio units are indicative of elevated levels of IgG antibodies which suggest that the patient has been recently exposed to dengue virus due to secondary infection.
As WHO recommends use of PRNT90 titers to minimize serum cross-reactivity with other dengue serotypes and flaviviruses prevalent in DENV endemic areas [12,18], we opted for PRNT90 method for this study. Due to resource constraint, we decided to process 120 indirect ELISA positive samples for PRNT. The selection of samples was based on Panbio units of IgG-positives (Indirect ELISA) arranged at the interval of 5 units and represented comparable proportions of total positives in each category.
We followed WHO guidelines for the PRNT90 test. However, since we were interested in assessing neutralizing antibodies (NAbs) against the currently circulating Indian strains, necessary modifications were made. The DENV strains used were DENV-1 (S19) (Accession no. MG053115), DENV-2 (S15) (Accession no. MG053142), DENV-3 (S111) (Accession no. MG053151) and DENV-4 (1028) (Accession no. MG272272) isolated during 2016 in Pune city [19]. The viruses actually used for PRNT were passaged 4–5 times, titrated using plaque assay and stored at -80°C at smaller aliquots. The test included two controls in duplicates; cell control without any virus or serum and virus control for different serotypes, without serum were used in the assay. For the test, early passage Vero cells (CCL-81, ATCC) were seeded at the density of 1 x 105 cells/mL in Minimum Essential Medium (MEM) (GIBCO) with 10% Fetal Bovine Serum (FBS, GIBCO) in 24-well plate (1mL/well). The following day, serum samples (diluted 1:5 in MEM with 2% FBS) were heat inactivated at 56°C for 30 min and then serially diluted 4-fold in the same diluent in 96-well microtiter plates. Serially diluted serum samples were mixed with an equal volume i.e, 1:2 of diluted virus that gives 40–100 plaques/control well with each serotype. The final serum dilutions were 1:10 to 1:2560. After incubation for 1 hr at 37°C, 5% CO2 incubator, the medium was removed from 24-well plate and 100µl of each dilution of serum/virus mixture was added onto the cells in duplicate. The plates were then incubated for 1 hr for DENV-1, 2, 3 and 2 hr for DENV-4 at 37°C, 5% CO2 incubator to allow virus adsorption. After adsorption, 1ml of overlay media containing 1% Aquacide-II (Calbiochem) were added onto the cells and incubated for 3 days at 37°C, 5% CO2 incubator. Three days post infection, the overlay medium was discarded from the plates, and the cell monolayer was fixed with formalin for 30minutes at RT and permeabilized with 0.2% Triton X-100 in PBS for 5min. The cells were washed three times with PBS-T (0.02% Tween-20 in PBS) and stained with HB112 pan-flavivirus mouse monoclonal antibody (D1-4G2-4-15, ATCC) at 1:500 dilution in PBS for 2 hr. Cells were washed three times with PBS-T and incubated with goat anti-mouse IgG horseradish peroxidase (HRP) at 1:1500 dilution in PBS for 1 hr. After washing three times with PBS-T and two times with PBS, cells were stained with True Blue peroxidase substrate (KPL, Sera Care, MA, USA) and blue color staining of virus infected cells were counted as plaques. PRNT90 titer was calculated using NIH LID Statistical Web tool.[20] PRNT90 titer ≥ 1:10 to one dengue serotype at least was considered seropositive. A monotypic response was defined by the presence of NAbs against only one of the four DENV serotypes. A multitypic response was defined as a concomitant detection of NAbs against more than 1 serotype.
Statistical analyses were performed using ‘R’ Version 3.4.1., Microsoft windows Excel 2010, SPSS v. 17.0 (SPSS Inc.,USA) and Graphpad Prism v.7.0 (Graphpad Software USA).[21] The logarithm (Log10) values of antibody titers of the serotypes were used for analysis and graphical representation. The statistical comparison of the means of the antibody titers of the serotypes was performed using analysis of variance (ANOVA). The association between the numbers of DENV serotypes (one, two and three simultaneous serotype infections) and mean age was performed using ANOVA with POST HOC Least significant difference (LSD) test, whereas their association with gender was tested through χ2 (chi-square) test for trend. Mann-Whitney U test was performed to check the association of PRNT90 titers against all 4 serotypes across different age groups.
Studies were conducted at Interactive Research School for Health Affairs (IRSHA), a constituent unit of Bharati Vidyapeeth (deemed to be University), Pune. The study was approved by the Institutional Ethics Committee (IEC/2017/04). Written consent/assent to participate in the study, reviewed and approved by the Ethics committee, was administered to each participant or to their legal guardian. All data were handled anonymously and confidentially.
In this study, 1434 participants were recruited from 15 wards of Pune city. Of these, 723 (50.4%) were men and 711 (49.6%) women, 401(28.0%) were children ≤18 years and 1033 (72.0%) were adults >18 years. The age ranged from 1 month to 85 years with a mean of 31.2 years and a median of 29 years (Table 1).
Ward-wise sample seropositive for anti-DENV IgG antibody by indirect IgG ELISA is presented in Fig 2. Overall percent seropositivity was 81%. The median age of seropositives was 33. The percent seropositivity between wards was significantly different (p< 0.001). The proportion of seroprevalence varied among the wards from moderate high in Aundh (61.8%) to very high in Wanawadi (94.9%).
The difference in percent seropositivity between males (81.5%) and females (80.7%) was not significant (p = 0.745). Similarly, there was no significant difference (p = 0.786) in percent seropositivity between the participants visiting GP clinics (80.96%) and Corporation clinics (82.12%). Only 92 of 1,205 seropositive individuals (7.6%) could remember having dengue in the past.
Distribution of seropositive samples in different age group is presented in Table 1. There was an increasing trend with age, from 21.6% among < 36 months group to 77.3% in age group 16–18 years. The positivity was significantly different (p<0.001) in different age groups in children ≤ 18 years but not significantly different in adults (Fig 3A and 3B). In adults > 70 yrs (n = 42) all the persons were seropositive. A third order linear polynomial model is best fit to the overall data (R2 = 0.97). Our estimated seroprevalence at 9 years age (SP9) was 54.17% (95% CI: 49.13% - 58.97%), which is classified as a low-to-moderate DENV transmission intensity.
This test is designed to detect high levels of anti-DENV IgG antibodies indicative of a secondary infection. A total of 150 of 1,363 samples tested were positive (11.01%; 95% CI: 9.3%-12.6%). Overall seropositivity was highly variable between wards, ranging from 2.91% in Kondhwa to 20.95% in Hadapsar (S1 Table and S1 Fig).
Only 4 of 229 children in age group ≤ 10 (1.7%) were seropositive suggesting a very low rate of secondary infection in young children. The seropositivity in older age groups varied between 11.2 and 15.9%. Overall distribution of positive proportions was non-linear suggesting age independent phenomenon (Fig 4).
A total of 120 indirect IgG ELISA positive samples were tested for the presence of neutralizing antibodies by PRNT. Of these, 119 samples were confirmed to be seropositive via the presence of neutralizing antibodies and PRNT90 titers of ≥ 10. One sample had a PRNT90 titer of <10 against all 4 DENV serotypes and was considered seronegative (Table 2). Over 69.2% samples were positive for DENV 1–4 followed by 11.7% samples which were positive for 3 serotypes, DENV2, DENV-3 and DENV-4. There was significant difference in the percent positivity for different serotypes (p<0.01).
Percent PRNT positives for all DENV serotypes in different age groups are shown in Table 3. Amongst PRNT positives, DENV-2 was the most prevalent serotype across all age groups (94.4–100%). Only 5–6% individuals of age group up to 15 years were susceptible to DENV-2 and all individuals > 44 years of age were seropositive to this virus (Table 3). DENV-3 and DENV-4 follow age dependent linear distribution suggesting endemic nature of these serotypes for long duration but introduced late in comparison to DENV-2. DENV-1 was also prevalent across all the age groups in comparatively lower proportion and follows time independent distribution suggesting recent introduction. There is significant difference for percent positivity among different DENV serotype for age group 15 to 44 years (p = 0.006) and age group 60 years and above (p = 0.026) (Table 3). The sample size for PRNT was not enough for serotype specific model building for force of infection.
Higher titers of neutralizing antibodies (log10 PRNT90) were detected in individuals infected with DENV-2 (2.524; 95% CI: 2.407–2.641) compared to other serotypes. The titer for DENV-4 was lowest among four serotypes (1.943; 95% CI: 1.844–2.041). There was significant difference between overall neutralizing antibody titer of the four serotypes (p<0.05; F = 23.568). Post hoc (LSD) test showed that this difference is because of neutralizing antibody titer of DENV-2 (2.524; 95% CI: 2.407–2.641) which was significantly higher than the titers of all other serotypes (p<0.05) and there is a significant difference between titer of DENV-3 (2.109; 95% CI: 1.987–2.231) and DENV-4 (1.943; 95% CI: 1.844–2.041) (Fig 5).
The titer of DENV-2 was highest across all age groups followed by DENV-3. In younger age groups, DENV-1 exhibited lower titer and in higher age groups DENV-4 showed lowest titer. However, differences in the titers across all age groups were not significant (Fig 6).
To estimate the transmission intensity of dengue, two catalytic models were fitted to the age-specific seroprevalence for indirect ELISA data, time constant (model A) and time varying (model B) forces of infection (Fig 7). As per model A, dengue naive children seroconverted at the rate of 7.81% per year (95% CI: 7.24%-8.43%). Under model B, annual rate of seroconversion was 8.68% (95% CI: 7.52%-9.95%) in younger population ≤ 18 years and 7.51% (95%CI: 6.87%-8.20%) in individuals > 18 years. The LR test showed non-significant association to favor any model B (p = 0.091) (Table 4). We also fitted model B with 15 and 12 years age break point without any significant difference in FOI. Thus the model B was consistent with a significantly higher force of infection during the period 2000–2016 (λ = 0.868; 95%CI: 0.752–0.099).
Our estimated basic reproductive number (R0) for dengue in Pune is 4.23. These estimates assume endemic circulation of 4 serotypes and were derived using the FOI estimates and census data. The sample size was not enough to calculate FOI for individual wards. We pooled data by zone. Each zone consists of 3 adjoining wards. The mean R0 for Pune was 4.23 (95% CI: 3.58–4.87), lowest 3.41 in zone 3 and highest 5.25 in zone 1 (S2 Fig).
For estimation of the burden of disease, we have taken FOI as 8.68% for primary infection and 7.51% for secondary infections (Table 4). Accordingly in Pune city with estimated population of ~3.99 million in 2016, we estimate that this leads to approximately 65,800 (95% CI: 57009–75507) primary infections and 242,716 (95% CI: 222,032–265,016) secondary infections per year. Assuming that 69% immune population is positive for all 4 serotypes (Table 2), only 31% of secondary cases are likely to give rise to active dengue cases. Therefore, 75,242 secondary infections only are considered potentially secondary infections for estimation of dengue cases. The ratio between in-apparent and symptomatic dengue cases is highly variable, ranging from 1:1 to 3:1. Considering a ratio of 3:1, Pune city is burdened by about 47,000 symptomatic dengue cases each year.
We found 11.1% seropositivity in Capture IgG ELISA. This test is designed to detect high levels of anti-DENV IgG antibodies indicative of secondary infection. This translates to 358,741 secondary infections; 111,210 potential secondary cases and 59,000 dengue cases each year in Pune.
Dengue was first reported in India from Calcutta in 1912.[27] Now, it is a well established endemic disease in majority of Indian cities with occasional epidemics.[28,29] In Pune city, sporadic cases were reported in 1970s and 80s. Seasonal outbreaks have been recorded from 90s in different localities of the city with hemorrhagic involvement in some cases.[30] In spite of high prevalence of clinical disease, limited information is available on prevalence and incidence of the disease in India. Overall 81% IgG positivity by indirect ELISA with ~ 100% positivity in age groups > 45 years reported by us is much higher than 43% and 59% reported from 2 villages near Pune. In our study, seroprevalence of dengue was 50% in children of 6–10 years age group. In the same age group, high positivity is reported in Mumbai (80%), Delhi (60.2–66.5%), Wardha (69%), Bangalore (62%), Hyderabad (58%) and low positivity in Kalyani (23%).[31] The seropositivity of 79.3% in age group from 5–40 years is lower than 93% reported in Chennai.[32] Seropositivity of 11% by Capture ELISA and 81% by Indirect ELISA in our study is similar to the report from Hyderabad.[33] High seropositivity is also reported in Asian countries like Thailand, Bangladesh, Indonesia etc. [34–36]
Human population density is reported to be an important variable associated with a high historical incidence of dengue.[37–40] The level of seroprevalence seems to be also associated with the population size of the city. Small places like a village near Pune, population (2,621) and Kalyani in WB (population 100,575) reported low seroprevalence. Hyderabad and Pune similar in population pattern have nearly similar dengue prevalence; Chennai, Mumbai and Delhi, the metropolitan cities reported higher seroprevalence.
The lowest FOI and R0 for Indian subcontinent reported was based upon the data from Andamans island collected in 1988–89.[6,41] Population of this region was also very small on individual islands. In our study, only 7.6% participants could recall having the disease which is suggesting of a high frequency of unapparent infection or mild undifferentiated fever in agreement with other epidemiological studies.[42–45]
We estimated FOI, seroconversion rate, 8.68% in younger age group ≤ 18 years and 7.51% in older age groups. This is very different from 23% in Chennai. The reported seroconversion is highly variable in different places. In Sri Lanka, 8% seroconversion was reported in children ≤ 12 years age, 11 to 17% among children aged 2 to 15 years in Vietnam, 2.1 to7.9% in Thailand, 10% in Bangladesh, 13.1% primary infection per year in children in Indonesia, 17% in children aged 3 years in Salvador, Brazil.[22,32,35,46–52]
Our estimated R0 for dengue, 4.3 is lower than 5.3 estimated in Chennai and is comparable to estimates in hyperendemic settings in Thailand and Brazil.[6,53,54] As reported in other places, we also found significant heterogeneity between different wards.
In India, this is the first study to provide data on PRNT90, the test recommended by WHO for survey for neutralizing antibodies. Over 69% indirect ELISA positive samples were positive for all 4 serotypes followed by 11.7% positive for 3 serotypes, DENV-2, DENV-3 and DENV-4. DENV-2 was the most prevalent (94.4%) serotype across all age groups. This suggests widespread circulation of all the serotypes in Pune for quite some time. In another study, we reported active circulation of all the serotypes in Pune during 2016 dengue season.[19] Only other report from India, based on PRNT50 titers in children of 5–10 years age groups, reported overall positivity of 97.2% for at least one serotype, 79.7% for all four serotypes. DENV-1 was dominant serotype in Delhi; DENV-2 in Mumbai, Wardha, Bangalore and Hyderabad; DENV-3 in Kalyani. There is ample evidence that all 4 serotypes have been circulating in majority of the Asian countries.[55,56]
For analysis of neutralization tests, some investigators used PRNT60, others PRNT50 [31,51] making it difficult to compare the results of different studies. Following WHO recommendations,[12] we used PRNT90. In case of DENV-2, the multitypic response was positively associated with age because of the diversity of antibodies generated as a result of ongoing exposure. Highest neutralizing antibody titers observed for the DENV-2 suggests ongoing activity of this virus over the years and a multitypic response caused by the booster effect.[52,57,58]
One of the limitations of the present study is that PRNT was performed in a subset of individuals since it is expensive and laborious. Therefore the study population may not be true representative of the entire city. However, in spite of limitations our results provide valuable data on previous immunity at population level. For estimation of number of dengue cases, whether average seropositivity (11.1%) in capture ELISA for estimation of secondary cases in population can be extrapolated or not is an important issue. According to the manufacturer of the Panbio kit and others an IgG result of 22 Panbio units correlates with an HI titer of 1:1280, the cut-off used to distinguish between primary and secondary dengue infection.[59–61] Therefore, percent positivity in capture ELISA was used for estimation of secondary dengue under assumption that the high titered antibodies wane to below 22 Panbio units within a year and before next dengue season. However, there is a need to generate region specific data on decay pattern of these high titered antibodies in population. Further, in absence of testing for IgM antibody, possibility of primary infection in some cases cannot be ruled out.
Currently, the use and deployment of vector control as part of dengue outbreak response strategies is managed by public health in Pune city. It is highly unreliable and unsustainable due to limited resources and difficulties in management of human resources involved in vector control measures. There is no impact assessment in place for such a measure. There is also growing evidence that vector control is not a logical solution for control of dengue in large cities.[62] In the absence of specific drugs and limited usefulness of vector control measures, suitable vaccines are eagerly awaited.[63]
Dengvaxia, a live attenuated (recombinant) tetravalent vaccine is a licensed vaccine for dengue in several countries for children 9 years of age or older living in DENV endemic areas having high endemicity among 9 year-olds. Children who are seronegative at the time of first vaccination may be primed for future risk of severe dengue illness in areas of low to moderate (SP9 = 30%-50%) and even moderate to high (SP9 = 50%-70%) endemicity.[11] Therefore it was suggested that average seropositivity of 70% may be minimum requirement for introduction of the vaccine because of variability from locality to locality. With estimated average SP9 = 54% in present study. This vaccine is not suitable for Pune at this stage for the specified age group ≥ 9.
It has been well-documented that passive surveillance involving case notifications does not accurately reflect the burden of dengue in most of locations. Cohort studies in different provinces of Thailand and in Nicaragua had revealed higher numbers of prospectively determined dengue incidences as compared with national reported figures, with a discrepancy of 8 to 21.3-folds.[64–66] According to Shepard et. al. (2014) disease burden of dengue in India is 282 times the reported number per year, substantially more than captured by officially reported cases.[9] In this study, we estimated 47,000 to 59,000 cases per year in Pune city alone. As per official government report, only 6,792 cases of dengue were reported from whole of the Maharashtra in 2016. Therefore, it is strongly recommended that for a disease like dengue, serosurveys should be conducted periodically. It could shed light on the true dengue infections in the population and can be a good tool to monitor impact of interventions at population level.
|
10.1371/journal.pgen.1005326 | Ribosomal Protein Mutations Result in Constitutive p53 Protein Degradation through Impairment of the AKT Pathway | Mutations in ribosomal protein (RP) genes can result in the loss of erythrocyte progenitor cells and cause severe anemia. This is seen in patients with Diamond-Blackfan anemia (DBA), a pure red cell aplasia and bone marrow failure syndrome that is almost exclusively linked to RP gene haploinsufficiency. While the mechanisms underlying the cytopenia phenotype of patients with these mutations are not completely understood, it is believed that stabilization of the p53 tumor suppressor protein may induce apoptosis in the progenitor cells. In stark contrast, tumor cells from zebrafish with RP gene haploinsufficiency are unable to stabilize p53 even when exposed to acute DNA damage despite transcribing wild type p53 normally. In this work we demonstrate that p53 has a limited role in eliciting the anemia phenotype of zebrafish models of DBA. In fact, we find that RP-deficient embryos exhibit the same normal p53 transcription, absence of p53 protein, and impaired p53 response to DNA damage as RP haploinsufficient tumor cells. Recently we reported that RP mutations suppress activity of the AKT pathway, and we show here that this suppression results in proteasomal degradation of p53. By re-activating the AKT pathway or by inhibiting GSK-3, a downstream modifier that normally represses AKT signaling, we are able to restore the stabilization of p53. Our work indicates that the anemia phenotype of zebrafish models of DBA is dependent on factors other than p53, and may hold clinical significance for both DBA and the increasing number of cancers revealing spontaneous mutations in RP genes.
| The p53 tumor suppressor is the most commonly mutated gene in human cancers. However, cancer cells exploit multiple mechanisms to silence the p53 pathway in addition to inactivation of the p53 gene. We previously reported that one of these mechanisms is found in tumor cells with ribosomal protein (RP) gene mutations. These cells transcribe wild type p53 mRNA yet do not stabilize p53 protein when exposed to DNA damaging agents. In this work we demonstrate that this loss of p53 protein is due to its constitutive degradation. This degradation is due to impairment of the AKT pathway, which normal signals for p53 to stabilize when the DNA is damaged. By re-activating the AKT pathway in RP-mutant cells we are able to restore p53 stabilization and activity, which may hold clinical significance for cancer treatment.
| The stabilization of the p53 tumor suppressor is a pivotal event in the programmed cell death response. Levels of p53 protein are normally kept very low through its physical association with the MDM2 protein, an E3 ubiquitin ligase that constitutively ubiquitinates p53 and targets it for proteasomal degradation [1]. Many kinds of cellular stress, including DNA damage and oncogene presence, activate different signaling pathways that result in the dissociation of p53 and MDM2. p53 then stabilizes and translocates to the nucleus where it targets genes that arrest the cell cycle and turn on DNA repair, or genes that induce apoptotic cell death if the damage is deemed irreparable [2].
The stabilization of p53 has been reported to trigger human bone marrow failures such as dyskeratosis congenita and Fanconi anemia [3,4]. While Fanconi anemia is predominantly linked to mutations in DNA repair enzymes, several genes found mutated in dyskeratosis congenita patients have a known role in the rRNA maturation steps of early ribosome biogenesis. The mutation of these latter genes in zebrafish stabilizes p53, as does the mutation of several other genes important for the processing of rRNA [5–7]. In human bone marrow failures syndromes linked to RP haploinsufficiency such as Diamond-Blackfan anemia (DBA) and 5q-myelodysplastic syndrome, the loss of hematopoietic progenitor CD34+ cells by p53-induced apoptosis is believed by some to be the major cause of cytopenia [8]. However, the contribution of p53-induced apoptosis specifically to the cytopenia phenotype remains controversial. Recent studies demonstrated that patient CD34+ hematopoietic progenitor cells carrying mutations in the most commonly mutated gene linked to DBA (RPS19) do not reveal any hallmarks of apoptosis as they are induced to differentiate into erythrocytes [9]. This work also showed that while the co-depletion of p53 with RPL11 in CD34+ cells reduced some apoptotic effects, it did not restore the proliferation capacity lost upon RPL11 depletion alone. Therefore the contribution of p53 stabilization to the loss of erythrocytes in DBA patients is possibly less significant than previously thought.
In addition to being important for erythrocyte production, there also exist several reports indicating a role for RPs as tumor suppressors. Human patients with 5q-MDS or DBA are more predisposed to developing malignancies, both acute myeloid leukemia (AML) and solid tumors, respectively [10,11]. Importantly, the advent of exome sequencing has unveiled a surprising number of somatic RP gene mutations in an array of human cancers. These recent exome sequencing reports have identified mutations in RPL5 and RPL10 in T-cell acute lymphoblastic leukemia, mutations in RPL5 in gliomas, and mutations in RPL22 in human endometrioid endometrial cancer and colorectal cancer [12–14].
Embryonic zebrafish mutants and morphants are popular models of DBA and RP loss. In mutant lines where RP genes are disrupted by murine virus integrations, the homozygous embryos reveal a progressive reduction of the RP over the first 3 days post fertilization (dpf) coupled to a failure of hemoglobin-expressing cells to develop [15,16]. In contrast, haploinsufficient RP embryos reveal no cytopenia or any other conspicuous phenotype except for a mild growth defect that does not affect their development into adulthood [17]. Once reaching adulthood however, many of the haploinsufficient mutant lines reveal the formation of malignant peripheral nerve sheath tumors (MPNSTs), a tumor type rarely observed in laboratory strains of zebrafish [18]. Interestingly, this exact tumor type arises with the same frequency in zebrafish carrying homozygous mutations of p53 in a highly conserved residue within one of the DNA-binding domains (M214K) [19]. Further study into this revealed that while wild type p53 mRNA was normally transcribed in the RP-mutant tumor cells, p53 protein was unable to be detected [20]. This observation was despite the application of ionizing radiation and/or proteasome inhibition with MG132, two conditions that were found to stabilize p53 in zebrafish tumor cells carrying wild type RP genes.
Actinomycin D is a drug that disrupts ribosome biogenesis by inhibiting rRNA polymerase and results in the stabilization of p53 [21]. It was recently reported that this induction of p53 requires the activity of the survival factor AKT/PKB [22], a kinase that responds to the stimulation of many growth factor receptors, including insulin receptors [23]. While activated AKT has many functions, one major event of insulin stimulation directly downstream of AKT is the phosphorylation and inhibition of glycogen synthase kinase-3 (GSK-3) [24]. Under normal conditions GSK-3 phosphorylates and activates MDM2 in a way that promotes p53 degradation [25]. However upon DNA damage by ionizing radiation, the phosphorylation of AKT results in the inactivation of GSK-3 [26]. This AKT-mediated inactivation of GSK-3 in response to ionizing radiation begins with the DNA-dependent protein kinase (DNA-PK), a protein that recognizes the double-stranded breaks on DNA and signals through a cascade that ultimately results in p53 stabilization and programmed cell death [27]. Thus one mechanism of the p53 stabilization in response to ionizing radiation is through the activation of DNA-PK and AKT inhibiting the downstream activity of GSK-3 and MDM2.
We previously reported that the loss of RP genes in mammalian cells and in zebrafish embryos results in a loss of AKT activity that could be overcome by the addition of insulin [16]. This observation led us to consider the possibility that the AKT pathway was involved in the regulation of p53 in RP mutant cells, and that the impairment of the DNA damage pathway through AKT may have a role in the predisposition to malignancy caused by RP gene mutations.
The RP mutant zebrafish lines we used for this study were generated by viral insertions in the introns of RP genes, two of which (rpS7 and rpL11) have homologues found mutated in DBA patients [28,29]. We find that at 2 dpf these embryos display normal expression of the scl transcription factor required for the genesis of hematopoietic stem cells in both primitive and definitive hematopoiesis, a result that was also observed in zebrafish embryos with deletions in the rpS19 gene (S1 Fig) [30–33]. However, RP mutants reveal a marked decrease in the expression of the βE1-globin gene, which (at this stage of development) normally becomes up regulated as cells commit to the erythrocyte lineage (S1 Fig) [34]. Zebrafish embryos carry maternal stores of RPs such that these mutants reveal a progressive loss of RP expression. The rpS7 mutants express about 50% of wild type rpS7 levels at 1 dpf while this percent reduction is not observed in rpL11 mutants until 3 dpf [16,35]. This likely explains why despite the fact that embryos from both mutant lines reveal hematopoietic and developmental phenotypes, these are more severe in the rpS7 mutants [16,35]. Because these phenotypes at 1 dpf are much more pronounced in rpS7 embryos, we selected them for the following analysis of apoptosis. S2 Fig provides images of the mutants compared to wild types and shows the morphological features that we use to initially select the mutants, such as the smaller head and eye, inflated hindbrain, and the dent in the mid-hind brain barrier.
We measured overall levels of cell death in developing embryos carrying the rpS7 mutation with acridine orange (AO), a stain commonly used to detect cells undergoing apoptosis in zebrafish embryos [36]. Fig 1A and 1B show that at 1 dpf, rpS7 mutant embryos reveal a significantly larger number of apoptotic cells compared to wild type controls. Closer visualization of an rpS7 mutant in Fig 1C reveals that these AO-stained cells are found clustered in the brain area and evenly distributed on the entire surface of the tail. The injection of a translation-blocking morpholino that we have previously demonstrated is specific to zebrafish p53 (p53 MO) but not a missense control morpholino (mis MO), completely rescued the number of cells undergoing apoptosis in the rpS7 mutants, reinforcing a role for p53 in cell death as a result of RP loss (Fig 1A and 1B) [37]. The p53 MO also results in the rescue of morphological phenotypes commonly observed in ribosome biogenesis mutants such as the inflation of the hindbrain vesicle and pericardial edemas, a rescue effect that has been previously demonstrated in other zebrafish models of RP loss (S2 Fig) [15,38]. To determine if the p53 MO was sufficient to rescue the defective hematopoiesis of the mutants we used o-dianisidine, which stains hemoglobin-expressing cells evident at 2 dpf. Staining with o-dianisidine revealed that despite the rescue of apoptotic cells by p53 depletion observed at 1 dpf in the rpS7 mutants, the p53 MO was not able to rescue hematopoietic development to any appreciable degree (Fig 1D).
The levels of caspase-induced apoptosis in human CD34+ cells with RP loss vary depending on the RP [9]. To determine if the loss of RPs triggers caspase-induced apoptosis in zebrafish embryos we measured both the basal levels of activated caspase 2 or 3/7 and their levels in response to ionizing radiation, comparing wild type embryos with those carrying mutations in rpS7 and rpL11, as well as rpS3 and rpL36 genes. Fig 2A and 2B indicate that in 2 dpf embryos the caspase 2 and 3/7 basal levels in the mutants is equivalent to wild types, and the caspase response to DNA damage is severely impaired in all the mutants. In fact, the RP mutant lines show the same suppressed caspase 2 and 3/7 response to ionizing radiation as the homozygous p53M214K/M214K mutant line (Fig 2A and 2B), which is severely impaired in its ability to induce apoptosis [19]. To determine if other hallmarks of apoptosis such as DNA fragmentation are present in RP mutants we performed TUNEL assays on rpS7 embryos at 2 dpf. Fig 2C and 2D show that while there is no appreciable difference in the levels of TUNEL-positive cells between the rpS7 mutant and wild type embryos, the exposure of mutants to ionizing radiation results in a similar significant increase of TUNEL-positive cells as observed in the wild types. Closer visualization of these DNA-damaged embryos reveals a localization of TUNEL-positive cells in the brain area similar to what is seen with the AO staining, however the localization of TUNEL-positive cells in the tail region is found much more restricted to the dorsal area above the notochord (Fig 2E). This may be due to the limits of penetration of the AO stain, or may suggest that the cells in this dorsal area are especially sensitive to DNA damage-induced apoptosis, as are the rapidly proliferating cells in the brain.
The increased transcription of the zebrafish p53 gene and its p53Δ113 isoform (a target gene of stabilized p53) has been described in several models of RP loss and likely reflects the early response of p53 that triggers an up regulation of its own transcription and the transcription of p53Δ113 [15,35,38–41]. In line with these results, we found using real-time quantitative PCR analysis with primers that amplify both full-length p53 and p53Δ113 that p53 mRNA levels were significantly higher in rpS7 and rpL11 mutants at 1 and 2 dpf both in the presence and absence of ionizing radiation compared to untreated wild type embryos (Fig 3A and 3B). Semi-quantitative PCR analysis of p53 mRNA levels in several other RP-mutant embryos (rpS3a, rpL23a, and rpL36) at 2 dpf similarly revealed equivalent levels of p53 transcription in the mutants compared to wild types (S3 Fig). However, when we performed western blotting analysis using a zebrafish p53-specific antibody, we were unable to detect any appreciable amount of p53 protein in the rpS7 or rpL11 mutants in either the presence or absence of ionizing radiation at either 1 or 2 dpf (Fig 3C). This is the case in all the mutant RP lines we tested including in rpS3a, rpL23a, and rpL36 (S3 Fig). We often observe what may be a p53-specific isoform such as p53Δ113 on the western blots, but this may also be a p53 degradation product and in this work we cannot be certain of its exact identity. Taken together, the results suggest that although the p53 response to the RP mutation on a transcriptional level may function normally, an additional level of p53 post-translational regulation exists in the presence of RP mutations that serves to reduce p53 protein.
We recently demonstrated that AKT phosphorylation activity is impaired in zebrafish embryos carrying mutations in rpS7, rpL11, and rpS3a [16]. Since the loss of AKT activity would theoretically lead to an increase of p53 proteasomal degradation due to an increase of GSK-3 phosphorylation of MDM2, we used the proteasome inhibitor MG132 to determine if we could restore p53 expression. Fig 4A illustrates that MG132 is able to stabilize p53 in wild type embryo cells as expected, and moreover it is able to stabilize p53 in the rpS7 mutants. This supports the notion that the loss of p53 in the RP mutants is the result of excessive proteasomal degradation. We therefore hypothesized that stimulating AKT in the presence of ionizing radiation may rescue the p53 response to DNA damage. Fig 4B (lanes 1–4) shows p53 stabilization approximately 6 hours after exposure of wild type embryos to ionizing radiation when the embryos are treated with the anti-oxidant Trolox, insulin, or both. Fig 4B (lanes 5–8) illustrates that the addition of insulin, but not Trolox, immediately following the ionizing radiation results in a rescue of the stabilization of p53 in rpS7 mutants. These data suggest that overcoming the RP mutation-induced inhibition of AKT with insulin, which would stimulate AKT more directly than Trolox, is able to rescue the p53 stabilization response to ionizing radiation. A diagram illustrating how the impairment of the AKT pathway can lead to p53 protein degradation through GSK-3 is shown in Fig 4C.
We hypothesized that the impairment of AKT activity observed in cells with RP mutations could cause constitutive activation of GSK-3, phosphorylation of MDM2, and ultimately resulting in the constitutive degradation of p53. Therefore we reasoned that the inhibition of GSK-3 with lithium chloride (LiCl) in cells with RP mutations could restore p53 stabilization. Fig 5A and 5B show a partial rescue of p53 stabilization in rpL11 and rpS7 mutant embryos exposed to ionizing radiation followed by 6 hours of LiCl treatment. Finally, p53 expression is completely restored in rpS7 haploinsufficient MPNST tumor cells when the cells are plated overnight with different dosages of LiCl (Fig 5C). This figure also shows LiCl induces the expression of what may be different isoforms of p53 such as p53Δ113 resulting from restored transcriptional activity of p53 in the tumor cells, for LiCl treatment of MPNST tumor cells expressing the transcriptionally impaired p53M214K/M214K mutant does not result in the appearance of same bands (Fig 5D). However we are still unable to state for certain what the exact identity of these bands are, although they appear to be specific to p53 activation. These data strongly implicate the impairment of AKT, in particular the loss of AKT inhibition of GSK-3, as a driving force behind the high levels of constitutive degradation of p53 observed in tumor cells with RP gene haploinsufficiency.
While stabilization of the p53 tumor suppressor has long been held the culprit for the cytopenia phenotype observed in human diseases linked to RP gene mutations, it has been a decade-long mystery as to why no p53 protein is detectable in RP haploinsufficient tumor cells. In the context of the RP mutant zebrafish embryos, we show that a similar loss of p53 protein is evident as the embryos age to 2 dpf, the maternal stores of RPs are depleted, and the RP deficiency resulting from the mutation becomes more severe. We do not believe that this loss of p53 in RP mutants is simply due to the inability of the cells to make protein per se, for in other zebrafish models of ribosome biogenesis deficiency such as nucleostemin, gnl2, and nop10 mutants we are able to visualize robust p53 stabilization by western blot analysis even in the absence of ionizing radiation up until 4 dpf [5,6]. Interestingly, the nop10, gnl2, and nucleostemin genes all code for proteins with important roles in early ribosome biogenesis and the processing of rRNA. While it has been shown that there are indeed detectable defects in rRNA processing in DBA patient CD34+ cells with RP mutations, these are by and large more subtle than the defects we observe upon the loss of nop10, gnl2, or nucleostemin [5,6,42]. This suggests that, as with actinomycin D, defective rRNA processing is a critical mediator of p53 stabilization and may in fact be a causative agent in the cytopenia phenotype of human diseases such as dyskeratosis congenita. However, our data suggest the molecular pathology underlying the anemia in diseases linked to RP mutations is likely to implicate mechanisms that go beyond stabilization of the p53 protein.
The fact that others and we observe no difference in early hematopoietic stem cell expression in RP mutant zebrafish embryos coupled to the anemia failing to be rescued upon p53 loss suggests that the anemia phenotype cannot be explained by the organism’s general p53 response to the ribosome biogenesis defects induced by RP loss [30]. Clearly the embryos younger than 1 dpf are experiencing some p53-induced apoptosis otherwise we would see no AO staining rescue upon injection of the p53 MO and no partial rescue of the morphological phenotypes. We therefore propose that p53 has such an early effect in the developing RP mutant embryo that by the time of 1 dpf (approximately 30 hours post fertilization is when the experiments would begin) all that we are able to observe by laboratory techniques are the apoptotic cells in the wake of a brief p53 activation. The remaining live cells at 1 dpf appear devoid of p53 protein, and if the apoptotic cells do still carry p53 protein we cannot detect it by western blotting. We were unable to detect p53 by western blotting in any embryo younger than 1 dpf, but this is likely a technical issue reflecting the abundance of yolk protein still present at this early stage. The apoptotic cells rescued by p53 loss that we are able to observe at 1 dpf also do not appear to be important for red cell development, as we detect no rescue of the anemia phenotype upon p53 loss.
The results of TUNEL assay suggests that some DNA fragmentation upon acute DNA damage is still possible in the 2 dpf RP mutant cells despite no evidence of p53 stabilization or caspase activation. Other studies have reported similar findings of caspase-independent DNA fragmentation in cancer epithelial cells in response to diverse toxins, and the authors of this work suggest that the DNA damage induced by these toxins may lead to noncaspase-mediated proteolytic activation of DNases [43]. Interestingly, studies in Drosophila have shown that ionizing radiation leads to cells acquiring RP haploinsufficiency, and that these cells undergo apoptosis that is also p53-independent [44]. Taken all together, our work supports a model where the impairment of mature red blood cell formation in organisms with RP deficiency is due to cell loss that is not dependent on either p53 stabilization or caspase activation. However, it should be pointed out here that the work in this study is entirely based on zebrafish models of DBA, which carry homozygous RP mutations (as noted in the introduction, the haploinsufficient mutants have no embryonic phenotype except a slight growth defect). Therefore it may be that p53 has a more prominent role in cells with true RP haploinsufficiency, and this remains an issue that must be kept in mind when interpreting data generated from zebrafish models of DBA.
While other studies have suggested that the p53 MO or p53M214K/M214K mutant background rescues both general apoptosis and the number of hemoglobin-expressing cells in embryos deficient for RPs [15,38], we believe our genetic models are more consistent than using morphants and that our approach to quantifying the mutant phenotypes are both more robust (we use sample sizes N > 100) and reliable (we genotype the embryos after the blind scoring to resolutely identify the homozygotes). Other studies support us by demonstrating that p53 independent pathways are contributing to the anemia phenotype of RP-mutant zebrafish embryos, and that the loss of p53 rescues the morphological abnormalities but not the anemia phenotype of embryos with reduced RP expression [45–47].
It remains an open question as to what the major pathways beyond p53 and caspase activation contribute to the death of RP deficient cells and the anemia phenotype in zebrafish models of DBA. Interestingly, a recent study has shown that the mutation of rpS19 in zebrafish embryo erythrocytes specifically reduces the translation, but not the transcription of the hemoglobin gene hbbe3 [30]. We also recently reported that the up regulation of autophagy, the cellular process of self-digestion that is tightly regulated during hematopoiesis, is observed in both zebrafish models of DBA and in human DBA cells with RP haploinsufficiency [16]. It may therefore be that RP loss in maturing erythrocytes derails their proper differentiation by failing to translate critical mRNAs, or by the constitutive activation of autophagy resulting in the erythrocyte progenitor cells essentially eating themselves before they are able to fully differentiate. The former possibility is supported by several findings suggesting that L-leucine, an amino acid that increases translation by activating the mTOR pathway, is able to partially rescue the anemia phenotype of zebrafish RP morphants and increases the number of erythroid cells in red cell culture assays where CD34+ cells are infected with shRNAs against RPS19 or RPS14 [47,48]. The latter possibility is enticing for one would not expect constitutively up regulated autophagy to elicit an apoptotic response. Or the cells may be undergoing an as of yet unknown p53- and caspase-independent mechanism of cell death that awaits identification.
We had previously reported that the addition of MG132 to the MPNST cells with RP gene haploinsufficiency was not able to restore p53 stabilization [20]. At the time this led us to believe that the block in p53 expression was at the level of protein synthesis, since MG132 should be able to stabilize any protein undergoing ubiquitination and proteasomal degradation. However our present study suggests otherwise, that in fact the MG132 treatment is not sufficient to restore p53 stabilization in RP mutant MPNST cells. There may be additional factors at the tumor level contributing to the degradation of p53 that are as of yet not known, or it may be that the robustness of the degradation in the tumors is much stronger than in the embryos. While our present study is not able to delineate between these possibilities, it may be that more powerful proteasome inhibitors are capable of restoring p53 to the same levels as we observe with the LiCl treatment of tumor cells, and these studies will be of great future interest. It should also be pointed out here that given the very wide range of effects that LiCl has on many cellular signaling pathways, there may be other effects beyond the AKT pathway that are contributing to the restoration of p53 that we observe.
Very recent work suggests that a mechanism for malignant transformation in RP-haploinsufficient cells involves cells acquiring oncogenic mutations that allow for the bypass of a 60S ribosomal subunit integrity checkpoint [49]. We propose that these mutations (which have yet to be identified) coupled to excessive degradation of wild type p53 would be sufficient to promote malignant transformation. Interestingly, the T-ALL cells with RPL5 and RPL10 mutations, similar to the RP-haploinsufficient zebrafish tumor cells, the p53 gene remains wild type (personal communication with Dr. De Keersmaecker). We thus posit that cells with RP deficiencies are able to exert selective pressure to overcome the p53 activity not just by acquiring p53 gene mutations but also by suppressing AKT activity.
DBA is a rare disease with a prevalence that is approximately 7 in 1 million live births [50]. In addition, since the majority of DBA patients die early in life from bone marrow failure or from complications stemming from chronic blood transfusions, the number of DBA patients who ultimately experience malignant transformation is very low (it would be next to impossible for us to obtain DBA patient-derived tumor tissue for p53 analysis). In the absence of a mammalian RP mutant cancer model, the zebrafish RP mutants to date remain the best possible option for molecular studies of RP mutation-driven tumors. That said, we believe our results may still have some important implications for human leukemia. For example, acute myeloid leukemia (AML) has very low rates of p53 genetic inactivation (similar to the RP mutant MPNSTs) compared to very high rates in many other tumor types [51,52]. Inhibition of GSK-3 with small molecules in AML cells leads to increased differentiation, impaired growth and proliferation, and the induction of apoptosis [53]. Interestingly, the immunomodulating agent lenalidomide, used to treat multiple myeloma, results in the phosphorylation of GSK-3 at the same serine residues that inhibit GSK-3 phosphorylation of MDM2 [54]. This drug has also been reported in 5q-MDS patients to result in increased survival and a reduced risk of transformation to AML [55], raising the possibility that one mechanism of action of lenalidomide is to inhibit GSK-3 phosphorylation of MDM2 and restore a normal p53 response to the acquisition of oncogenic mutations. In sum, we suggest the new mechanisms driving p53 loss that we report here may be useful pathways to target in some cancers.
Zebrafish mutants were created and maintained as described [19,28,56]. Animal experiments were conducted in accordance with the Dutch guidelines for the care and use of laboratory animals, with the approval of the Animal Experimentation Committee (Dier Experimenten Commissie) of the Royal Netherlands Academy of Arts and Sciences (Koninklijke Nederlandse Akademie van Wetenschappen [KNAW] (Protocol # 08.2001).
Performed as previously described using probes against scl [31] and βE1-globin [57]. At least three independent stains were performed with clutches > 60 embryos. Clutches were stained simultaneously and mutants confirmed afterwards using Sanger sequencing, as previously described [58].
Embryos were sorted by gross phenotype and photographed with a Leica MZ FLIII microscope. At least three independent clutches were analyzed and the expected Mendelian ratio of mutants was always observed. Mutants were confirmed afterwards using Sanger sequencing, as previously described [58].
Embryos were microinjected with ~1ng MO at the one-cell stage. The p53 MO (Gene Tools, Inc. Philomath, OR, USA) is 5’-GCGCCATTGCTTTGCAAGAATTG-3’ and has been previously described [59]. The missense MO sequence is 5'-CATGTTCAACTATGTGTTAGCTTCA-3' (Gene Tools, Inc. Philomath, OR, USA)
Live 1 dpf embryos were incubated in E3-embryo medium + 10μg/mL AO stain (Sigma) in the dark for 30 min. Photos were obtained using a Leica MZ FLIII microscope and cells were blindly counted within a defined area that included the tail starting at the dorsal end of the yolk extension using Image J v1.44 software. At least 8 animals per condition were used for counting. Embryos were genotyped afterwards using Sanger sequencing, as previously described [58].
At least 100 embryos per clutch of a mating between two heterozygous hi1034b fish were injected at the one-cell stage with MOs as described above. At 2 dpf they were stained with o-dianisidine (Sigma) as previously described [16]. Scoring of the phenotype severity was done blindly. Embryos were genotyped afterwards using Sanger sequencing as previously described [58].
2 dpf embryos were untreated or exposed to 25 Gy of ionizing radiation. Six hours after irradiation TUNEL assays were performed as previously described and TUNEL-positive cells counted as for AO staining [6].
Embryos at 2 dpf were either untreated or subjected to 25 Gy ionizing radiation, and mechanically lysed 6 hours later with a P200 pipette (each sample = 3 embryos per well of a 96-well plate) using 100μL of the western blotting lysis buffer described above. The Caspase-Glo® 2 or 3/7 assay (Promega, Madison, WI, USA) was then performed per the manufacturer’s instructions.
5 embryos per sample were added to 100μL Trizol (Life Technologies, Carlsbad, CA, USA), RNA was isolated and used to make cDNA with the iScript cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA) per the manufacturer’s instructions. Primers used were as follows: p53 forward 5’- GCTTGTCACAGGGGTCATTT-3’, p53 reverse 5’-ACAAAGGTCCCAGTGGAGTG-3’, GAPDH forward 5’-GGATCTGACAGTCCGTCTTGAGAA-3’, GAPDH reverse 5’- CCATTGAAGTCAGTGGACACAACC, actin forward 5’-GCCCATCTATGAGGGTTACG-3’, actin reverse 5’-GCAAGATTCCATACCCAGGA-3’. Quantitative PCR (qPCR) was performed using iQ SYBR Green Super Mix and a MyiQ Single-Color PCR thermal cycler (Biorad, Hercules, CA, USA). Each experiment was performed in biological triplicate. p53 mRNA expression in mutants relative to wild types was normalized to GAPDH and calculated according to the Cτ method. Semi-quantitative analysis was performed with a standard PCR method using either p53 or actin primers, the products run on a 1% agarose gel. Statistics were performed with a Student’s t-test.
Embryos were subject to 25 Gy ionizing radiation and then lysed 6 hours later. 350nM insulin, 10mM Trolox or 100μM LiCl (all from Sigma) was added to the E3 media immediately following the ionizing radiation. For the MG132 (Sigma) experiment, embryo cells were dissociated mechanically with a 200μL pipette tip and added to DMEM media +10% FCS with or without 20μM MG132 for 6 hours at 28°C before lysing the cells. The zebrafish specific p53 antibody and western blotting of zebrafish embryos and MPNST cells has been previously described [20]. For MPNST cells the tumor was dissected, cells mechanically dissociated with a P200 pipette, and plated in a 6 well dish with the indicated concentration of LiCl at 28°C in DMEM + 10% fetal calf serum (Life Technologies, Carlsbad, CA, USA) media overnight. 50μg of total protein from the MPNST cells was used for each sample, measured by Bradford assay (BioRad, Hercules, CA, USA). Other antibodies used at 1:1000 dilution included anti-phospho-AKT (Cell Signaling #9275) and anti-actin (Santa Cruz Biotech, #sc-1616, Santa Cruz, CA, USA). Antibodies used at 1:5000 included donkey anti-goat IgG-HRP (Santa Cruz Biotech, #sc-2020, Santa Cruz, CA, USA) and sheep anti-mouse IgG-HRP (GE Healthcare, #NA931, Little Chalfont, Buckinghamshire, United Kingdom). Quantifications of western blots were performed using Image J v1.44 software. Unless otherwise stated, all embryos subject to western blotting were 2 dpf.
|
10.1371/journal.pgen.1000443 | Combined In Silico and In Vivo Analyses Reveal Role of Hes1 in Taste Cell Differentiation | The sense of taste is of critical importance to animal survival. Although studies of taste signal transduction mechanisms have provided detailed information regarding taste receptor calcium signaling molecules (TRCSMs, required for sweet/bitter/umami taste signal transduction), the ontogeny of taste cells is still largely unknown. We used a novel approach to investigate the molecular regulation of taste system development in mice by combining in silico and in vivo analyses. After discovering that TRCSMs colocalized within developing circumvallate papillae (CVP), we used computational analysis of the upstream regulatory regions of TRCSMs to investigate the possibility of a common regulatory network for TRCSM transcription. Based on this analysis, we identified Hes1 as a likely common regulatory factor, and examined its function in vivo. Expression profile analyses revealed that decreased expression of nuclear HES1 correlated with expression of type II taste cell markers. After stage E18, the CVP of Hes1−/− mutants displayed over 5-fold more TRCSM-immunoreactive cells than did the CVP of their wild-type littermates. Thus, according to our composite analyses, Hes1 is likely to play a role in orchestrating taste cell differentiation in developing taste buds.
| The sensation of taste is composed of five basic modalities: sweet, bitter, umami, sour, and salty. Specialized taste cells perceive the various chemical cues within food. About 100 taste cells assemble into onion-shaped clusters called taste buds, which are located on taste papillae in the tongue epithelium and on oral mucosa. Of the five taste modalities, the taste stimulants responsible for sweet, bitter, and umami tastes are recognized by a group of G protein–coupled taste receptors, and the signal transduction pathways utilized following receptor stimulation share common molecules. However, it is still largely unknown how these molecules are regulated during taste cell development. We performed computer analyses based on previously known information about signal transduction pathways involved in the taste-sensing system to identify taste stem cells/progenitor factors of type II taste cells (responsible for sweet, bitter, and umami taste sensations). We found several transcription factors likely to bind to the regulatory regions of taste-related calcium signaling molecules (TRCSMs), and identified Hes1 as a potential candidate for common regulatory factors of TRCSMs. In vivo analyses using wild-type and Hes1 mutant mice confirmed that Hes1 regulates differentiation of bitter-, sweet-, and umami-sensing cells.
| Taste is one of the major chemosensory systems enabling animals to perceive crucial environmental stimuli. It performs the vital role of helping animals to identify favorable nutrition sources, as well as to avoid toxic substances, making taste a fundamental sensory recognition system that is required for survival [1],[2]. While the ontogeny of the other special sense organs has been studied in depth at a molecular level [3]–[5], the development of taste remains to be clarified.
Taste buds are the sensory end organs for gustation, and are located on the epithelium of the tongue and palate. On the tongue, they reside on three types of papillae, i.e., fungiform, foliate, and circumvallate [2],[6],[7]. In adult mammals, each taste bud comprises groups of 50–100 spindle-shaped epithelial cells and a small number of proliferative cells [8],[9]. Taste bud cells are heterogeneous in terms of gene expression profiling of individual taste cells, as well as in their ultrastructural characteristics. [8], [10]–[14]. Ultrastructual studies have revealed three distinct anatomical types of spindle-shaped epithelial cells within each taste bud: type I (dark), type II (light), and type III (intermediate) cells [8],[10],[11]. Type II cells have a characteristic large round nucleus and are responsible for the sweet, bitter, and umami taste sensations [2], [6]–[8],[10]. These cells express a number of G protein–coupled receptors and common downstream transduction components called taste receptor calcium signaling molecules (TRCSMs; e.g., PLCβ2, gustducin [GNAT3], and IP3R3 [ITPR3]) [2],[6],[7],[10]. Although several studies have examined the lineage of taste cells [8],[10],[15],[16], the molecular mechanisms of cell differentiation in developing taste buds have remained elusive. We took a novel approach toward investigating taste cell development in mice by combining in silico and in vivo analyses of the TRCSM transcription regulatory network in type II taste cells.
We examined the expression of TRCSMs in the epithelium of presumptive circumvallate papillae (CVP) during mouse embryogenesis. The papilla structure of CVP is already visible before embryonic day 14 (E14) [17]. We examined expression of five TRCSMs—PLCβ2, gustducin, IP3R3, Ggamma13 (GNG13), and Trpm5—in developing CVP by immunohistochemistry and/or in situ hybridization. We identified the appearance of cells expressing these TRCSMs (which are widely accepted as representative markers for differentiated taste cells) in serial sections from the posterior one-third of the embryonic tongue (Figure 1) [2],[6],[7],[10]. One series of sections from an entire CVP was subjected to each combination of antibodies or probes, such as PLCβ2 and IP3R3 antibodies, and more than five CVPs were subjected to histological analysis with each combination of markers at each developmental stage.
In previous studies, cytokeratin-8/Troma1 (CK8) staining revealed that the taste bud primordia in CVP appear from E15 onward during early development of the tongue epithelium in mice [18]; this is before the morphology of taste buds becomes evident, approximately 2 d postnatal (P2). Although a search was carried out for TRCSMs within the mouse tongue epithelium from E11 to E16, none were detected; TRCSM-positive cells appear as single isolated cells among the cell population immunoreactive against CK8 at E17 (Figure S1 and Table S1). These results indicate that TRCSM-positive cells appear in the developing CVP just before birth in mice (Figure 1, S1 and S2). Around the time of birth (E18 to P0), two or three TRCSM-positive cells were observable within the entire CVP (Figures 1, 4 and S2). These cells were not considered to be fully differentiated taste cells because they lacked certain crucial taste cell markers such as taste cell receptors. While previous studies reported incomplete overlapping of five TRCSMs in taste buds in the CVP of adult mice [19]–[21], we detected 100% colocalization of these TRCSMs (PLCβ2, gustducin, IP3R3, Ggamma13, and Trpm5) in the developing tongue epithelium, from E17 to at least P5 (Figures 1, S2 and Table S1). These results suggest that TRCSMs are expressed simultaneously in the same cell population during early development of the taste cell lineage in CVP.
The synchronous cellular colocalization of TRCSMs led us to investigate the regulatory mechanisms of TRCSMs, under the hypothesis that these genes are involved in the same regulatory network and share common regulatory factors, at least in the early phase of taste cell development. We analyzed the promoters of the five TRCSMs in silico to identify any common transcription factors that bind to regulatory sequences of taste stimuli signaling components. A series of putative transcription factor binding sites to these DNA sequences were identified by the Match program [22], which searched for regulatory sequences up to 5 kb upstream of each of the five TRCSMs (Figure 2). We further sieved common transcription factors through interspecies comparisons based on information acquired from mouse, rat, and human DNA sequences. Using these computational predictions, we identified 94 transcription factors as putative common transcription regulators (Figure 2A). These factors, which included candidates for factors implicated in the taste developmental system, are listed in Table S2. To evaluate this approach, we further performed a bibliographic and database search for gene expression within the embryonic oral epithelium. Because transcription repressors are presumably required to suppress the expression of TRCSMs in stem or precursor cells, we focused on transcription repressors within our list of identified candidates, in an effort to identify the regulator for taste stem cells or precursor cells. Ultimately, Hes1, a basic helix-loop-helix type of transcription factor, emerged as the most likely candidate from our different sets of informatics screenings (Figure S3).
To confirm that HES1 binds to the Plcβ2 and Ip3r3 promoter regions (Figure 2B), we ran chromatin immunoprecipitation assays (ChIP) using an antibody against HES1. We designed several pairs of primers to amplify putative HES1 binding sites in these promoter regions, as predicted by our in silico analyses. As controls, we also designed pairs of primers that did not contain the HES1 binding sequence (Figure 3). ChIP with the p1 primer pair yielded a higher recovery of chromatin than did ChIP with the control p2C primer pair (Figure 3). Similarly, the ip1 and ip2 primer pairs also yielded a higher recovery of chromatin than did the control ip3C primer pair. The ip3C control primers showed a relatively high recovery of chromatin, most likely due to the close location to the third HES1 binding site, and to their position between two HES1 binding sites within the Ip3r3 promoter region (Figure 3). These results suggest that HES1 bound the predicted sequences in the Plcβ2 and Ip3r3 promoter regions.
Because our composite approach to identifying factors in the regulatory network of taste system development picked up Hes1 as a strong candidate, we further analyzed the role of Hes1 in taste system development. In situ hybridization analyses against tongue epithelium from 3-weeks-old animals (W3) revealed that large numbers of cells within taste buds exhibited Hes1 transcript (Figure S4) [23], and that expression of Hes1 overlapped with the TRCSMs (data not shown). This observation contradicts somewhat the hypothesis that HES1 directly represses the expression of TRCSMs in taste buds; therefore, we performed detailed immunohistochemical analyses using HES1 antibody on CVPs from P0 animals and W3 animals (Figure 4). The TRCSM-positive cells observed at P0 showed a reduction in HES1 immunoreactivity within nucleus, suggesting that HES1 protein had evacuated from nuclei (Figure 4). In W3 animals, HES1 localized in the cytoplasm of most taste bud cells (Figure 4). This cytoplasmic HES1 can be considered to be nonfunctional as a transcription regulator. The few cells that showed HES1 localized in the nucleus as well as in the cytoplasm exhibited no IP3R3 expression (Figure 4; indicated by white arrows), while the cells with cytoplasmic HES1 only also expressed IP3R3 (Figure 4; indicated by arrowheads). This suggested that regulation of the subcellular localization of HES1 was important for taste cell differentiation.
Because HES1 represses transcription from bound promoters, cells positive for HES1 within the nucleus may be either precursor cells (including stem cells) of TRCSM-positive cells (type II; responsible for sweet, bitter, and umami taste) or other cell types within the taste cell lineage, such as type I or type III cells [10],[11]. Therefore, it is important to observe HES1 colocalization with markers for other differentiated cell types within taste buds [12],[24]. Similar results in the case of IP3R3 (Figure 4) were obtained with the blood type H antigen and SNAP25, which represent type I and type III cells, respectively, within taste buds (Figure 4) [12],[24]. Our results raise the possibility that HES1 is commonly expressed in precursor cells involved in the cell type differentiation pathway within CVP, and that Hes1 activity is required in the precursor or stem cell population in taste system development.
To clarify the potential role of Hes1 during development of the taste recognition system in vivo, we performed analyses of taste cell differentiation in mouse Hes1 mutants. Because Hes1−/− mice die at the newborn stage, observations of entire CVP by serial section were conducted around the time of birth. In wild-type littermates, PLCβ2/IP3R3-positive cells appeared as single, isolated cells (Figure 5A). However, in Hes1−/− embryos, the PLCβ2/IP3R3-positive cells were relatively small in shape, increased in number, and in contact with one another, forming cell clusters within the CVP of E18 embryos (Figure 5A). The total number of PLCβ2- and/or IP3R3-positive taste cells in the entire CVP was more than 5-fold greater in Hes1−/− embryos than in their wild-type littermates at E18 and P0 (Figure 5).
Previous lineage tracing studies have indicated that taste cells are derived from as-yet unidentified stem cells that reside outside of taste buds, and that immature but postmitotic progenitors derived from these stem cells enter taste buds before the last division and final round of differentiation step [8]–[11]. Thus, the HES1 that we observed in cells within the taste buds (Figure 4) suggests that it may play a role in repressing TRCSMs in these progenitor cells (Figure 6). These observations support our hypothesis that Hes1 functions as a repressor of TRCSMs in taste cell precursor cells.
Despite its importance, research regarding the molecular mechanisms of the development of the taste system has lagged behind that of the other special sense organs [3]–[5]. In our investigation of the molecular regulation of taste cell differentiation, we isolated key regulators of taste cell differentiation in early development by combining computational and experimental biology.
Sharing gene expression regulatory components is an efficient way of regulating molecules within the same signal transduction pathway. TRCSMs are indeed expressed in the same population of cells, at least during the course of early taste system development. We performed in silico analysis of stretches of sequence up to 5 kb upstream of TCRSMs, some of which had been shown previously to drive TCRSM expression in taste cells in transgenic mice [25],[26]. Using computational analysis to determine which transcription factor binding sites were commonly found in the promoters of genes involved in the same regulatory network, we identified a number of putative transcription regulators. Similar procedures could be applied to analyses of other systems.
It has been proposed that the development of taste buds is regulated by epithelial-mesenchymal interactions involving several different signaling pathways, such as Notch, Shh, Wnt, and BMP [23], [27]–[32]. Recent analyses of β-catenin and Sox2 suggest that they are involved in taste cell development, although the steps involved in differentiation have yet to be clarified [27]–[29]. The expression patterns of Notch signaling pathway–related genes indicate that the Notch signaling cascade may have a role during morphological differentiation of CVP [23]. Here, we report that the number of TRCSM-positive cells is more than 5-fold greater in Hes1−/− embryos than in their wild-type littermates at stages E18 and P0 (Figure 5). Although we believe that the increase in TRCSM-positive cells observed in Hes1−/− mutants is due to premature expression of these marker proteins in the taste cell lineage, we cannot not exclude other possibilities, such as an increase in the total number of cells in CVP, or ectopic expression in cell types other than taste cells, in which expression of TRCSMs is normally repressed by HES1. Previous studies have proposed that a precursor population in the developing central nervous and hematopoietic systems expresses Hes1 to maintain its undifferentiated state, and that downregulation of Hes1 leads to differentiation [33]–[37]. Hes1 may have a similar function in the taste cell lineage, and a reduction in nuclear HES1 would thus trigger taste cell differentiation in CVP epithelium. In addition, we observed a reduction in nuclear HES1 in blood type H antigen– and SNAP25-positive cells (corresponding to type I and type III taste bud cells, respectively) in older animals (Figure 4). These observations support the possibility that Hes1 is indeed a common regulator of taste bud cell differentiation.
Our computational analysis yielded several transcription factors that may be involved in the TRCSM regulatory network (Table S2). Our investigation of HES1, one of the candidate transcription factors, provides support for the utility of the computational approach. Our list of TRCSM regulators will be a valuable resource for future studies of taste development, leading to a better understanding of the process of taste cell differentiation. Further, it may be useful for designing therapies for taste disorders, such as loss of taste.
Hes1 mutant animals were kindly provided by Ryoichiro Kageyama [37]. Developing CVP were fixed with neutralized 10% formaldehyde and embedded in paraffin. The histological protocols were described previously [38],[39]. The sections were 7.5 µm thick. Serial sections were prepared from the tongue, including entire CVP. A series of serial sections was subjected to immunohistochemistry or in situ hybridization with each combination of antibodies or probes (PLCβ2 and IP3R3, gustducin and IP3R3, Plcβ2 and Ggamma13, and Plcβ2 and Trpm5). For each combination of antibodies or probes, more than five serial section series were used for staining. Overall, tongues from more than 80 animals (four combinations of markers at stages E16, E17, P0, and P5) were analyzed to observe colocalization of TRCSMs (Figure 1).
The mouse, rat, and human sequences 5 kb upstream of the TRCSMs that we investigated were retrieved from the Ensembl v46 (Aug 2007) database: Plcβ2 (ENSMUST00000077829), Trpm5 (ENSMUST0000009390), gustducin (Gnat3) (ENSMUST000000030561), Ip3r3 (ENSMUST00000049308), and Ggamma13 (Gng13) (ENSMUST00000026836). We utilized vertebrate-specific profiles of transcription factor binding sites (TFBSs) in the TRANSFAC Professional database 11.3 (10 September 2007). We searched for putative TFBSs in the promoter sequences of five TRCSMs in mouse, rat, and human using the MATCH program (version 10.4) [22], with the option of minimizing the number of the error rates of false positives and false negatives. Among the putative TFBSs that we discovered in these cross-species searches, we identified putative TFBSs that were conserved among all three species. We also performed a database search for genes expressed during stages E16–E18 in mouse undifferentiated oral epithelium in the Mouse Genome Informatics page of Jackson laboratory (as salivary gland precursor cells and oral epithelium at Theiler's Stage (TS) 24–TS26) (http://www.informatics.jax.org/) (Figure S3).
ChIP experiments were performed in accordance with previous reports [40],[41] and a technical protocol established by the Farnham laboratory (http://genomecenter.ucdavis.edu/farnham/farnham/protocols/tissues.html). We used stage P19 embryonal carcinoma cells as a substrate for ChIP assays, and an antibody against HES1 (Chemicon, AB5702: antibody raised against a synthetic peptide).
The following primers were used: p1 pair, 5′-TGTTAGAACGCTGGAGTTCAAG-3′ and 5′-ATCAGGCTCAGCTTTCCCATG-3′; p2C pair, 5′-AAAGTCTCTCGGACACCCAGC-3′ and 5′-TCTTAGGCTGTGAGGCAGCTG-3′; ip1 pair, 5′-GAGCAGAATGAGATCCGCATC-3′ and 5′-ACTGGGTAGCTGCTGCTACAG-3′; ip2 pair, 5′-CTCATTGACACCTGGGAGGAG-3′ and 5′-GGAATCTACATCCCTCAGTGG-3′; and ip3C pair, 5′-GTTGGGTCCAGAGTCAGAGAC-3′ and 5′-CTCACCTTCTAGGATCTCAGG-3′.
We used antibodies to PLCβ2 (Santa Cruz, SC206: antibody raised against amino acids 1170–1181 of PLCβ2 of human origin), gustducin (Santa Cruz, SC395: antibody raised against amino acids 93–112 of gustducin of rat origin), IP3R3 (BD Transduction Laboratories, 610312: antibody raised against amino acids 22–230 of IP3R3 of human origin), SNAP25 (Abcam, ab24737: antibody raised against full length protein-the critical epitope lies amino-terminal of the C-terminal peptide), and human blood type H antigen (AbH) (Abcam, ab3355: antibody raised against human colon cancer cell line SW-403). An antibody against HES1 was raised in this study using a polypeptide corresponding to amino acids 24–41 of HES1 (TPDKPKTASEHRKSSKPI) to immunize a rabbit and produce anti-HES1 antibody (Figure S5). Antiserum was purified by the same polypeptide. We verified the specificity of this antibody by western blotting and immunohistochemistry on the spinal cords of wild-type and Hes1−/− embryos (Figure S5). Our anti-HES1 antibody recognized nuclear localized HES1 in neurons from the embryonic spinal cord in wild-type animals (Figure S5).
All sections were treated with HistoVT One solution (Nakalai Tesque, 06380–05) for antigen retrieval. Images were captured with LSM51 confocal microscopy (Zeiss), and their optical thicknesses are 1 µm [42].
All PLCβ2-, gustducin-, and IP3R3-positive cells were counted in 7.5 µm serial immunohistological sections from whole CVP. Immunoreactive cells were counted only when nuclear staining with DAPI was clearly observed in the same cell. Immunofluorescence that appeared at a similar position in two successive sections was counted as one positive cell. All immunoreactive cells were observed with an LSM51 confocal microscope (Zeiss) and the optical thicknesses of images are 1 µm [42].
|
10.1371/journal.pntd.0007779 | Evaluation of nitazoxanide treatment following triclabendazole failure in an outbreak of human fascioliasis in Upper Egypt | Fascioliasis is a neglected zoonosis with major public health implications in humans. Although triclabendazole (TCBZ) is the drug of choice, there are records of TCBZ failure worldwide. TCBZ-resistant fascioliasis is treated with alternative approved drugs including nitazoxanide (NTZ), with varying levels of efficacy. Data on NTZ efficacy after TCBZ failure in Egypt is scarce. This study evaluated the efficacy of NTZ in cases of TCBZ failure during an outbreak of fascioliasis in Assiut governorate of Upper Egypt.
This prospective study included 67 patients from the outpatient clinic in Manfalout locality of Assiut governorate with clinical manifestations of acute fascioliasis. These included high eosinophilia (> 6% eosinophils in peripheral blood), positive anti-Fasciola antibodies, and hepatic focal lesions (HFL) or ascites on abdominal ultrasound or computed tomography. All patients initially received TCBZ at recommended doses. Patients were followed up after 1 month to assess response. According to the responses, patients were categorized as non-responders and responders. The non-responders received a trial of NTZ and were re-assessed for response based on clinical manifestations, eosinophil count, and abdominal ultrasound. Patients not responding to NTZ received additional doses of TCBZ.
One month after initial TCBZ treatment, 37 patients responded well to TCBZ, while 30 patients failed to respond with persistence of fever, abdominal pain, high eosinophilia, and HFL. Most non-responders were male (56.7%); females predominated among TCBZ responders (62.2%). The mean age of the non-responders was relatively lower, at 20.57 ± 14.47 years (p = 0.004). Following NTZ therapy, HFL disappeared in 9/30 (30%) patients and eosinophil counts normalized in only 2 (6.7%) patients, indicating an overall efficacy of 36.6%. The remaining cases received additional doses of TCBZ with complete clinical, pathological, and radiological resolution.
Nitazoxanide was partially effective in TCBZ failure in acute human fascioliasis in Upper Egypt. Further studies with larger samples are highly encouraged and further research is urgently needed to find new therapeutic alternatives to TCBZ.
| Fascioliasis is a neglected zoonosis with major public health implications in humans. Triclabendazole (TCBZ) is the drug of choice, but alternative approved drugs are necessary in cases of TCBZ failure. Nitazoxanide (NTZ) is an alternative used in such cases. However, the efficacy of NTZ in TCBZ-failure cases among patients in Egypt remains unclear. In this study, the efficacy of NTZ was evaluated in cases of TCBZ failure during an outbreak of human fascioliasis in Assiut governorate of Upper Egypt. This study enrolled 67 patients diagnosed with fascioliasis based on clinical, laboratory, and radiological findings. These patients were referred from the outpatient clinic in Manfalout locality of Assiut governorate in Egypt. All patients received TCBZ at recommended doses as initial treatment. Those failing to respond were treated with NTZ at standard doses; following therapy, lesions in the liver and high eosinophil counts were resolved in 30% and 6.7% patients, respectively, indicating an overall efficacy of 36.6%. Therefore, in this outbreak of human fascioliasis in Upper Egypt, NTZ was found to be partially effective in cases with TCBZ failure.
| Fascioliasis has emerged as a notable zoonotic disease with considerable impact on veterinary and public health. This prompted the World Health Organization (WHO) to include human fascioliasis among the important neglected tropical diseases (NTDs) [1]. It is a foodborne disease caused by trematodes belonging to the genus Fasciola (F. hepatica and F. gigantica). In the past few decades, the incidence of human fascioliasis has considerably risen in different parts of the world. The rise has been particularly remarkable in South America, Asia, and Africa including Egypt, where the two common species of Fasciola coexist [2]. Recent studies have revealed a large number of cases of fascioliasis (2.4 to 17 million cases) worldwide [3].
Fasciolids are parasites of the hepato-biliary ducts, and the disease is mostly confined to the liver. Therefore, the main pathogenic sequelae are hepatic lesions, fibrosis, and chronic inflammation of the biliary passages. The pre-patent period together with the time to onset of signs/symptoms of the disease may range from a few days to 2–3 months or longer. There are 2 main clinical stages in fascioliasis. The acute stage coincides with larval migration and mechanical destruction of the liver tissue. This stage extends till worm maturation in the hepatic tissues, and lasts for 2–4 months. The chronic stage coincides with the persistence of adult Fasciola worms in the bile ducts and may last for months or even years [4]. Eosinophilia is the most common clinico-pathological feature against fascioliasis in both stages.
In Egypt, fascioliasis has probably been prevalent for a very long period, since the times of the pharaohs [5, 6]. High levels of infestation have been widely described in livestock, [7] resulting in considerable economic losses and expenditure for purchase of anthelmintics, liver condemnation, loss of production due to mortality, lower production of meat, milk, and wool, reduced weight gain, and impaired fertility [8].
The mainstay of treatment in fascioliasis affecting animals and humans is triclabendazole (TCBZ), which targets both the immature stages and mature adult worms [9]. Older drugs, such as tetrachloride, tetrachlorethylene, and bithionol, are currently considered to be less effective, unacceptably toxic, or both [10]. Although TCBZ is the only effective treatment for fascioliasis, it is currently registered for human use in only 4 countries [11].
The widespread use of TCBZ in the livestock industry led to the emergence of resistance in fluke populations affecting ruminants in both, developed and developing countries including Ireland, Spain, Australia, Peru, and Argentina [10]. The zoonotic nature of fascioliasis may raise concerns regarding the transmission of resistant strains to humans, particularly in endemic areas such as Peru, Bolivia, and Egypt [12]. In recent years, a few reports have described the occurrence of TCBZ resistance in humans. The first case was reported in a livestock farmer in the Netherlands, followed by 4, 1, and 7 cases in Chile, Turkey, and Peru, respectively [13–16]. Unfortunately, despite the prevalence of TCBZ resistance in Egypt, a review of the literature does not reveal any published data.
Reliance on monotherapy poses a risk for the treatment of fascioliasis, particularly in the absence of a vaccine for the prevention of the disease [9]. As cases of TCBZ resistance are continuously being documented from livestock, human cases of TCBZ-resistant fascioliasis are most likely to occur. This is a serious challenge for treatment in humans, with considerable public health implications [8] and emphasizes the urgent need for developing new fasciocidal drugs [17].
Several trials were conducted in the search for effective alternative drugs for fascioliasis. Nitazoxanide (NTZ), which is a broad spectrum antiparasitic agent, has been found to be well tolerated by humans, with adverse effects similar to that of placebo [10]. Across different studies, its efficacy has ranged from 40–100% [18].
The aim of the present study was to investigate the efficacy of nitazoxanide as a treatment for fascioliasis in the face of incomplete response to triclabendazole in Upper Egypt.
This prospective study was conducted between August and November 2018.
The study protocol was approved by the Institutional Review Board of the Faculty of Medicine of Assiut University, Egypt. Written informed consent was obtained from all patients prior to participation in this study.
A total of 74 patients with diagnosed or suspected fascioliasis were recruited in the study. All these patients were referred to the outpatient clinic in the Tropical Medicine and Gastroenterology Department at the Al-Rajhi Liver University Hospital during an outbreak of fascioliasis in Manfalout locality of Assiut Governorate in Upper Egypt.
The included patients had symptoms and signs suggestive of fascioliasis such as fever, abdominal pain, jaundice, and hepatomegaly. The complete blood count (CBC), including eosinophil percentage and absolute eosinophil count was individually assessed using the ADVIA 2120i Hematology System (Siemens Healthcare Diagnostics Inc. Tarrytown, NY 10591, USA). Stool examination was also performed for all cases for the qualitative diagnosis of fascioliasis using the native lugol and formalin ethyl acetate sedimentation method [19]. Stool samples were also examined for the presence of other co-existing intestinal parasites that could potentially cross-react or overlap with fascioliasis. Liver function tests were also performed.
Further investigations included; serological analysis was done by F. hepatica IgG Enzyme-linked immune sorbent assay (ELISA) kits (DiaColon Tech Houston, USA) for qualitative diagnosis of fascioliasis. The result was read photometrically at 450 nm (TECAN Sunrise Absorbance Reader). (values greater than 10.0 AU/ml were interpreted as seropositive, cut-off value 0.25 according to the manufacturer’s instructions) Indirect hemagglutination assay (IHA) using Distomatose Fumouze (Laboratories Fumouze Diagnostic, Levallois Perret, France) was also done to compare antibody titers (a titer ≥ 1/320 was considered to be positive). Abdominal ultrasound (US), and abdominal computerized tomography (CT) were also done. Endoscopic retrograde cholangiopancreatography (ERCP) was performed in cases presenting with obstructive jaundice and a dilated common bile duct (CBD) on abdominal US and/or CT.
All patients received a double dose of triclabendazole (Egaten, Novartis Pharma AG) at a dose of 10 mg/Kg/dose, at 12-hour interval in a joint venture with the WHO. Patients were advised to avoid vegetables that posed a risk for re-infection.
The endpoints for treatment response were evaluated on follow up after 1 month. Evaluation was based on 3 parameters, namely, resolution of clinical symptoms and signs, normalization of eosinophil counts, and improvement of hepatic lesions on US. According to the WHO criteria, persistence of symptoms or signs with either eosinophilia (> 6% eosinophils in peripheral blood) or hepatic focal lesions, was considered to be a probable indicator of treatment failure with TCBZ [1]. Patients were then divided into 2 groups according to treatment response. The first group included the patients who did not respond to TCBZ and were administered NTZ (non-responders), while the second group included patients who successfully responded to TCBZ (responders).
The non-responders received NTZ at a dose of 500 mg orally every 12 hours for 7 days. Patients were clinically assessed for response after 1 month. Resolution of both, eosinophilia in the CBC and/or hepatic focal lesions on US were indicative of response. Patients who failed to respond to NTZ were re-treated with TCBZ at doses similar to the initial dose and were followed up for response.
Patients who received any other anthelminthic drugs within 1 month before TCBZ or NTZ therapy including bithionol, praziquantel, albendazole, dihydroemetine, or emetine hydrochloride, and patients who showed hypersensitivity to nitazoxanide were excluded from this study.
Data entry and analysis were performed using the IBM SPSS Statistics for Windows, Version 20.0. (Armonk, NY: IBM Corp) software package. Data were presented as numbers, percentages, means, and standard deviations. The Chi-square and Fisher’s exact tests were used to compare qualitative variables. The Mann-Whitney test and the Wilcoxon signed rank test were used to compare variables between independent and dependent groups, respectively. In case of non-parametric data, the Wilcoxon signed rank test was used to compare the quantitative variables before and after treatment. A P-value < 0.05 was considered statistically significant.
In this prospective study, 74 patients were initially recruited. Among them, 67 patients with symptoms and signs suggestive of acute fascioliasis were included for the NTZ trial; 7 patients were excluded as they presented with obstructive jaundice and a dilated CBD on ultrasound (suggesting chronic fascioliasis). These 7 patients underwent endoscopic sphincterotomy and extraction of the adult worm by ERCP followed by TCBZ therapy. The included patients received initial treatment with a double-dose of TCBZ.
The pretreatment demographic, clinical, and laboratory data of the studied patients are shown in Tables 1–3, respectively. The cohort comprised 31 male and 36 female patients with a mean age of 26.27±15.3 (range: 4–60) years. The patients presented with one or more of the symptoms and signs of acute infection, which include fever, abdominal pain, hepatomegaly, splenomegaly, and ascites. Laboratory data showed mild anemia (hemoglobin [Hb]: 11.8± 0.7 g/dl), high eosinophilia (41.1 ± 15.7%), high alanine transaminase (ALT) and aspartate transaminase (AST) levels, and a positive serological titer (936.1±387.2). As depicted in Fig 1, radiological investigations showed the presence of hepatic focal lesions (HFL) in 25 patients (37.3%). Stool examination was positive for Fasciola eggs in 7 of 67 patients (10.4%) with absence of other co-existing parasitic infections that could, potentially, construct immunological cross-reactions or clinical symptoms overlapping with fascioliasis.
The studied patients were followed up after 1 month to evaluate the response to first line TCBZ. A total of 37 cases (55.2%) showed good response to TCBZ (the responder group) as evidenced by disappearance of signs and symptoms, normalization of peripheral eosinophil counts, and resolution of HFL. The remaining 30 cases (44.8%) (the non-responder group) showed persistence of infection, as evidenced by persistence of clinical manifestations, high eosinophilia, and HFL. This group received nitazoxanide and were followed up after 1 month (Fig 2).
The demographic, clinical, and sonographic characteristics of both groups, as summarized in Table 4, showed that most patients in non-responder group were male (56.7%), while females were predominant in the responder group (62.2%). Also, the mean age in the responder group (30.89 ± 14.57 years) was significantly higher than that of the non-responder group (20.57 ± 14.47 years) (p = 0.004). However, the clinical presentation and sonographic evidence of HFL were not significantly different between the groups.
As shown in Table 5, the hematological, biochemical, serological and parasitological parameters of patients at baseline were not considerably different between patients in both groups, except for total leucocyte count, and levels of ALT and AST, that were significantly higher in the non-responder group (p = 0.008, p = 0.026, and 0.047, respectively).
Furthermore, the assessment of response to first line TCBZ showed complete resolution of the clinical manifestations in all patients in the responder group; patients in the non-responder group had persistent fever and abdominal pain. The pre-treatment eosinophil counts were not significantly different between the groups (p = 0.081). After treatment, limited improvement in eosinophil counts was observed in the non-responder group, with a reduction from 26.72% ± 13.21 to 20.00% ± 11.28. In the responder group, the counts reduced from 30.47% ± 15.18 to 3.6% ±1.7, showing statistically significant difference between the groups (p = 0.000).
After NTZ treatment in the non-responder group, HFL disappeared in 9/30 patients (30%) as opposed to all patients in the TCBZ responder group; this difference was statistically significant (p = 0.015) (Table 6). In addition, eosinophil counts normalized in only 2 (6.7%) patients after NTZ therapy. Patients who did not show improvement after NTZ therapy received an additional dose of TCBZ, similar to the initial dose, with complete clinical, laboratory, and radiological resolution. Therefore, based on the improvement of eosinophil counts and HFLs in patients with TCBZ failure, nitazoxanide was effective in 11/30 patients (36.6%).
Owing to its activity against juvenile and adult forms of the parasite, TCBZ is the drug of choice in the treatment of F. hepatica and F. gigantica infections in humans [1]. Mass control programs for human fascioliasis in Egypt, Vietnam, Bolivia, and Peru have used TCBZ, which was donated through an agreement between the WHO and the manufacturer [20].
Several previous studies have documented the clinical efficacy of TCBZ with various treatment regimens in different regions including Egypt [21–23]. The results of these drug trials are indicative of a dose–response relationship. The WHO currently recommends the administration of a single dose of TCBZ at a dose of 10 mg/kg for the treatment of human fascioliasis, and a double dose of 10 mg/kg, 12 hours apart, in severe cases [1]. In a randomized open-labeled study conducted in Egypt, which compared 1- and 2- dose regimens of TCBZ at 10 mg/kg, the 2-dose regimen showed more favorable results [21].
Indeed, TCBZ is the only first-line medication with reports of high efficacy in humans. Therefore, the effective management of resistance to this drug is of utmost importance [9]. Clinical trials on alternatives to TCBZ are limited. This is probably the first study to evaluate the efficacy of NTZ in the management of cases of acute fascioliasis with TCBZ failure in Egypt.
In the current study, all cases of acute fascioliasis were defined based on clinical manifestations, high eosinophilia, and radiological signs with positive anti-Fasciola antibodies. However, stool examination was positive in only 7 cases (10.4%) with a low egg burden. In the present study we could not rely only on coprological examination for the diagnosis and follow up of cases. This is attributed to many factors, including prepatent or acute infections (where the patients were symptomatic prior to the appearance of eggs in the stool) [24], the inability of adult Fasciola worms to produce eggs (due to its lack of adaptation to the human host), encapsulation of eggs in granulomas or abscesses in the liver, and low egg shedding related to low infection burdens [25]. Coprological examination may also overestimate the response to treatment since the age of the fluke or its anatomic location, which may be associated with increased susceptibility to treatment, may impact the results [26].
In a study previously conducted in Egypt including 23 cases, Fasciola eggs were detected in only 2 cases (8.6%) as the patients were diagnosed in the hepatic phase [27]. An immature worm feeds on liver tissue without producing eggs; the only evidences of infection are eosinophilia and HFL, which are observed in early stages of the infection [28].
The detection of anti-Fasciola antibodies by the ELISA test is a reliable and sensitive test for diagnosis of fascioliasis compared with stool examination. The main advantage is that results are positive as early as 2 weeks post infection. However, since serum antibodies may persist for 4–5 months after successful treatment, it is not a reliable test in the evaluation of response during follow up [24]. Eosinophilia as a host defense mechanism is a common feature of fascioliasis and is encountered in 14%- 82% of patients, and may rise and fall during the chronic stage [29].
As described by Marcos et al., the primary outcome measures for clinical cure after treatment are defined by resolution of the clinical picture and eosinophilia during follow up [30]. Therefore, in the current study, post-treatment follow-up was based mainly on the persistence of clinical manifestations with either high eosinophilia with or without radiological signs.
In the present study, as evidenced by the disappearance of signs and symptoms, normalization of peripheral eosinophil counts, and resolution of HFL, 37 patients (55.2%) showed good response to TCBZ. The remaining 30 cases (44.8%) were suspected to have TCBZ failure and were treated with NTZ. The mean age of the non-responder group was lower than that of the TCBZ responders; this may have had an impact on the treatment response. A double blinded placebo-controlled trial in Peru, which employed NTZ for the treatment of chronic fascioliasis, has shown a low cure rate in children (40%) and a slightly higher efficacy in adults (60%) [9].
In our cohort, most non-responders were male (56.7%); females were predominant among the responders (62.2%). The gender of the studied patients did not significantly differ between groups. However, previous studies have indicated differences in sensitivity to flukicides depending on the sex of the host animals infected with F. hepatica [31].
Notably, in the current study, patients who did not initially respond to TCBZ in the acute stage, responded to the subsequent trial of TCBZ administered 2 months after the initial dose, in the chronic stage. This relationship between response to TCBZ and the stage of the disease has also been previously mentioned by Marcos et al. [30], who reported the amelioration of eosinophil counts after a single dose of TCBZ in 10 patients with acute Fasciola infection. However, parasitological cure (the absence of eggs in the stools) was not reported during follow up.
The difference in TCBZ susceptibility between juvenile and adult parasites has been previously described in an in vitro study with Fasciola hepatica infection [17]. However, this has not been thoroughly described in case series including patients with acute fascioliasis [32, 33].
According to our results, 30 patients showed clinical evidence of the presence of TCBZ-resistant F. hepatica infection, which is considered a large number. They received a trial with NTZ at a dose of 500 mg twice daily for 7 days, that showed an overall efficacy of 36.6% (11/30 patients), based on the improvement of eosinophil counts and HFLs.
NTZ has been widely used in the management of different parasitic infections with reportedly high efficacy and tolerability. The efficacy of NTZ against Fasciola has been studied in rabbits experimentally infected with F. gigantica. NTZ was found to be partially effective (47%) against the juvenile stages of the parasite, but completely effective (100%) against the adult stage [34].
A few clinical trials have been conducted on the efficacy of NTZ in the treatment of human fascioliasis with considerably variable efficacy. In Egypt, an open-label clinical study including 125 Egyptian patients with chronic fascioliasis demonstrated 97% clearance of F. hepatica eggs in the stool on day 30 after treatment with NTZ; the serological and eosinophilic patterns had also improved [35]. A second report from Egypt showed a slightly lower cure rate with NTZ (82.4%) [36]. Similar results were observed in a study conducted on schoolchildren in Mexico that documented the efficacy rates of NTZ against chronic fascioliasis to be 94.0% and 100% after the first and second treatment courses, respectively [37].
A much lower efficacy rate was observed in a double-blinded placebo-controlled study in northern Peru, where 50 adults and 50 children infected with F. hepatica received a 7-day course of NTZ. Compared to the placebo group, 60% adults and 40% children were cured [10]. These results suggest that NTZ may be a reasonable option at least in the chronic stage of fascioliasis, and is a good alternative to TCBZ.
Conversely, some studies have revealed a lack of efficacy of NTZ in 24 cases of liver fluke infection in Cuba [38] and in a patient with apparent TCBZ failure in the Netherlands [13]. Cabada and colleagues have reported that a cohort of 7 patients, infected by ingesting watercress in the Cusco region of Peru, had failed to respond to multiple courses of TCBZ in combination with NTZ [16]. The wide variances in fasciolid susceptibility to NTZ may be attributed to differences in geographical strains of Fasciola in various regions [16]. This indicates the urgent need for further controlled clinical trials to evaluate the efficacy of NTZ in the control of fascioliasis.
Although TCBZ resistant fascioliasis has been widely described in livestock, the understanding of the mechanism of resistance to TCBZ remains incomplete, with a knowledge gap in terms of its capacity to spread and strategies for control [39]. It has been suggested that resistant fasciolid strains may have alterations in drug uptake, efflux, and detoxification, including the conversion of TCBZ sulfoxide into the less active forms. However, this has not been verified in large studies using other parasite strains. Poor response to TCBZ may also be attributed to its poor water-solubility and limiting drug concentration in the organs [40–42].
In contrast to veterinary medicine where other treatment options for Fasciola exist, there is no documented strategy for the management of TCBZ treatment failure in humans. To minimize the development of drug resistance, the use of synergistic drug combinations has been suggested [43]. However, this approach carries the risk of building up resistance to multiple drugs [44].
Although the small sample size has limited the scope of this study, to the best of our knowledge, this is the first report of TCBZ failure in humans with acute fascioliasis in Egypt. Further multicenter randomized studies including larger sample sizes are required to evaluate predictors of TCBZ failure. This will help to determine the optimum timing for repeating TCBZ after failure of the initial dose. Also, further research is urgently needed to find new therapeutic alternatives to TCBZ for controlling fascioliasis.
In this first report of TCBZ failure in acute human fascioliasis in Upper Egypt, NTZ proved to be partially effective.
|
10.1371/journal.pgen.1007269 | Glial loss of the metallo β-lactamase domain containing protein, SWIP-10, induces age- and glutamate-signaling dependent, dopamine neuron degeneration | Across phylogeny, glutamate (Glu) signaling plays a critical role in regulating neural excitability, thus supporting many complex behaviors. Perturbed synaptic and extrasynaptic Glu homeostasis in the human brain has been implicated in multiple neuropsychiatric and neurodegenerative disorders including Parkinson’s disease, where theories suggest that excitotoxic insults may accelerate a naturally occurring process of dopamine (DA) neuron degeneration. In C. elegans, mutation of the glial expressed gene, swip-10, results in Glu-dependent DA neuron hyperexcitation that leads to elevated DA release, triggering DA signaling-dependent motor paralysis. Here, we demonstrate that swip-10 mutations induce premature and progressive DA neuron degeneration, with light and electron microscopy studies demonstrating the presence of dystrophic dendritic processes, as well as shrunken and/or missing cell soma. As with paralysis, DA neuron degeneration in swip-10 mutants is rescued by glial-specific, but not DA neuron-specific expression of wildtype swip-10, consistent with a cell non-autonomous mechanism. Genetic studies implicate the vesicular Glu transporter VGLU-3 and the cystine/Glu exchanger homolog AAT-1 as potential sources of Glu signaling supporting DA neuron degeneration. Degeneration can be significantly suppressed by mutations in the Ca2+ permeable Glu receptors, nmr-2 and glr-1, in genes that support intracellular Ca2+ signaling and Ca2+-dependent proteolysis, as well as genes involved in apoptotic cell death. Our studies suggest that Glu stimulation of nematode DA neurons in early larval stages, without the protective actions of SWIP-10, contributes to insults that ultimately drive DA neuron degeneration. The swip-10 model may provide an efficient platform for the identification of molecular mechanisms that enhance risk for Parkinson’s disease and/or the identification of agents that can limit neurodegenerative disease progression.
| Glutamate (Glu) is an important signaling molecule used by nerve cells to communicate information, although excessive Glu signaling can overexcite neurons to the point where they degenerate, a phenomenon termed excitotoxicity. Glu induced excitotoxicity has been linked to neurodegeneration arising in the context of stroke, amyotrophic lateral sclerosis (ALS) and Parkinson’s disease (PD). Glial cells, that surround neurons, and their processes have been shown to limit Glu-induced excitotoxicity in mammals. Here, we demonstrate that C. elegans glia limit progressive degeneration of dopamine (DA) neurons that arises in the context of mutation of the protein, SWIP-10, and that this degenerative process relies on Glu signaling, altered Ca2+ homeostasis and apoptotic pathway genes. Our findings reveal a novel molecular contributor to glial maintenance of DA neuron viability, provide a genetically-tractable example of Glu-dependent cell death, and encourage further evaluation of SWIP-10 linked pathways for mechanistic insights into neurodegenerative diseases and their treatment.
| Across phylogeny, the amino acid glutamate (Glu) plays multiple, important roles including contributions to protein synthesis, intermediary metabolism, and chemical neurotransmission [1–4]. At neuronal synapses, Glu signals through both metabotropic receptors that initiate G-protein coupled signaling [5–7] as well as ionotropic receptors that flux ions such as Na+ and Ca2+, altering membrane excitability [5, 8–10]. Excessive ionotropic Glu signaling in the mammalian brain has been implicated in a variety of brain disorders including addiction, schizophrenia, amyotrophic lateral sclerosis (ALS), and Parkinson’s disease (PD) [11–14], as well as the neuronal death that arises in the context of stroke and glioblastoma [15, 16]. Acute treatment of neurons with high, non-physiological, levels of Glu can induce signs of cell death within minutes, characterized by intense vacuolization and cell swelling characteristic of necrosis [17–20]. Chronic hyper-activation of neurons by Glu, within physiological limits, can drive apoptotic mediated neural degeneration, particularly if other genetic or environmental risk pathways are engaged [21–23]. Glu activation of Glu receptors can lead to prolonged alterations in intracellular Ca2+ homeostasis, driving Ca2+-dependent proteolysis and activation of apoptotic programs [24].
Although cell autonomous mechanisms remain a focal point for many investigations seeking insights into determinants of neurodegeneration, increasing attention has been given to astrocytic mechanisms that can sustain neuronal viability, in the context of constant Glu stimulation that could otherwise lead to cell death. These mechanisms include the shuttling of metabolic intermediates such as lactate to neurons that can help sustain ATP synthesis [25–27], the buffering of extracellular ions such as K+, since excess extracellular K+ due to chronic ion channel activation and Na+/K+ ATPase dysregulation can contribute to excess neuronal activation [26, 28, 29], and the efficient clearance of extracellular Glu that both limits the amplitude of synaptic and extrasynaptic Glu signaling but also Glu-driven neuronal degeneration [26, 30, 31]. Astrocytic Glu clearance is mediated by multiple Na+-dependent Glu-transporters of the SLC1 family (e.g. GLT1/rodents, EAAT2/humans) that terminate Glu signaling via binding and uptake of Glu in proximity to synaptic release sites [13, 31, 32]. A second astrocytic Glu transporter that participates in extracellular Glu homeostasis is xCT (SLC7A11), the transporter subunit of a dimer that supports intracellular Glu exchange for extracellular cystine. xCT is generally thought to act oppositely to SLC1 transporters, balancing control of extrasynaptic Glu levels with the provision of precursor (cysteine) for astrocytic glutathione synthesis [33–36].
Due to their significant impact on synaptic and extrasynaptic Glu homeostasis, Glu transporters and exchangers have been widely studied to determine their contribution to Glu-induced neural degeneration as well as in efforts to manipulate their activity and expression for therapeutic ends [36–38]. For example, Rothstein and coworkers identified β-lactam antibiotics, typified by the cephalosporin-type agent ceftriaxone (Cef), as capable of elevating GLT1 expression in vitro and in vivo, protecting neurons from Glu toxicity, and enhancing longevity in an ALS mouse model [13]. Subsequently, many investigators have demonstrated the neuroprotective activity of Cef administration in rodents [39–41], with evidence supporting antibiotic modulation of both GLT1 and xCT expression [13, 36, 42], although candidates targeted by the antibiotic in glia to induce transporter expression have, until recently, been unidentified.
In a screen for novel genes that control DA signaling in the nematode, C. elegans [43], we identified a glial-expressed gene, swip-10, whose mutation induces hyper-excitability of DA neurons and elevates rates of vesicular DA release, culminating in the hyperdopaminergic phenotype, Swimming induced paralysis (Swip) [44]. These studies also demonstrated a critical role for Glu signaling in establishing the paralytic phenotype of swip-10 mutants [44]. Swip-10 is conserved across phylogeny, with the unstudied gene, Mblac1, as the putative mammalian ortholog. Both SWIP-10 and MBLAC1 proteins are metallo β-lactamase domain (MBD)-containing proteins [44], with residues key for metal binding and catalysis conserved across worm and vertebrate proteins. Although the substrate hydrolyzed by SWIP-10/MBLAC1 enzymatic activity is currently unknown, we recently established that MBLAC1 is a specific, high-affinity target for Cef [45]. We presented evidence that Cef binding activity in brain lysates could be totally eliminated by MBLAC1 immunodepletion therefore supporting the hypothesis that MBLAC1 may be the exclusive, non-microbial target of Cef in vivo. These findings also suggest that further study of SWIP-10/MBLAC1 may reveal mechanisms normally engaged to protect neurons from chronically elevated extracellular Glu and a path to the identification of novel neuroprotective agents. A key piece of data lacking in this hypothesis, however, is evidence that loss of SWIP-10/MBLAC1 either induces Glu-dependent neural degeneration or eliminates the neuroprotective actions of Cef.
Here, we capitalize on the ease of monitoring the morphology and degeneration of C. elegans DA neurons engineered to stably express green fluorescent protein (GFP) to examine a requirement for swip-10 expression in limiting Glu-dependent DA neuron degeneration. We find that swip-10 mutants demonstrate a striking, progressive degeneration of DA neurons that can be suppressed by glial expression of wild type swip-10 and by mutation of Ca2+ permeable Glu receptor mutants. Through our studies, we provide evidence that a cell non-autonomous action of SWIP-10 sustains DA neuron viability in the context of excess Glu signaling and elevations of cytosolic Ca2+ that we hypothesize leads to increased cellular stress and, ultimately, apoptotic cell death. Our findings support SWIP-10 (and by extension MBLAC1) as a key protective agent whose further study may yield important insights into risk factors for progressive neurodegenerative disorders and their treatment.
Given the Glu signaling-dependent, Swimming-induced paralysis (Swip) phenotype present in swip-10 mutants [44], and evidence from the latter study that swip-10 DA neurons are hyper-excitable, as assessed by a cytoplasmic Ca2+ reporter (GCamp), we sought to determine whether these animals might display signs of excitotoxic neural degeneration. We examined the DA neurons of multiple mutant swip-10 alleles crossed to BY250, a strain that stably expresses the integrated transcriptional fusion pdat-1::GFP (vtIs7) (Fig 1A) [46]. We focused our evaluations on CEP DA neurons, and quantitatively evaluated degeneration by three distinct morphological assessments: 1) neurite truncations and breaks in GFP-labeled dendrites (Fig 1B and 1C), 2) shrunken cell soma (Fig 1D) and 3) missing cell soma (Fig 1E), as previously described [47, 48]. From these categories, we also calculated an overall degeneration score where the appearance of any of the components qualifies an animal as displaying CEP degeneration [48]. We found that all three available swip-10 alleles (vt29 and vt33 from our forward genetic screen, and the larger deletion allele, tm5915) exhibited elevations in the degeneration index, relative to wildtype animals (Fig 1F–1I). To further support that mutation of swip-10 induces morphological changes in DA neurons, versus a sequestration or inactivation of cytoplasmic GFP, we corroborated our findings using a DA neuron-targeted, membrane-bound reporter (pdat-1::myrRFP) which also yielded evidence of tm5915 DA neurodegeneration (S1 Fig). Interestingly, evaluation of swip-10 impact on C. elegans glia broadly (marked by the ptr-10 promoter driven myrRFP) or on CEPsh glia that ensheath CEP DA neurons specifically (marked by phlh-17::GFP) failed to reveal evidence for gross morphological changes (S2 Fig). These findings suggest that swip-10 mutation induces a localized, cell non-autonomous effect on the integrity of neighboring DA neurons.
To be sure that our fluorescent reporters of DA neuron morphology were faithfully reporting structural changes in DA neurons, we assessed CEP cilia of swip-10 via electron microscopy (EM). Previously, we used this approach to document damage to CEP dendrites in the context of 6-OHDA induced DA neuron degeneration [49]. The tm5915 deletion allele was selected for EM studies of swip-10 induced neural degeneration, though as noted above, all mutants demonstrated comparable degeneration. The morphology of CEP neuronal processes is well characterized at the ultrastructural level [50], especially the specialized cilium at the tip of the CEP dendrite, which can be visualized in transverse thin sections through the lips of adult C. elegans (Fig 2A) [51, 52]. Using relative position and the defined morphological characteristics of CEP DA neurons, such as the electron dense cuticular branch or nubbin associated with their cilia to anchor the dendrite to the cuticle [52] and the presence of the electron dense clumps of tubule-associated material (TAM) previously shown to be characteristic of CEP cilium [51], we were able to identify multiple anomalies in tm5915 CEP structure. These defects include changes in the size and appearance of the nubbin (Fig 2B), loss or misplacement of TAM and microtubules (Fig 2C and 2D), and the presence of large or small vacuoles in several locations either below or above the axoneme (Fig 2E and 2F). A summary of the swip-10 mutant CEP cilium defects is depicted in Fig 2G. In addition to the defects described above, half of the CEP dendrites of swip-10 mutants were missing any cilium that extended beyond the axoneme. These TEM studies confirm that swip-10 mutation results in striking DA neuron morphological changes.
In order to determine whether the DA neuron degeneration observed in swip-10 animals represents a late onset phenomenon and/or might arise from a progressive perturbation across development, we assayed DA neuron degeneration in swip-10 mutants across various post-embryonic ages. We observed that tm5915 animals display time-dependent indications of DA neuron degeneration that are distinct from the changes seen with wildtype animals (Fig 3D). In wildtype animals, signs of DA neuron degeneration are evident only in older, adult animals whereas signs of degeneration are already evident in tm5915 animals by day 1 (L1 stage) of larval development (Fig 3D). A breakdown of the components that comprise the overall degeneration score of tm5915 mutants is revealing, where non-uniform patterns are evident across measures. Although we were unable to follow individual DA neuron morphological changes over time, our population findings are suggestive of a progressive form of degeneration at the single neuron level, with dendritic breaks and truncations as earliest signs of degeneration (Fig 3A), followed by the appearance of shrunken soma (Fig 3B), and then by missing soma (Fig 3C). Overall similar patterns are evident with wildtype animals, just appearing much later in life. Together our findings indicate that swip-10 mutation begins to disrupt the health of DA neurons early in development with the appearance of indices of morphological perturbations arising in distal processes that progress to neuronal death.
Although the Swip behavior of swip-10 mutants at the L4 stage arises as a consequence of excess DA signaling [43], this paralysis is a cell non-autonomous consequence of glial, and not DA neuron, expression of swip-10 [44]. To determine whether the degeneration of DA neurons is similarly a consequence of mutation of glial swip-10, we expressed a full length wild type swip-10 cDNA fused to GFP (swip-10::GFP) under control of glial and DA neuron promotors. Fig 4A demonstrates that pan-glial swip-10 expression, as achieved through use of the ptr-10 promoter [53], robustly rescues DA neuron degeneration of tm5915 animals, comparable to that achieved with a genomic construct that encodes swip-10 and the upstream elements needed to achieve full rescue of Swip [44]. Significant rescue of DA neural degeneration was also achieved with the CEP sheath glia-specific promotor hlh-17 [53]. In contrast, DA neuron specific expression of swip-10, driving expression with the dat-1 promoter, failed to restore normal morphology. Together, these findings support the conclusion that glial expression of swip-10 is required to maintain the normal viability of DA neurons.
Although not explored extensively, we sought to understand whether neural degeneration in swip-10 mutant animals is limited to the DA neurons. We chose to evaluate swip-10 mutant (tm5915) animals bearing reporters to demarcate OLL and BAG neurons. Glutamatergic OLL neurons are similar in location and morphology to dopaminergic CEP neurons, are mechanosensitive like CEP neurons, and share an association with glia (OLLsh) that ensheath OLL processes. Carbon-dioxide sensing, glutamatergic BAG neurons are similar in location and morphology to the CEP neurons, although not associated with direct ensheathing or socket glia. We observed degeneration of glutamatergic OLL neurons (Fig 4B) but not of BAG neurons (Fig 4C). These findings, along with rescue of DA neurodegeneration through glial re-expression of swip-10, reinforce a key role for glia in maintaining the viability of C. elegans DA neurons.
Mechanisms proposed to support DA neuron degeneration include mishandling of intracellular DA that can form cytotoxic quinones [54, 55]. Thus, elevations in cytosolic DA that arise with pharmacological blockade of the vesicular monoamine transporter (VMAT, cat-1 in C. elegans) by reserpine results in DA neuron degeneration [48], and a genetic reduction of VMAT2 expression causes progressive DA neuron degeneration in mammals [56]. The degeneration of DA neurons in swip-10 animals does not appear to arise as a consequence of elevations of intracellular DA as disruption of DA synthesis capacity arising from a loss of function mutation in cat-2, the C. elegans ortholog of tyrosine hydroxylase, the rate-limiting step in DA synthesis, did not alter tm5915 DA neuron degeneration (S3A Fig). Extracellular DA elevations can lead to the formation of toxic adducts with vital cell proteins [57] and our prior studies support excess DA secretion in swip-10 animals [44]. However, loss of extracellular DA clearance capacity achieved via mutation of the presynaptic DA transporter, dat-1, which triggers Swip, [58] did not induce DA neuron degeneration (S3B Fig).
Neural degeneration, more generally, can be triggered by extrinsic or intrinsic activation of cell death genetic programs, first elucidated at a molecular level in C. elegans [59–62]. Additionally, disruptions of vital cellular processes (e.g. ATP production, membrane permeability, ion gradients or cytoskeletal organization) by genetically encoded neurotoxins or following exposure to reactive chemical species [63–65] have been shown to lead to the death of neurons. Lastly, excitotoxicity, a form of neurodegeneration with features of both apoptotic and necrotic cell death, is well known in mammalian brain preparations and typically observed in the context of over stimulation of Glu-responsive, ionotropic receptors [10, 65, 66]. Our prior findings that DA neurons in swip-10 animals display elevated excitability that is dependent on Glu signaling [44] encouraged our consideration of the latter mechanism of DA neuron degeneration. We therefore quantified DA morphological changes in swip-10 animals bearing loss of function mutations in genes supporting synaptic Glu packaging and Glu signaling, as well as mutations in genes encoding transporters thought to modulate extracellular Glu levels. First, we examined contributions of vesicular Glu transporters (vGLUTs). These proteins are responsible for packaging Glu into synaptic vesicles prior to release [67, 68]. There are three genes that encode proteins homologous to VGLUTs in C. elegans (eat-4, vglu-2 and vglu-3) [69–71] with eat-4 being the only one functionally characterized to date [68, 72]. Loss of individual vGLUTs (Fig 5A) had no effect on DA neuron morphology. Interestingly, whereas eat-4 mutation significantly reduced the Swip behavior of swip-10 mutants [44], the same eat-4 allele failed to blunt the degeneration of DA neurons in tm5915 animals. vglu-2 mutation was also unable to reduce DA neuron degeneration. In contrast, loss of vglu-3 significantly, suppressed DA neuron degeneration (Fig 5A), suggesting a contribution of vesicular Glu signaling, directly or indirectly, to swip-10 DA neuron degeneration.
Mammalian glia express multiple Na+-dependent Glu transporters (GLTs) of the SLC1 family that support efficient clearance of Glu after release at synapses and their dysfunction figures prominently in investigations of Glu-dependent neuronal injury and death [31]. Additionally, our previous studies [44] demonstrated that mutation of several GLTs (glt1, glt3 and glt4) conferred DA-dependent Swip. However, we found that mutation of individual glt genes failed to induce DA neuron degeneration (Fig 5B).
A second, glial Glu transport system, xCT, regulates extra-synaptic Glu levels, acting as a cystine/Glu exchanger [36]. xCT imports extracellular cystine in exchange for intracellular Glu, and thus altering the expression or activity of this transporter can modulate extracellular Glu levels. xCT is a member of the mammalian heteromeric amino acid transporter (HAT) family, for which there are 9 C. elegans homologs, with the highest homology for xCT being to AAT-1 and AAT-3 [73]. To determine whether xCT-like proteins could contribute to DA neuron degeneration, we generated tm5915 double mutants with all available aat mutants. Of the 7 xCT homologs tested, we found that loss of aat-1 uniquely attenuated the DA neuron degeneration of tm5915 (Fig 5C). These findings implicate non-vesicular Glu release as a contributor to swip-10 DA neuron degeneration. To determine if both vesicular Glu release supported by VGLU-3 and transporter-mediated Glu release supported by AAT-1 act in parallel or via a shared pathway to support DA neuron degeneration, we examined DA neuron morphology in an aat-1;vglu-3 double mutant. We found no enhancement of the suppression of the tm5915 degeneration beyond that of the individual mutants (S4 Fig). These findings are consistent with common mechanisms, downstream of extracellular Glu availability through either vesicular or non-vesicular Glu secretion mechanisms, as determinant of the quantitative extent of swip-10 DA neuron degeneration.
Post-synaptically in both vertebrates and nematodes, Glu binds and activates ionotropic and metabotropic receptors (iGluRs and mGluRs, respectively) [74, 75]. To further pursue the hypothesis that mutation of swip-10 triggers DA neuron degeneration via excess Glu signaling, we examined DA neuron morphology in tm5915 lines bearing available mutant alleles for the iGluRs and mGluRs. Among the twelve GluR mutants tested, we found that loss of either the NMDA-type iGluR, nmr-2, or loss of the AMPA-type iGluR, glr-1 [76], significantly suppressed swip-10 DA neuron degeneration (Fig 6A). Interestingly, these GluRs are distinct from the GluRs previously shown to suppress the paralysis phenotype of swip-10 mutants (glr-4, glr-6 and mgl-1) [44]. A double mutant of glr-1 and nmr-2 did not further suppress tm5915 degeneration beyond that seen with either mutant alone, suggesting that these receptors support neurodegeneration through a common pathway (Fig 6B). To further substantiate that excess GluR signaling via NMR-2 and GLR-1 could support our swip-10 observations, we selectively overexpressed these receptors in DA neurons and examined CEP neuron morphology. As hypothesized, we detected statistically significant DA neuron degeneration, as compared to non-transgenic lines, similar to that observed in swip-10 mutants (Fig 6C).
The evidence presented above of a role of Ca2+-permeant iGluRs [77] in DA neuron degeneration, as well as our prior findings that swip-10 DA neurons demonstrate an exaggerated Ca2+ elevation in response to food contact [44], suggested to us that DA neuron degeneration in these animals could reflect activation of Ca2+-dependent programs linked to apoptotic and/or necrotic cell death [78–80]. Consistent with this idea, we found that loss of the primary endoplasmic reticulum (ER) Ca2+ storage/binding protein, calreticulin (crt-1) protected against swip-10 DA neuron degeneration (Fig 7A). Excessive activation of the Ca2+-activated protease calpain-1, has been shown to lead to cellular damage, including neurodegeneration, in both mammals and C. elegans [81–83]. In keeping with these findings, a loss of function mutation of clp-1, the C. elegans calpain-1 ortholog, significantly attenuated the DA neuron degeneration of tm5915 animals (Fig 7A). Together, these results support the hypothesis that inappropriate or excessive elevations of intracellular Ca2+ support swip-10 DA neurodegeneration. In mammals, aberrant excitotoxic Ca2+ signaling can generate reactive oxygen species (ROS) leading to activation of cell stress pathways that drive neuronal cell death [84, 85]. To explore this idea, we inspected swip-10 animals for signs of oxidative stress by monitoring reporter expression from Pgst-4::GFP. The gst-4 gene encodes a glutathione-s-transferase, and is a target for the ROS responsive transcriptional regulator SKN-1(C. elegans Nrf2 ortholog) [86, 87]. As shown in Fig 7B and 7C, tm5915 animals demonstrate a significant elevation in Pgst-4::GFP expression. As a measure of ER stress, we monitored the transcriptional reporter, Phsp-4:GFP [88, 89]. Although tm5915 animals did not show indications of basal ER stress with this marker (Fig 7D and 7E), they were more sensitive to the pharmacological ER stressor, tunicamycin, compared to N2 animals (Fig 7D and 7E).
Glu-induced excitotoxicity has been reported to arise from multiple mechanisms, including necrosis, autophagy, and apoptosis [90], processes that also contribute to cell death in the nematode [61]. Cells dying by necrosis exhibit cell swelling and vacuolization [61], which we do not observe in swip-10 animals (S5 Fig). In contrast, as described in Fig 1, DA neurons in swip-10 mutants display blebbing or breaks in processes (Fig 1C and 1F) and shrunken soma (Fig 1D and 1G), features characteristic of apoptosis [91]. Consistent with this idea, we found that gain of function ced-9 mutant animals [92] and loss of function ced-4 and ced-3 mutants, well-known contributors to programmed cell death [59], significantly suppressed tm5915 DA neuron degeneration (Fig 8A). Apoptosis in the context of normal developmental programmed cell death is tightly coupled to cell corpse engulfment [93], with two partially-redundant and parallel pathways involving ced-1/ced-6 [94, 95] and ced-10 [96] responsible for recognition of dying cells and initiation of cell corpse clearance. Little is known concerning the integration of death and engulfment programs in relation to DA neuron cell death, though Offenburger recently reported contributions from both ced-6 and ced-10 linked engulfment mechanisms in 6-OHDA induced DA neuron degeneration [97]. In contrast, we found that genetic disruption of individual genes associated with ced-1/ced-6 and ced-10 had no effect on measures of swip-10 DA neuron degeneration (Fig 8B). These findings suggest that swip-10 DA neuron degeneration arises from the activation of a cell-autonomous apoptotic pathway, one that draws little observable support from known engulfment mechanisms.
Overall, our findings reveal that loss of glial-expressed swip-10 results in DA neuron degeneration through a process supported by excess Glu signaling through Ca2+-permeant ionotropic Glu receptors and Ca2+-dependent cell death mechanisms that engage apoptotic cell death pathways, as summarized in Fig 9. Although, we predominantly characterized swip-10 DA neuron viability in gravid (egg-laying) adults, time-dependence studies indicate that degeneration is evident by day 2 post-hatching, and increases on all degeneration measures across the lifespan. A predominant display of fragmented or truncated dendrites in young animals versus shrunken or lost soma at later stages (Fig 3) suggests that degeneration in individual neurons is progressive, first emerging as altered neurite structure, followed by engagement of all compartments and eventually resulting in disappearance of some DA neurons altogether. This progressive pattern of degeneration is commonly seen with neurons suffering from energy depletion, that can be triggered by excessive stimulation or through metabolic poisoning [98, 99].
Most of our observations were obtained with a DA neuron-specific, cytosolic, fluorescent reporter, findings corroborated using a membrane anchored reporter (Figs 1 and S1). Subsequent studies of swip-10 mutants using electron microscopy to image DA dendrites and cilia provided clear evidence of physical alterations (Fig 2) that we believe reflect the declining health of the DA neurons, versus a direct action of swip-10 or its immediate effectors though further studies are needed. Additional EM studies would also be valuable in investigating the degeneration of swip-10 mutants at the level of the DA neuron cell bodies, axons, and synapses. There are significant technical challenges associated with identifying these cells and processes in the densely populated nerve ring, though future studies that make use of correlated light electron microscopy (CLEM) can be envisaged [100].
The discovery that DA neurons degenerate in swip-10 animals was initially surprising as our identification of swip-10 derives from a hyperdopaminergic behavioral phenotype [44], though we previously demonstrated reduced DA levels in these animals [43]. Since swip-10 DA neurons exhibit increased excitability and tonically-elevated DA secretion rates [44], we hypothesize that the degenerative process we have characterized likely contributes to a perturbation of mechanisms that insure a tight control over DA release (e.g. DA autoreceptors), along with a diminished capacity for DA clearance, leading to Swip. Alternatively, a degeneration-induced loss of DA signaling capacity could result in a hypersensitivity of postsynaptic DOP-3 DA receptors, such that DA release arising from water immersion then triggers excessive inhibition of motor neurons and Swip. In support of the latter possibility, movement of swip-10 animals on plates reveals a heightened sensitivity to exogenous DA [43].
Having generated evidence for an age-dependent degenerative process impacting the morphology of swip-10 DA neurons, we pursued mechanistic studies through a combination of genetic and imaging techniques. Such approaches have provided for a systematic elucidation of mechanisms underlying both programmed and environmentally-triggered cell death [59, 101–104]. In addition to the apoptotic pathways that drive programmed cell death during development, molecular determinants of later stage necrotic neuronal death, that arise as a result of the constitutive activity of mutant ion channels [105, 106] and ligand-gated Glu receptors [107], have been investigated. A potential role for excess Glu signaling in swip-10 DA neuron degeneration seemed plausible given the contribution of Glu receptors and Glu transporters to Swip reported in our prior study [44]. In this regard, the groups of Driscoll and Mano have provided evidence that necrotic cell death arises with excess Glu signaling that occurs from a combined loss of Glu clearance and a hyperactive, constitutively active form of the alpha subunit of the G-protein, Gs [108–110]. Although swip-10 DA neuron death shares features associated with the degeneration described in Mano’s studies, specifically the contributions from the iGluR, glr-1 (Fig 6A), and the intracellular Ca2+ sequestering protein, crt-1 (Fig 7A), our analysis also reveals a number of differences. Thus, besides a lack of morphological evidence of swollen, vacuolated soma seen in prior studies (S5 Fig), we found no evidence for a contribution to DA neuron degeneration of the adenyl cyclase ortholog, acy-1, nor could we implicate the autophagy-associated, cell death protein kinase, dapk-1 (S6 Fig) [108, 109]. We also obtained evidence that the damaging effects of swip-10 mutation are quite distinct from those observed with 6-OHDA induced DA neuron cell death. For example, the degeneration of DA neurons that arises within a day following 6-OHDA administration to wildtype worms lacks contributions from genes that participate in programmed cell death mechanisms [49], whereas, as we discuss below, contributions of these genes are evident in the swip-10 model. Moreover, recent studies indicate that ced-6 and ced-10 dependent engulfment pathways support 6-OHDA induced loss of DA neurons [97], whereas we found no contribution of these engulfment genes to swip-10 effects (Fig 8B). Moreover, Offenburger and colleagues have reported that 6-OHDA induced DA neuron death is exacerbated by mutation of the Ca2+ chaperone crt-1 whereas we demonstrated crt-1 mutation confers neuroprotection [111]. Together, these findings indicate that the DA neuron degeneration induced by swip-10 mutation is an altogether unique form of neural degeneration as compared to prior glutamatergic and exogenous neurotoxin models.
A critical step in defining the mechanism associated with swip-10 DA neuron degeneration is to determine the site(s) where wildtype swip-10 expression is required to support normal DA neuron morphology. As with the rescue of DA-dependent Swip behavior [44], we found that glial swip-10 re-expression, both genomic and cDNA, rescued swip-10 DA neuron degeneration, whereas DA neuron expression of swip-10 was without effect. These findings attest to a cell non-autonomous mode of action and raise the possibility that glial loss of swip-10 may damage the glial cells themselves, rendering them unable to engage in secretory or contact-mediated support for DA neurons. Although non-quantitative, we detected no obvious morphological differences between wildtype and swip-10 glia (S2 Fig), which may indicate that swip-10 expressing glia are deficient in a capacity to provide specific trophic or metabolic support to DA neurons, versus participating in critical cell autonomous mechanisms. Future studies using higher resolution, EM-based methods should be pursued to refine this analysis. Importantly, we obtained rescue of swip-10 DA neuron degeneration with a promoter driving wildtype swip-10 expression in CEP sheath glia. Moreover, degeneration was apparent in OLL neurons that like CEPs are ensheathed by glia but not in BAG neurons, which lack these contacts. These studies reinforce a contribution of glia to the cell non-autonomous actions of swip-10 to sustain neuronal viability and suggest that neurons in close apposition to ensheathing glia may preferentially depend on the activity of swip-10.
Our prior studies [44] assessing Swip behavior in swip-10 mutants provided evidence of perturbed glial control of extracellular Glu that we hypothesized was responsible for the iGluR and mGluR dependence of Swip in these animals. We therefore considered the possibility that perturbed buffering of extracellular Glu by swip-10 glia also underlies DA neuron degeneration. Mammalian literature emphasizes the critical role of glial Glu buffering mechanisms as protective against Glu excitotoxicity. As first described by Olney and colleagues, Glu excitotoxicity derives from excessive synaptic Glu acting on post-synaptic iGluRs [112–114], a process recapitulated by the actions of iGluR agonists such as kainic acid and ibotenate [115–117]. Moreover, inhibition of Glu transporters and increased extracellular Glu recapitulates the pathological hallmarks of PD in animal models, including DA neuron degeneration [118]. Our findings that mutations in the Ca2+ permeant iGluRs, glr-1 and nmr-2, protect against swip-10 DA neuron degeneration and that overexpression of these receptors leads to DA neuron degeneration in wildtype animals (Fig 6C) provide strong supportive evidence that glial mechanisms dictating the availability of extracellular Glu are likely disrupted in swip-10 animals.
Mammalian glia have been reported to modulate extracellular Glu by vesicular release [119], Glu-permeable channels [120], synaptic clearance of Glu by Na+-coupled Glu transporters (GLTs) [31], and extrasynaptic Glu buffering by the cystine/Glu exchanger (xCT) [34, 35]. We found that a mutation in the vesicular Glu transporter vglu-3 attenuates swip-10 DA neuron degeneration (Fig 5A). We were surprised that an eat-4 mutation did not contribute to swip-10 induced degeneration, as such a mutation reduced Swip behavior [44]. Although the expression pattern and role for vglu-3 is undetermined, these findings raise the possibility that EAT-4 supports Glu signaling in the neural circuitry that drives DA neuron excitation in response to water, whereas VGLU-3 contributes to Glu release directly onto DA neurons, including SWIP-10 expressing glia, and drives tonic activation of Glu receptors on DA neurons and over time, excitotoxicity. Consistent with this model, distinct Glu receptors support Swip (GLR-4, GLR-6 and MGL-1) versus DA neurodegeneration (GLR-1 and NMR-2). Although we did not observe DA neurodegeneration with genetic loss of single GLT orthologs in the nematode (Fig 5B), unlike Swip [44], this may be due to genetic redundancy among the six GLTs. Indeed, studies by the Driscoll lab demonstrated that loss of one or two GLTs is insufficient to drive Glu-dependent neurodegeneration [121]. In contrast to our inability to implicate specific GLTs, we found that genetic disruption of the xCT related gene, aat-1, significantly reduced swip-10 DA neuron degeneration (Fig 5C). As with vglu-3, the expression pattern for aat-1 in the worm is undefined, and thus additional studies are needed to determine site(s) of expression that contribute to our results. The effects of aat-1 mutation were not additive with those of vglu-3, suggesting that both genes act to support DA neurodegeneration through a common mechanism, which we propose is through the control of tonic, extracellular Glu providing tonic excitation of DA neuron expressed Glu receptors. Finally, it is important to note that mammalian xCT is upregulated by the β-lactam antibiotic ceftriaxone [36, 42, 122], which we have shown binds directly to the putative swip-10 ortholog MBLAC1 [45]. Moreover, research, initiated by findings of Rothstein and colleagues [13], has demonstrated that ceftriaxone is neuroprotective, including in models of DA neuron degeneration [41].
Although not exclusive, Glu-induced neural degeneration often involves activation of Ca2+-permeable NMDA type iGluRs [19, 123, 124] and, as noted, our studies demonstrate an important contribution of Ca2+-permeable C. elegans iGluRs, the NMDA-type iGluR, nmr-2, as well as the AMPA-type iGluR, glr-1 in swip-10 neural degeneration[77] (Fig 6A). Expression profiling data provides evidence that nmr-2 and glr-1 are expressed in DA neurons [125]. Since swip-10 mutant animals with loss of both nmr-2 and glr-1 do not demonstrate enhanced suppression of DA neural degeneration as compared to single receptor mutations (Fig 6B), we suggest that the flux of Ca2+ through one of these receptors is sufficient to increase intracellular Ca2+ sufficiently to initiate downstream signaling pathways that lead, over time, to neurodegeneration. Aberrant intracellular Ca2+ regulation and signaling has been implicated in excitotoxic cell death [126], with evidence supporting a role for Na+/Ca2+-permeable degenerin/epithelial sodium channels (DEG/ENaCs) [127–129], Ca2+-dependent proteases such as calpain [130, 131], and deficiencies in ER Ca2+ buffering [80, 106] in cell death mechanisms. We found that disrupting ER Ca2+ storage, by mutation of crt-1, or mutation of the C. elegans calpain ortholog, clp-1, significantly rescued swip-10 DA neural degeneration (Fig 7A). Ca2+ dysregulation following excessive Glu stimulation has also been shown to engender multiple indications of cell stress including oxidative stress and ER stress [132, 133], which swip-10 mutants display.
Finally, although acute Glu excitotoxicity has been more typically associated with necrosis [17, 19], evidence suggests that chronic dysregulation of Glu signaling and altered intercellular Ca2+ homeostasis can lead to activation of apoptotic pathways [134, 135], and a recent study by Anilkumar and colleagues has demonstrated that external factors, such as nutrient availability, determine whether or not excess Glu signaling triggers apoptotic or necrotic cell death (Anilkumar 2017). Consistent with this idea, genetic disruption of apoptosis in C. elegans [59] significantly reduced the DA neurodegeneration of swip-10 mutants (Fig 8A). The progressive DA neuron degeneration we detect in swip-10 animals supports the occurrence of a chronic insult and thus is in line with our genetic findings of apoptotic program engagement. However, our data suggests that swip-10 involvement of apoptotic cell death associated genes differs from the involvement of these genes in developmental programmed cell death, as loss of genes critical for cell-corpse engulfment during programmed cell death did not alter the levels of swip-10 DA neuron degeneration (Fig 8B). Although lack of a reliance on engulfment genes could be a reflection of the partial redundancy of the two major engulfment pathways, we suspect that these findings are indicative of a slower engagement of apoptotic genes in the swip-10 model. Additionally, the majority of our assays are conducted at a mid-point, with degeneration in progress, to capture various degrees of degeneration in swip-10 animals, it is possible that we have simply not assessed the correct temporal window for engulfment.
Although we present clear evidence for a significant role of excess Glu signaling in the degeneration of swip-10 DA neurons, other mechanisms besides changes in extracellular Glu homeostasis are likely to contribute to our observations since Glu homeostasis and signaling mutants afford incomplete suppression of swip-10 DA neurodegeneration. The elucidation of the normal role and genetic pathway for wildtype swip-10 in C. elegans glial cells will likely clarify other contributors to swip-10 induced neural degeneration. For example, mammalian glia have been shown to support neurons by buffering ions such as potassium (K+) and hydrogen (H+) [136], and by providing metabolic support via lactate, glutathione, and ATP shuttling [26]. Although only limited data speaks to glial-neuronal crosstalk in worms, we suspect that one or more of these mechanisms contribute to the diminished viability of DA neurons in swip-10 animals. As our transcriptional stress reporter data indicate a systemic increase in cellular stress mechanisms (Fig 7B–7E), it seems entirely likely that the perturbations induced by swip-10 mutation extend beyond the deficits observed in CEP (and OLL) neuron viability. Since wholesale degeneration is not evident, we suspect that the premature degeneration of DA neurons reflects a more dependent relationship of these cells on glia. The selective loss of nigrostriatal DA neurons in idiopathic PD has been suggested to derive from an intrinsic vulnerability to stress, possibly arising from the reactivity of DA itself, as well as inefficient anti-oxidant protection, ultimately rendering these cells more vulnerable than others to Glu-induced cell death [137]. Since genetic elimination of the capacity to synthesize DA did not reduce swip-10 DA neuron degeneration, we feel it more likely that excess Glu signaling drives degeneration in combination with a parallel loss of glial metabolic or trophic support required by DA neurons.
In summary, our findings reveal a previously unreported dependence of DA neurons on C. elegans glia, one that when disrupted leads to neuronal degeneration. DA degeneration triggered by glial loss of swip-10 appears to be progressive and dependent on excess Glu signaling through Ca2+ permeant iGluRs. We propose that these effects lead to perturbed intracellular Ca2+ homeostasis and, progressively, the engagement of apoptotic cell death pathways. Our work adds support to studies in mammals that indicate a critical role of proper glial function in DA neuron viability [138–141] and reveals a new worm model of Glu excitotoxicity, one likely amenable to pharmacological manipulation that could provide insights to novel therapeutics to treat human neurodegenerative disorders.
Strains were maintained as described previously [142]. We thank J. Rand (Oklahoma Medical Research Foundation); the Caenhorhabditis Genetics Center; Shohei Mitani of the National Bioresource Project at Tokyo Women’s Medical University; and Shai Shaham, Niels Ringstad, and Oliver Hobert for providing the strains used in this work. N2 (Bristol) served as our wild-type strain, and unless specified otherwise, we utilized the proposed null allele, TM5915, of swip-10 [44]. Strains used in this study are enumerated per figure appearance in S1 Table.
In all cases, insertion of the DNA fragment of interest and the fidelity of the vector was confirmed by sequencing and all PCRs were performed using KAPA HiFi HotStart ReadyMix (Kapa Biosystems). All constructs resulted in C-terminal cDNA fusion to an unc-54 3’ UTR. For the membrane bound transcriptional reporter, we used overlap PCR [143] and Gibson Assembly (NewEngland Biolabs) to subclone the 700bp dat-1 promotor into the myrRFP containing backbone from pptr-10:myrRFP (gift from Shai Shaham) to create pRB1349 (pdat-1:myrRFP). For transgenic swip-10cDNA::GFP rescue experiments, DA neuron, pan-glial, and CEPsh glial expression was achieved using the previously described plasmids, pRB1157, pRB1158, and pRB1159, respectively [44]. Genomic full-length swip-10 rescue experiments were conducted as previously described [44]. For the DA neuron-specific Glu receptor experiments, a PCR product (20ng/μL) was amplified by overlap PCR [143] to include the 700bp dat-1 promoter and genomic glr-1 from the ATG start to 2890 of genomic nmr-2 from the ATG start to 2974 fused to unc-54 3’ UTR for injection, along with punc-122:RFP (35ng/μL) and pdat-1:myrRFP (35ng/μL).
Crosses were performed using publicly available, integrated fluorescent reporter strains to mark chromosomes in trans. Single worm PCR was performed to confirm the presence of the indicated mutation. For all deletions, we used a three primer multiplex strategy that produces PCR amplicons with a 100–200 bp difference between N2 and mutant. This method was highly effective in eliminating preferential amplification of a lower-molecular-weight species. In all cases, a synthetic heterozygous control was used to ensure that heterozygous clones could be identified. We identified recombinant lines by PCR genotyping of single worm genomic DNA lysates. All genotyping PCRs were performed with the KAPA Genotyping Kit (KAPA Biosystems). In some cases, alleles were sequenced with sequence-specific primers to verify mutation homozygosity (GeneHunter and EtonBioscience).
Confocal microscopy of mutants on the BY250 strain background was performed using a Nikon A1R confocal microscope in the FAU Brain Institute Cell Imaging Core using a 20x or 60x oil-immersion objective and Nikon Elements capture software. Worms were immobilized using 30mM levamisole in M9 on a fresh 2% agarose pad and cover-slipped with a 1mm cover glass before sealing with paraffin wax [144].
The neurodegeneration assay was adapted from a previously described method [145]. In our case, we transferred 20 worms to normal NGM/OP50 plates as L4s and incubated these plates for 48hrs at 19°C until animals reached the gravid adult stage, unless otherwise noted. We then picked 15 worms into 20μL of 30mM levamisole in M9 on slides prepared with a 2% agarose pad. For imaging, we utilized a Zeiss Discovery V12 inverted fluorescent microscope outfitted with a Xenon UV light source and GFP/YFP/RFP filter sets. We used a Zeiss mono FWD 16mm objective lens to visualize Green Fluorescent Protein (GFP) containing integrated transgenes, vtIs7[Pdat-1::GFP], nsIs242[Pgcy-33::GFP], wgIs328[Pser2prom3::GFP] selectively expressed in DA, BAG, and OLL neurons respectively, allowing us to examine neurodegeneration in a cell-specific manner. For the DA neurons, analysis was limited to CEP neurons, because out of the 8 DA neurons in C. elegans, the 4 CEP neurons display the clearest and most distinct dendritic projections and can be readily identified via both light and electron microscopy (see below). Neurons were examined for the presence of 1) breaks in the CEP dendrites 2) shrunken or 3) missing somas. Worms were counted as displaying degeneration if one or more of these features were present. Normal N2 CEP, BAG, and OLL neurons lacked any of these abnormalities at the gravid adult stage. Total animals with degeneration, shrunken and missing somas, or neurite breaks were calculated for each trial. The percentage of animals exhibiting each morphological trait was determined for graphical analysis. Animals were tested 15 animals/day on 7–9 separate days (n = 90–135 animals assayed per genotype) blinded to genotype.
N2 and swip-10 mutant animals were raised and maintained at 20°C on E. coli OP50/NGM plates and 2-day adult animals (fixed 2 days after the L4 stage) were fixed and embedded for transmission electron microscopy (TEM) following a chemical immersion protocol [146, 147]. Briefly, animals were first cut open in a cacodylate-buffered osmium tetroxide fixative, then en bloc stained in uranyl acetate, and dehydrated and embedded in Spurr resin. Thin sections were collected onto Formvar-coated slot grids and examined on a Philips CM10 electron microscope. Digital images were collected with an Olympus Morada camera on the TEM, and figures were created using Photoshop.
All fluorescent stress reporter stains were a generous gift from Dr. Matt Gill (Scripps Research Institute, Jupiter, FL). All stress reporter strains were imaged as gravid adult animals grown at 19°C for 48hrs after transfer to a fresh OP50/NGM plate at the L4 stage. To determine levels of stress we used the transcriptional reporter strains, dvIs19 [pgst-4:GFP] and zcIs4 [phsp-4:GFP] to measure oxidative stress and ER stress respectively. We adapted previously described methods [87, 88]. Briefly, the overall pgst-4:GFP fluorescence intensity/μm per 15–20 3 day adult swip-10 animals and 15–20 3 day adult N2 animals (with subtracted background fluorescence per animal) was determined, and the fold change GFP intensity compared to N2 signal was calculated for all animals assayed from one population and subsequently averaged over 4 independent days (n = 60–75 animals assayed). As a positive control for oxidative stress, we picked 15–30 L4 N2 animals to OP50 plates 2mM paraquat (Sigma) mixed with the NGM agar [148]. phsp-4:GFP fluorescence intensity/μm was assayed as described above. To determine susceptibility of swip-10 mutants to ER stress, we transferred 15–30 L4 N2 and swip-10 animals to NGM plates containing 10μg/mL tunicamycin (Sigma) [89]. For each of the stress reporters, images were acquired using identical imaging settings across blinded genotypes and drug treatments, via a Nikon A1R confocal microscope in the FAU Brain Institute Cell Imaging Core using a 4x objective and Nikon Elements capture and analysis software.
All statistical tests were performed and graphs generated using Prism version 7.0. Data were analyzed by Student’s t-tests, one-way ANOVAs followed by Sudak or Dunnet’s post-hoc tests and two-way ANOVAs, where appropriate. A P < .05 was taken as evidence of statistical significance in all cases.
|
10.1371/journal.pgen.1002022 | Sensing of Replication Stress and Mec1 Activation Act through Two Independent Pathways Involving the 9-1-1 Complex and DNA Polymerase ε | Following DNA damage or replication stress, budding yeast cells activate the Rad53 checkpoint kinase, promoting genome stability in these challenging conditions. The DNA damage and replication checkpoint pathways are partially overlapping, sharing several factors, but are also differentiated at various levels. The upstream kinase Mec1 is required to activate both signaling cascades together with the 9-1-1 PCNA-like complex and the Dpb11 (hTopBP1) protein. After DNA damage, Dpb11 is also needed to recruit the adaptor protein Rad9 (h53BP1). Here we analyzed the mechanisms leading to Mec1 activation in vivo after DNA damage and replication stress. We found that a ddc1Δdpb11-1 double mutant strain displays a synthetic defect in Rad53 and H2A phosphorylation and is extremely sensitive to hydroxyurea (HU), indicating that Dpb11 and the 9-1-1 complex independently promote Mec1 activation. A similar phenotype is observed when both the 9-1-1 complex and the Dpb4 non-essential subunit of DNA polymerase ε (Polε) are contemporarily absent, indicating that checkpoint activation in response to replication stress is achieved through two independent pathways, requiring the 9-1-1 complex and Polε.
| The maintenance of genome stability is an essential process which needs a careful control. Indeed, the checkpoints are surveillance mechanisms sensing alterations in the integrity of the genome and preventing the replication and segregation of defective DNA molecules. The DNA integrity checkpoint is a signal transduction cascade conserved from yeast to man, and the apical factors in the pathway are protein kinases, called Mec1/Tel1 in Saccharomyces cerevisiae and ATR/ATM in mammals. DNA integrity can be challenged by lesions caused by a variety of chemical/physical agents, or by replication stress caused by special DNA structures, or by a limited supply of deoxyribonucleotides (dNTPs). The mechanisms leading to checkpoint activation in response to DNA damage are better understood compared to the processes leading to activation as a consequence of replication stress. We investigated the mechanisms required for Mec1 activation in response to dNTPs depletion caused by hydroxyurea treatment. We found that Mec1 activation occurs through two independent pathways: one acting through the PCNA-like 9-1-1 complex and the second through Dpb11 and DNA polymerase ε. The existence of these two pathways suggest a model possibly reflecting a DNA strand specificity in the detection of replication stress.
| The DNA replication machinery can experience various types of stress during S phase. This can happen when the replisome encounters DNA lesions that hinder its progression, while traversing slow replication zones corresponding to genomic regions difficult to replicate [1] or when encountering replication fork barriers [2]. Replication stress can also be induced by inhibiting ribonucleotide reductase (RNR) with hydroxyurea, which causes a global replication arrest by reducing the dNTPs pools [3].
Under replication stress conditions, eukaryotic cells trigger a signaling cascade, known as the replication checkpoint, which, in budding yeast, culminates with the phosphorylation of Rad53 [4]. This protein kinase is essential for the activation of the molecular mechanisms required to cope with replication arrest: it promotes stabilization of stalled replication forks and allows DNA replication re-start after removal of the blocking agent [5], [6], [7], [8]. Rad53 is also responsible for inducing the transcription of RNR genes by inhibiting the transcriptional repressor Crt1 and promoting the degradation of the RNR inhibitor Sml1 [9], [10]. Finally, Rad53 prevents the firing of late replication origins [11] and restrains spindle elongation thus preventing mitosis [12], [13], [14].
The DNA damage and replication checkpoints are genetically distinct pathways; however, they are partially overlapping since they share several of the factors involved. In fact, replication stress activates Mec1, the same apical kinase triggered by DNA damage, which is recruited to RPA-covered ssDNA by its binding partner Ddc2 [15]. After damage, Mec1 phosphorylates the Rad9 adaptor protein, which has been loaded onto DNA via chromatin-dependent and -independent pathways: the former requiring methylation of H3-K79 and the latter depending on the 9-1-1 complex and Dpb11 [16], [17], [18], [19], [20]. Phosphorylated Rad9, in turn, recruits Rad53, which becomes hyperphosphorylated in a Mec1-dependent manner. Differently, in the case of HU-induced checkpoint activation, the Rad9 adaptor protein is dispensable and its function is performed by Mrc1, a constitutive member of the replisome complex [21], [22].
It is now clear that following genotoxin treatments, primary lesions are generally recognized by specific repair factors that process them to generate ssDNA regions, which elicit the DNA damage response. On the other hand, the actual mechanism acting in the activation of the replication stress response is poorly understood. In budding yeast, it has been suggested that replication proteins may be involved in sensing blocks of the replication fork. Indeed, in addition to Dpb11, the initiation factor Sld2/Drc1 and Polε itself are required for efficient checkpoint activation in response to HU treatment, although the corresponding mutants are only mildly sensitive to the drug [23], [24], [25].
Sld2 is an essential CDK1 target required for initiation of DNA replication. Its phosphorylation and subsequent interaction with Dpb11 is essential for the loading of Polε and the firing of replication origins [26], [27]. Polε consists of four subunits: Pol2 and Dpb2 are essential for cell viability while Dpb3 and Dpb4 appear to be non-essential. These last two factors contains a histone-like fold motif and are also implicated in transcriptional regulation [28], [29]. The Polε holoenzyme is composed of two structurally distinct domains: a globular domain, made of the N-terminus of the catalytic Pol2 subunit and a tail-like domain containing the other three factors, bound to the Pol2 C-terminus [30], [31]. The catalytic subunit contains an N-terminal polymerase domain followed by a C-terminal region, where the checkpoint-defective mutations of POL2 map [24]. Surprisingly, deletion of the polymerase domain does not cause cell lethality, whereas the checkpoint domain is essential for cell viability [32].
It has been established that in response to DNA damage, the 9-1-1 clamp is loaded onto the 5′ primer-template junction adjacent to RPA-coated ssDNA [33], [34]. In higher eukaryotes, 9-1-1 then recruits TopBP1 which, through an interaction with ATRIP, stimulates the ATR kinase activity [35], [36], [37], [38]. Recent work in yeast demonstrated that Mec1 activation can proceed also through a 9-1-1-dependent, but Dpb11-independent pathway, mediated by an activation domain present in the Ddc1 subunit of the 9-1-1 complex [39]. Indeed, it has been reported that S. cerevisiae 9-1-1 can directly activate the Mec1-Ddc2 kinase in vitro [40]. The in vivo balancing between these two pathways has been recently studied, following Rad53 phosphorylation [39], which is influenced not only by Mec1 activation, but also by the Rad9 mediator [39].To determine directly the relative contributions of Ddc1 and Dpb11 to Mec1 activation in different cell cycle phases, and particularly in response to replication stress, we analyzed a direct target of Mec1 kinase, histone H2A, whose phosphorylation is not dependent upon Rad9.
In this study we found that, in G1 yeast cells, Mec1 activation induced by UV irradiation completely depends on the 9-1-1 dependent pathway, whereas Dpb11 only plays a minor role. Conversely, in response to replication stress, Mec1 activation is achieved through two independent pathways which rely on the 9-1-1 complex and Dpb11, respectively. At least one of these two pathways is necessary to efficiently activate Mec1 and to allow cell growth in the presence of HU. Finally, we provide evidence that the DNA polymerase ε complex and Sld2 are required to establish the 9-1-1 independent branch of Mec1 activation and we suggest that this could reflect strand-specificity in detecting replication stress.
We have previously shown that, in M phase, Dpb11 is required to recruit the Rad9 adaptor protein to UV-damaged DNA in a pathway that is parallel to that controlled by histone modifications [16], [20]. Dpb11 was also found to stimulate Mec1 kinase activity in vitro and this function appears to be modulated by its interaction with the 9-1-1 complex [41], [42]. To dissect the Mec1-activation role of Dpb11 in vivo and to determine the relative contribution of Dpb11 and 9-1-1 to this mechanism in different cell cycle phases, we analyzed histone H2A phosphorylation as an assay for Mec1 activity. After UV damage H2A is phosphorylated directly on serine 129 (γH2A) by Mec1 kinase; indeed mec1-1 mutant cells fail to phosphorylate H2A after DNA damage and a strain deleted in TEL1, coding for a second sensor-kinase, does not show any significant reduction in γH2A levels (Figure S1A and S1B).
We used a yeast strain carrying a C-terminal deletion of Dpb11 (Δ583―764) encoded by the dpb11-1 allele, which removes almost entirely the ATR Activation Domain (AAD) and a strain carrying the deletion of DDC1, the gene encoding the 9-1-1 subunit involved in Mec1 activation [40]. WT, dpb11-1, ddc1Δ and ddc1Δdpb11-1 cells were arrested in G1 with α-factor and in M phase with nocodazole and UV irradiated. As it is shown in Figure 1A, histone H2A is extensively phosphorylated after UV treatment in G1 and this damage-dependent modification requires the presence of a functional 9-1-1 complex, while the contribution of the AAD domain of Dpb11 is only minor. The quantification of the signal (shown in the lower panel of Figure 1A), indicates that the level of phosphorylated histone H2A (γH2A) in dpb11-1 is ∼50% of that found in WT cells.
In M phase cells the basal level of phosphorylated H2A-S129 is much higher (Figure S1C), and this likely influences the magnitude of the increase measured after UV-irradiation. In these conditions, Dpb11 plays a minimal role, if any, in H2A phosphorylation and also DDC1 deletion reduces γH2A only partially (∼50%) (Figure 1B). However, the residual H2A phosphorylation observed in a ddc1Δ mutant strain is lost when TEL1 is deleted, (Figure 1C). On the other hand, deletion of TEL1in the dpb11-1 background does not significantly influence H2A phosphorylation (Figure S1D)
To further elucidate the balancing between 9-1-1-dependent and Dpb11-dependent Mec1 activation in S phase, we decided to analyze this process after replication stress induced by HU. This allowed us also to minimize the side effects due to the involvement of Dpb11 in Rad9 recruitment because, during HU treatment, Rad9 does not become hyperphosphorylated and is not expected to play any role in checkpoint activation [22]. WT, dpb11-1, ddc1Δ and ddc1Δdpb11-1 cells were synchronized in G1, released into fresh medium supplemented with 200 mM HU, and checkpoint activity was monitored by measuring Rad53 phosphorylation (Figure 2A). Differently from what found in G1 and G2 cells, strains lacking either a functional 9-1-1 complex or the Dpb11 C-terminal region were fully able to phosphorylate Rad53. In these experimental conditions, ddc1Δ dpb11-1 double mutant cells showed a very severe defect in Rad53 phosphorylation, similar to that found in a Mec1-defective strain. These results suggest that a dpb11-1 ddc1Δ double mutation virtually abolishes UV-induced Mec1 activation differently from what previously reported [39], In addition, the double mutant strain showed synthetic lethality on HU plates (Figure 2B and [43]). To confirm that the dpb11-1 and ddc1Δ mutations directly affect Mec1 activity, we monitored γH2A levels in the same conditions. As shown in Figure 2C, the ddc1Δ and dpb11-1 mutations showed a synthetic defect in the ability to phosphorylate H2A-S129 (Figure 2D).
Although displaying a severe defect in Rad53 phosphorylation, ddc1Δdpb11-1 still displays a residual low level of phosphorylated Rad53, which may be dependent upon a residual Mec1 activity. However, Figure 2E and Figure S2A show that the residual Rad53 phosphorylation in the double mutant is instead due to Tel1. Indeed, an additional mutation eliminating Tel1 function completely abolishes Rad53 phosphorylation in a dpb11-1 ddc1Δ strain and strongly sensitizes cells to HU treatment, as shown in Figure S2B. These findings further support the hypothesis that Mec1 cannot become activated in response to replication stress in the absence of both Ddc1 and Dpb11-AAD.
To verify the possibility that in dpb11-1 mutant cells an unscheduled, Ddc1-dependent, DNA damage response is triggered as a consequence of the inability to properly activate the replication stress response, similarly to what happens in an mrc1Δ strain [22], we monitored DNA damage checkpoint activation looking at Rad9 hyperphosphorylation. As shown in Figure 2F, differently from what found in the mrc1Δ control strain, no Rad9 hyperphosphorylation was detectable in ddc1Δ, dpb11-1 single or double mutant strains. Consistently, rad9Δdpb11-1 double mutant cells are far less sensitive than the ddc1Δdpb11-1 strain to HU treatment (Figure 2B and [43]).
Rad53 kinase activity is required to stabilize stalled replication forks [7]. To verify whether the increased HU sensitivity of ddc1Δdpb11-1 double mutant cells was due to their inability to fully activate Rad53 and thus to stabilize the replisomes, we performed a recovery assay. Briefly, WT, dpb11-1, ddc1Δ, ddc1Δdpb11-1 and mec1-1sml1 mutant strains were blocked in G1, released and exposed to HU for 90 minutes; cells were then washed and shifted into fresh medium lacking HU and allowed to recover. As shown in the control strain mec1-1 sml1, when Rad53 activity is impaired, cells transiently exposed to HU loose the ability to resume DNA synthesis and complete DNA replication once the drug has been removed ([6] and Figure 3A). Unexpectedly, we found that not only dpb11-1 and ddc1Δ single mutant cells, but also the double mutant strain, which has a severe Rad53 hyperphosphorylation defect, were able to recover from the HU treatment with a WT kinetics (Figure 3A). Moreover, with lower HU concentrations, ddc1Δ dpb11-1 cells were capable of completing a round of DNA replication, as demonstrated by the re-entering of the replicated chromosomes in a pulsed-field gel system (Figure 3B).
Another marker of checkpoint activation by HU is the arrest of cell cycle, preventing mitosis. When exposed to HU, checkpoint mutants fail to delay the onset of mitosis and display elongated spindles [14]. To address the hypothesis that ddc1Δ dpb11-1 cells may die as a consequence of a premature mitosis, we measured spindle length 90 minutes after HU addition. ddc1Δ dpb11-1 double mutant cells prevent spindle elongation in the presence of HU, a process which is clearly defective in a mec1-1 mutant strain (Figure S3A), suggesting that the replication checkpoint can delay mitotic entry in the double mutant [10].
In agreement with all these data, the HU sensitivity of ddc1Δ dpb11-1 double mutant cells can be observed only to chronic exposure to the drug, while it is virtually undetectable if cells are transiently exposed to HU (Figure 3C).
ddc1Δ dpb11-1 mutant cells exhibit extremely low levels of Mec1 and Rad53 activation and, despite being sensitive to exposure to even low concentrations of HU (Figure 2B), they do not show some of the most common phenotypes observed in replication checkpoint defective cells. To better characterize the sensitivity to the drug, we monitored cell growth in the presence of 100 mM HU. The single and double mutant ddc1Δ dpb11-1 yeast strains were synchronized in G1, released into fresh medium supplemented with HU and cell cycle progression followed by FACS analysis. The double mutant ddc1Δ dpb11-1 showed a small delay in progressing through S-phase in the presence of HU, compared to WT and single mutant cells. Significantly, at late times (20 hours) after the release, a large fraction of double mutant cells appeared to be arrested at different stages of S-phase, while WT and single mutant cells had regained a FACS profile with 1C and 2C peaks (Figure 4A). Consistently, PFGE analysis of genomic DNA prepared from the various strains 20 hours after release from HU showed that in ddc1Δ dpb11-1 double mutant cells most of the DNA fails to enter the gel, suggesting the presence of branched intermediates (Figure 4B, 4C). It is important to note that, differently from what found in a mec1-1 strain, the ddc1Δ dpb11-1 strain did not accumulate cells with a<1C DNA content, or low molecular weight DNA fragments (Figure 4A–4C) indicating a correct segregation of chromosomes. Altogether, these findings may suggest that ddc1Δ dpb11-1 cells are unable to counteract the effect of HU by upregulating ribonucleotide reductase (RNR). Indeed, Rad53 regulates both the timely degradation of the RNR inhibitor Sml1 and the inactivation of Crt1, which represses the transcription of RNR genes [9], [10]. Consistently with this interpretation, ddc1Δ dpb11-1 cells show a modest delay in Sml1 degradation and, more significantly, CRT1 deletion suppresses, although not completely, the sensitivity of the double mutant strain to HU (Figure 4D, 4E).
Sld2/Drc1 and Polε participate in replication checkpoint signaling [24], [25]. Moreover, these factors were recently found to be part of the same pre-loading complex, together with Dpb11 and GINS [44]. An interesting possibility is that Sld2 and Polε exert their checkpoint function by controlling Dpb11-mediated Mec1 activation. To address this hypothesis we combined the drc1-1 allele with the DDC1 deletion. As it is shown in Figure 5A, similarly to what reported above for Dpb11, Sld2 also acts in a pathway that is parallel to that involving Ddc1; indeed, residual Rad53 phosphorylation present in ddc1Δ cells depends on Sld2. Moreover, drc1-1 cells do not show hyperphosphorylation of Rad9 in response to HU treatment, excluding the possibility of a secondary DNA damage response (Figure S4A). In agreement with these data, deletion of DDC1 displays a synergistic sensitivity to HU when combined with the drc1-1 mutation and the HU sensitivity of the double mutant strain is very similar to that observed for ddc1Δdpb11-1 cells (Figure 5B).
The checkpoint function of Polε appears to reside in the C-terminal domain of Pol2, which is bound, either directly or indirectly, by the three smaller subunits Dpb2, Dpb3 and Dpb4 and by Dpb11 [31], [45]. To assess if Polε participates in the Dpb11 signaling branch via its minor subunits, we deleted DPB4 in combination with the DDC1 deletion. Figure 5C shows that Rad53 phosphorylation is severely impaired in the double mutant ddc1Δdpb4Δ, closely resembling the phenotype of a ddc1Δdpb11-1 mutant. The same effect is measured by testing H2A phosphorylation in HU-treated samples (Figure 5D). The signals obtained for each time-point are quantified with respect to the signal detected in G1-arrested cells, in order to compensate for the higher basal level of γH2A observed in ddc1Δdpb4Δ double mutant cells in the absence of any treatment. Moreover, no unscheduled DNA damage checkpoint activation occurs, since no Rad9 phosphorylation is detected in dpb4Δ or dpb4Δ ddc1Δ cells treated with HU (Figure S4B). Finally, the ddc1Δdpb4Δ strain shows an HU sensitivity similar to that found in ddc1Δdpb11-1 cells (Figure 5E).
Apical checkpoint kinases (Mec1/Tel1 in budding yeast, ATR/ATM in humans) convert a structural signal coming from damaged DNA to a phosphorylation-based signaling cascade, and a large amount of work has been devoted to clarify the underlying mechanisms. Initially, the attention was focused on the recruitment of these kinases to damaged DNA [15], based on the assumption that binding to damaged chromatin sites would lead to their activation. More recently, the finding that Dpb11/TopBP1 stimulates Mec1 activity suggests a more complex scenario [40], [41], [42].
In vitro data obtained in Xenopus egg and mammalian cell extracts demonstrate the ability of TopBP1 to increase Mec1 kinase activity [35], [38]. The significance of this TopBP1 function does not appear to be specific for multicellular eukaryotes, since an interaction between Rad4/Cut5 and the checkpoint sensor kinase Rad3-Rad26 has also been found in S. pombe [46], [47]. More recently, in S. cerevisiae cells, Dpb11 has been demonstrated to contain an ATR activation domain (AAD), which is sufficient to promote Mec1 activation in vitro [41], [42]. These findings apparently contradict a previous observation that Mec1 can normally phosphorylate Ddc2 in a dpb11-1 mutant, lacking part of the AAD, after UV damage in M phase [16], while in our hands DDC1 deletion prevents Ddc2 phosphorylation (unpublished observation). Two explanations can be envisaged: in dpb11-1 mutant cells, Mec1 activity may be sufficient to phosphorylate Ddc2, while being defective towards other substrates; alternatively, Dpb11 may play only a marginal role in response to UV irradiation in M phase. We favored the second hypothesis because dpb11-1 mutant cells are mildly sensitive to UV irradiation and are proficient in the G2/M checkpoint; moreover, the 9-1-1 complex has also been identified as an activator of Mec1 in vitro [39], [40] and may play a prominent role in M phase. If this assumption is correct, Dpb11 could play a role in Mec1 activation in response to a different kind of damage or in other cell cycle phases. Interestingly, it was demonstrated that the dpb11-1 temperature-sensitive mutant is defective in checkpoint activation after replication stress caused by HU treatment at the restrictive temperature (36°C), while it is only mildly sensitive to the drug at permissive temperature ([23], [25] and Figure 2B).
To better understand the process of Mec1 activation in vivo after DNA damage or replication stress, we analyzed the relative functions of the two putative Mec1 activators: Dpb11 and the 9-1-1 complex. We extended our previous analysis by monitoring, in different cell cycle phases, a direct target of Mec1 kinase (histone H2A) as marker of Mec1 activity. We found that, both in G1 and in M phase, the 9-1-1 complex is absolutely required for Mec1 activation in response to UV treatment, while the contribution of Dpb11 AAD is only partial (∼50%) and restricted to G1. These in vivo findings are in agreement with the current activation model inferred from in vitro biochemical data [39], indicating that 9-1-1 can stimulate Mec1 through both Dpb11-dependent and -independent pathways in G1 (Figure 6, left). Differently, in M phase, the ATR activation domain of Dpb11 is dispensable for full Mec1 activation, which relies mainly on the presence of 9-1-1 (Figure 6, right). In fact, the residual UV-induced H2A phosphorylation detectable in the ddc1Δ strain, is dependent upon the Tel1 kinase (Figure 1). Different requirements for Mec1 activation in G1 and in M phase may reflect differences in CDK-controlled processing of DNA filament ends to generate the substrate detected by checkpoint factors [48], [49].
To complete studying of the pathways leading to Mec1 activation in different cell cycle stages, we analyzed the contribution of Dpb11 and Ddc1 to Mec1 activation in S phase cells challenged with replication stress. HU decreases the cellular concentration of dNTPs available for DNA synthesis and yeast cells respond by activating the replication checkpoint.
In vivo analysis of the phosphorylation state of two Mec1 substrates, H2A and Rad53, indicates that Dpb11 and 9-1-1 participate in Mec1 activation in response to HU treatment independently of each other in two parallel pathways. The possibility that dpb11-1 may cause problems to the replication process triggering a DNA damage response mediated by the 9-1-1 complex, similarly to what happens in mrc1Δ cells [22], seems unlikely. In fact, the Rad9 DNA damage-specific adaptor does not become hyperphosphorylated in both dpb11-1 and ddc1Δ single mutants. In agreement with such observation, rad9Δdpb11-1 cells are much less sensitive to HU than ddc1Δ dpb11-1 cells (Figure 2 and [43]).
We report that the HU sensitivity of ddc1Δ dpb11-1 strain is not due to replication fork collapse or premature elongation of the mitotic spindle (Figure 3 and Figure S2), two phenotypes characteristic of mutants defective in the replication checkpoint [7], [12]. Accordingly, the HU sensitivity of ddc1Δdpb11-1 double mutant cells, differently from that of a mec1-1sml1 strain, is not detectable in the case of transient HU treatment. This observation suggests that another Rad53 function activated by the replication checkpoint, and different from that responding to temporary fork arrest, is essential for sustaining growth in the constant presence of hydroxyurea. Indeed, ddc1Δ dpb11-1 double mutant cells grown in the presence of HU show defects in completing replication and accumulate replication intermediates. Moreover, ddc1Δ dpb11-1 cells are unable to counteract the effect of HU by upregulating ribonucleotide reductase. Interestingly, CRT1 deletion partially suppresses HU sensitivity of the double mutant strain (Figure 4E).
To obtain more insights on the pathways leading to Ddc1-dependent and Dpb11-dependent activation of replication checkpoint and to identify possible mechanisms specific for lagging or leading strand fork arrest, we analyzed mutants in the genes coding for proteins that are known to be involved in leading strand replication. During initiation of DNA replication, Dpb11 interacts with both Sld2 and Sld3 in a phosphorylation-dependent manner, a process that is required for origin firing [26], [27]. Moreover, temperature sensitive drc1-1 strains, mutated in Sld2, display the same checkpoint-deficient phenotype of dpb11-1 cells, when treated with HU at the non-permissive temperature, (Figure 5 and [25]). We tested whether Sld2 functions with Dpb11 in the same 9-1-1-independent pathway for Mec1 activation. Combining the drc1-1 allele with the DDC1 deletion, we found that ddc1Δ drc1-1 double mutant cells display the same Rad53 phosphorylation defect and the same HU sensitivity of a ddc1Δdpb11-1 strain, indicating that Mec1 activation by Dpb11 also requires Sld2 (Figure 5).
Mutants in the Pol2 C-terminus, the enzyme replicating the leading strand [50], are defective in the establishment of the replication checkpoint [24], [50] and this protein region of Pol2 was suggested to be involved in its interaction with other three Polε subunits: the essential Dpb2 protein and the non-essential Dpb3 and Dpb4 subunits [31], [45], [51]. Disruption of the DPB4 gene in a ddc1Δ background leads to identical phenotypes to the one observed in ddc1Δ dpb1-1 and ddc1Δ drc1-1, strongly suggesting that the 9-1-1-independent pathway involves leading strand replication factors. The observations that Dpb11 acts directly on Mec1 activity [41], [42] and that, in the dpb11-1 mutant, Polε seems to be normally loaded onto replication origins [52], strongly suggest that Dpb4, and possibly Sld2, function upstream of Dpb11 during checkpoint signaling. Unfortunately, it is impossible to perform a complete formal epistatic analysis as the dpb11-1 mutation also affects replication initiation and deletion of DPB4 or mutations in SLD2 are synthetic lethal when combined with the dpb11-1 allele [28], [53].
In conclusion our data suggest that during exposure to hydroxyurea, two independent pathways sense replication stress and signal for Mec1 activation. The first pathway depends on 9-1-1, which is known to be loaded at the 5′ of primer-template junctions, when RPA covers ssDNA ahead of the primer [34]. During unchallenged DNA replication these structures are normally formed on the lagging strand as a consequence of discontinuous DNA synthesis, and rapidly removed by refilling polymerase activity. Inhibition of DNA polymerization by HU likely stabilizes the 5′ DNA end providing the structure required for 9-1-1 loading. On the other hand, the higher processivity of leading strand synthesis makes it likely that the nearest 5′ end will be far away from the site of polymerase stalling, where ssDNA is generated and the Mec1-Ddc2 complex should be recruited. The absence of such structure could prevent the 9-1-1-dependent Mec1 activation. In this case a pathway requiring the leading strand factors Dpb4, Dpb11 and Sld2 becomes relevant to induce Mec1 activation (Figure 6, center). The hypothesis that Polε, Sld2 and Dpb11 work together in sensing replication stress is supported by the recent finding that an unstable complex containing Dpb11, Sld2, Polε and GINS is formed at the beginning of S-phase [44]. Moreover, the demonstration that under unstressed conditions Polε acts on the leading strand while Polδ works on the lagging strand [50], [54] supports the hypothesis that Polε and its interacting subunits may function in sensing replication stress on the leading strand, while the 9-1-1 complex may be more important to detect lagging strand fork arrest. Additional work will be needed to confirm this model and to identify the mechanisms leading to Dpb11 recruitment at the sites of replication fork stalling, since Dpb11 appears to co-localize with Polε during initiation of DNA replication, but not during elongation [52].
All of the strains used in this work are derivatives of W303 (K699 [MATa ade2-1 trp1-1 can1-100 leu2-3,12 his3-11,15 ura3]) and are listed in Table 1. Deletion strains were generated by using the one-step PCR system [55] or by genetic crossing.
Cells were grown overnight at 25°C to a concentration of 5×106 cells/ml and arrested in G1 with 5 µg/ml α-factor for three hours. 60 ml of cultures were spun and resuspended in the same volume of YPD supplemented with HU (200 mM or 100 mM, depending on the experiment). 20 ml samples were taken every 30 minutes after the release. In the case of untreated samples cells were released in fresh YPD +10 µg/ml nocodazole and every 5 minutes samples were taken for SDS-PAGE and FACS analysis.
Cells were grown in YPD medium at 25°C to a concentration of 5×106 cells/ml and arrested with nocodazole or α-factor (20 µg/ml). 50 ml of cultures were spun, resuspended in 500 µl of sterile water, and plated on a Petri dish (14-cm diameter). Rapidly, a 15 ml untreated sample was taken. Plates were irradiated at 75 J/m2 and cells were resuspended in 50 ml of YPD + nocodazole or α factor. Three 15 ml samples were taken every 10 minutes after irradiation.
Trichloroacetic acid protein extracts [56] were separated by SDS-PAGE; for the analysis of Rad9 phosphorylation, NuPAGE Tris-Acetate 3–8% gels (Invitrogen) were used following the manufacturer's instructions. Western blotting was performed with anti-Rad53, anti-H2A-S129 (Abcam #15083), anti-Actin (Sigma #A2066), anti-Sml1 and anti-Rad9 antibodies, using standard techniques. Values of phospho-H2A levels were obtained by quantifying the signal in the corresponding lanes using Quantity One software (BioRad) and normalizing it, first on the loading controls and then on the level of phospho-H2A in the untreated/G1-arrested sample of each strain.
1 ml of a 5×106 cells/ml culture were fixed overnight at 4°C with fixation buffer (3,7% formaldehyde, 0,1 M K-phosphate pH 6,4, 0,5 mM MgCl2). Cells were then washed three times with wash buffer (0,1 M K-phosphate pH 6,4, 0,5 mM MgCl2), one time with spheroplasting solution (1,4 M sorbitol, 0,1 M K-phosphate pH 6,4, 0,5 mM MgCl2) and resuspended in 200 µl of the same solution. Spheroplasts were prepared using 5 µl of 10 mg/ml Zymolyase at 37°C. Spheroplasts were washed with the same solution and used to prepare multi-well immunofluorescence slides which were incubated overnight with α-tubulin antibody (YOL1/34, Seralab) diluted 1∶100 in PBS-5%BSA.
HU plates were prepared by serial dilutions of the 2 M stock solution. Plates containing 25 mM, 50 mM and 100 mM HU were prepared. Overnight grown cultures were diluted to 1×106 cell/ml, then 10-fold serial dilutions were prepared and 10 µl of the suspensions were spotted on HU plates, which were incubated at 25°C. Images were taken 2 to 7 days later.
Agarose plugs containing yeast chromosomes were prepared as described previously [57]. These were incubated overnight at 37°C in 0.5 ml/plug TE containing 1 mg/ml RNAseA. After extensive washes with Wash Buffer (10 mM Tris-HCl pH 7.5 50 mM EDTA), plugs were loaded on 1% agarose gel and sealed in the wells with a solution of 1% LMP agarose in TBE 0.5X. Gels were run at 4°C for 24 h at 165 V, with 60 seconds pulses for 12 h and 90 second pulses for 12 h, using an Amersham Gene Navigator system.
|
10.1371/journal.ppat.1000351 | An Amphipathic α-Helix Controls Multiple Roles of Brome Mosaic Virus Protein 1a in RNA Replication Complex Assembly and Function | Brome mosaic virus (BMV) protein 1a has multiple key roles in viral RNA replication. 1a localizes to perinuclear endoplasmic reticulum (ER) membranes as a peripheral membrane protein, induces ER membrane invaginations in which RNA replication complexes form, and recruits and stabilizes BMV 2a polymerase (2aPol) and RNA replication templates at these sites to establish active replication complexes. During replication, 1a provides RNA capping, NTPase and possibly RNA helicase functions. Here we identify in BMV 1a an amphipathic α-helix, helix A, and use NMR analysis to define its structure and propensity to insert in hydrophobic membrane-mimicking micelles. We show that helix A is essential for efficient 1a–ER membrane association and normal perinuclear ER localization, and that deletion or mutation of helix A abolishes RNA replication. Strikingly, mutations in helix A give rise to two dramatically opposite 1a function phenotypes, implying that helix A acts as a molecular switch regulating the intricate balance between separable 1a functions. One class of helix A deletions and amino acid substitutions markedly inhibits 1a–membrane association and abolishes ER membrane invagination, viral RNA template recruitment, and replication, but doubles the 1a-mediated increase in 2aPol accumulation. The second class of helix A mutations not only maintains efficient 1a–membrane association but also amplifies the number of 1a-induced membrane invaginations 5- to 8-fold and enhances viral RNA template recruitment, while failing to stimulate 2aPol accumulation. The results provide new insights into the pathways of RNA replication complex assembly and show that helix A is critical for assembly and function of the viral RNA replication complex, including its central role in targeting replication components and controlling modes of 1a action.
| Positive-strand RNA viruses (one-third of all virus genera) transfer their genetic material between host cells as RNA of mRNA polarity, which are translated into proteins immediately upon entry. One immediate function of these proteins is to establish RNA replication compartments on intracellular membranes to copy the incoming viral RNA. Although much is known about the viral protein and RNA components in such replication complexes, little is understood about how the multiple protein–membrane–RNA interactions required for replication complex assembly are regulated. To study this, we used a well-established model virus that encodes only two replication proteins: an RNA polymerase enzyme that copies the viral RNA and an assembly-coordinating protein that guides the rearrangement of intracellular membranes to form replication compartments and recruits the viral RNA template and polymerase to these sites. We identified a small helix in this guiding replication protein that is essential for efficient association with and rearrangement of the correct intracellular membrane type and for regulating a switch between at least two different functional states of the replication guide protein. Mutations in this small helix interfere with separable guide protein functions, revealing new insights into the sequential steps in positive-strand RNA virus RNA replication complex formation.
| Positive-strand RNA viruses comprise over one-third of all virus genera and cause numerous diseases of humans, animals and plants [1]. Important human pathogens include hepatitis C virus (HCV), SARS coronavirus, Norwalk virus, West Nile virus, and the majority of common cold viruses, among others. Other positive-strand RNA viruses of animals, such as foot-and-mouth disease virus, and numerous plant viruses are of great veterinary and economic concern.
A universal feature of positive-strand RNA virus RNA replication is its close association with intracellular membranes. One or more viral nonstructural proteins target the viral replication complex to its preferred membrane type and often, if not always, induce membrane rearrangements. The responsible viral proteins can be true integral membrane proteins such as the flock house virus protein A that builds replication complexes on outer mitochondrial membranes [2] or HCV NS4B that targets HCV RNA replication to the endoplasmic reticulum (ER) membrane [3]. Alternatively, some viruses utilize peripheral membrane proteins such as the Semliki Forest virus nsP1 that locates to endosomal membranes [4] or HCV NS5A [5] and picornavirus 2 C [6], which associate with ER membranes.
Brome mosaic virus (BMV), a member of the alphavirus-like superfamily of human, animal, and plant viruses, is among the best-studied positive-strand RNA viruses for RNA replication. BMV has three genomic RNAs, RNA1, RNA2 and RNA3, and a subgenomic mRNA, RNA4. RNA1 and RNA2 encode nonstructural replicase proteins 1a and 2a polymerase (2aPol), respectively, which are required for RNA replication. RNA3 and RNA4 encode the 3a movement protein and the coat protein, respectively, required for systemic spread in plants [7]. BMV RNA replication and encapsidation can be fully reconstituted in the yeast Saccharomyces cerevisiae by expressing the viral RNA replication and/or capsid proteins together with one or more genomic RNAs [8],[9],[10]. BMV replication in yeast duplicates the major features of replication in BMV's natural plant hosts, and the powerful techniques of yeast genetics and molecular biology have greatly facilitated the investigation of BMV replication and host-virus interactions [11],[12].
In plant cells and yeast, BMV RNA replication occurs on the perinuclear region of the ER [13]. The only viral component in the BMV RNA replication complex that localizes independently to the ER is replicase protein 1a [14], a multifunctional protein with an RNA capping domain in its N-terminal half and an NTPase/ RNA helicase-like domain in the C-terminal half [15],[16],[17]. The other viral RNA replication components, the RNA polymerase 2aPol and RNA templates, depend on 1a for their recruitment to the ER membrane and into the RNA replication process [14],[18],[19],[20],[21]. In close linkage with this recruitment, 1a dramatically increases the in vivo stability (but not the translation) of viral genomic RNA3 [22], and similarly increases the accumulation of the 2aPol protein [19].
When 1a associates with ER membranes, it induces the formation of membrane-bound spherular invaginations, that we will refer to as spherules [13]. By electron microscopy, the 50–70 nm diameter spherules are bounded by a single lipid bilayer continuous with the outer ER membrane and containing condensed or fibrillar material. The membrane bounding of this compartment is almost complete except for a narrow neck-like opening that retains a connection to the cytoplasm [13]. By electron microscopy, spherules in yeast cells that express only 1a are indistinguishable from spherules in yeast co-expressing 1a, low copy numbers of 2aPol, and genomic RNA3, and that are actively replicating viral RNA [13]. Similar spherules are induced in association with RNA replication by many other positive-strand RNA viruses [12],[23].
The manner by which 1a interacts with ER membranes to induce these membrane invaginations, and the details of 1a's interactions with the other viral components remain poorly understood. We previously showed that BMV 1a has no trans-membrane domain(s) and resides fully on the cytoplasmic side of the ER membrane, but that amino acids 368–478 contain sequences important for ER membrane binding [24]. In this report we use NMR and other approaches to identify an amphipathic α-helix in this region, which is critically involved in 1a-membrane association, spherule induction and functional RNA replication complex assembly. The results also provide significant new insights into the pathways by which the RNA replication complex assembles and how different 1a functions are coordinated, revealing e.g. that 1a-induced membrane invagination and 1a-induced viral RNA protection are closely linked, while 1a interaction with and stimulation of BMV 2aPol accumulation does not require, and is in fact inhibited by, membrane rearrangements.
Previously, using membrane affinity and protease sensitivity assays, we showed that BMV 1a strongly localizes to the cytoplasmic face of the ER membrane despite lacking any detectable trans-membrane domain [24]. Membrane flotation assays of 1a deletion derivatives and GFP-fusion to truncated versions of 1a showed that a 105 amino acid (aa) region (aa 368–472, previously designated region E, Fig. 1A) plays a major role in 1a-ER membrane binding [24]. In this region, a stretch of 35 amino acids (aa 388–422) is predicted to be predominantly α-helical. Within this helical region, a putative amphipathic α-helix core peptide of 18 amino acids (aa 392–409) can be recognized, which we will refer to as “helix A”. One indication that helix A is likely important is that its amino acid sequence is evolutionarily highly conserved among the equivalent 1a replication proteins of other bromoviruses (Fig. 1A).
To test the functionality of helix A for membrane association, the 105, 35 and 18 aa regions described above were fused to the N-terminus of GFP to produce E-GFP, 35H-GFP, and 18H-GFP, respectively (Fig. 1B). Lysates of yeast cells expressing these fusion proteins were loaded under flotation gradients, which upon centrifugation were fractionated and analyzed by SDS PAGE and western blotting using anti-GFP antibodies. As a measure of membrane association, flotation efficiency was determined as the percentage of total GFP or 1a-GFP fusion protein in the gradient that was present in the top two fractions. In these assays, less than 3% of wild type cytosolic GFP floated to the top of the gradient with the membrane fraction. Fusing the 35 aa region to GFP greatly increased membrane association up to 45%, which was as efficient as membrane association directed by the full 105 aa E region fused to GFP. The smaller 18H-GFP fusion protein retained about 30% flotation efficiency (Fig. 1B). Thus, the 35 aa segment 388–422 accounts for essentially all of 1a's membrane association mediated by domain E, and the 18 aa helix A region retains most of this function and is sufficient to direct membrane association of GFP.
A helical wheel projection of the 18 aa helix A core region shows that it has the potential to form an amphipathic α-helical cylinder with one side (the right side in Fig. 2) having a cluster of hydrophobic, non-polar residues including three leucines (L396, L400, L407) and two nearby positive-charged lysines (K403, K406), and the other (left) side of the helix mostly hydrophilic and polar residues (Fig. 2, see also marked aa in Fig. 1A). To test these predictions, we used NMR to resolve the structure of an 18 aa peptide with the core sequence (aa 392–409) of helix A. NMR spectra of this peptide dissolved in water did not reveal a long term stable structure. However, upon including 100 mM SDS to provide lipid bilayer-mimicking micelles [25], the peptide showed NMR spectral changes consistent with a stable conformation (Fig. 3A). Based solely on 13C chemical shifts, NMR showed that aa 397–406 in the peptide had a >80% probability to be in a helical structure (Fig. 3B). To elucidate this further, the three dimensional structure of the peptide was calculated based on NOE distance constraints arising from spatial contact of hydrogen atoms observed to be closer than ∼5×. Additional dihedral angle constraints were derived from chemical shifts using the TALOS program [26]. The resulting structure (Fig. 3C) shows an α-helical conformation for aa 397–406, indicating that an amphipathic helix formed upon binding to the lipid membrane-mimicking SDS micelle. The constraints and overall quality of the structure are shown in Table 1.
Table 2 shows that 65% of the observed NMR signals were assigned to specific atoms in the peptide. Of these assigned signals, 80% were affected by the addition of 16-doxyl stearic acid (DSA), a paramagnetic molecule whose presence in SDS micelles causes nearby atoms' NMR signals to broaden and lose intensity, thus serving as an internal probe for the extent to which atoms on the surface of a labeled structure are immersed in the micelles [27]. In parallel with the distribution of hydrophobic amino acid residues (Fig. 2 and Fig. 3C), the N-terminal half of the peptide had a larger percentage of assigned atoms that showed DSA contact than the C-terminal half, i.e. 91% vs. 69%, respectively (Fig. 3C and Table 2).
Since the structure and DSA results implied that L396, L400, and L407 were positioned in the face of helix A most deeply immersed into the bilayer-mimicking micelle (Fig. 3C, bottom view, and 3D), we tested the importance of these three leucines for helix A-mediated membrane association. We introduced L to A mutations in the 18H-GFP fusion protein expression plasmid and tested their effects on membrane flotation efficiency. As shown in Fig. 4, the wt18H-GFP again had 30–35% flotation efficiency, while single L to A mutations reduced this to ∼7–15%. Of the three leucines, mutating the more N-proximal L396 and L400 more severely reduced membrane association than mutating L407, which paralleled the stronger micelle contact of the N-terminal half of the peptide (Fig. 3C and Table 2). These results might also explain in part the tolerance for an isoleucine at the 407-equivalent position in other bromovirus replicase proteins (Fig. 1A). A fusion protein with a combination of all three L to A mutations had near background level flotation, implying a complete loss of function of helix A in targeting cytosolic GFP to membranes. In contrast, K to E mutations reversing the charge of lysines 403 and 406 (the only basic residues in the 18 aa helix A core) showed K403E to only marginally decrease the flotation efficiency of 18H-GFP, while K406E had no significant effect (Fig. 4), consistent with the NMR observation that these amino acids have weak and no lipid contact, respectively (Fig. 3C). A double K to R mutation designed to retain the positive charge at these amino acid positions did not affect membrane association at all (Fig. 4), suggesting that K403 might contribute to membrane association via its positive charge, perhaps by neutralizing negatively charged lipid head groups. Overall, as mutations that change the leucine-rich non-polar face of the helix have more detrimental effects on membrane association than other amino acid substitutions, the results were consistent with the NMR-based structure of helix A and show that amphipathic helix A has a key role in membrane targeting.
To extend the results from helix A-GFP fusion proteins, the contribution of helix A to membrane association of full-length 1a was assessed using biochemical and cell imaging approaches. By membrane flotation gradient analyses, the flotation efficiency of wt 1a was ∼ 96% (Fig. 5A), confirming 1a's previously established high affinity for membranes [24]. Deleting the 35 aa or 18 aa helices reduced 1a-membrane association by over two-fold (Fig. 5A). The three L to A mutations, either as single mutations or as a triple combination, similarly reduced the flotation efficiency of full length 1a to ∼45%. Single alanine insertions immediately downstream from L396 and L400 reduced flotation efficiency to levels similar to full helix A deletions (Fig. 5A), confirming the importance of correct spacing to maintain the amphipathic characteristics of helix A. The importance of the charged lysines at positions 403 and 406 at the hydrophilic face of helix A was assessed using alanine or arginine substitutions. Single position substitution mutants and double mutants K403/406A and K403/406R maintained full flotation efficiency (Fig. 5A, single mutations not shown). 1a mutants K403E, K406E, and double mutant K403/406E retained intermediate flotation efficiencies showing that although the positive charge at these positions is not required, reversing it to a negative charge destabilizes membrane association (Fig. 5A). The K403/406E single and double mutations showed a somewhat greater inhibition of membrane association in the context of full length 1a (∼63% for the double mutant in Fig. 5) than in the context of the 18 aa helix fused to GFP (∼77%, Fig. 4), suggesting the possibility that residues outside of the 18 aa helix core might cooperatively influence membrane association.
Since none of the deletions and mutations completely abolished 1a-membrane association, we used confocal immunofluorescence microscopy to compare the sub-cellular localization of the 1a mutants with that of wt 1a (Fig. 5B). Wildtype 1a localized predominantly to the perinuclear ER membrane, co-localizing almost completely with the distribution of ER marker Sec63p. In contrast, the 1a protein mutants that lacked either the 35aa or 18aa helices no longer co-localized with Sec63p and displayed a mostly diffuse cytoplasmic localization (Fig. 5B). Confocal fluorescence images showed similar staining throughout the cytoplasm for 1a triple mutant L396/400/407A and the K403/406E double mutants, although in these cases a minority of 1a retained ER association. By contrast, the K403/406R mutant co-localized with Sec63p throughout, as for wt 1a (Fig. 5B).
Combined, the flotation and confocal results demonstrate that 1a has both helix A-dependent and -independent modes of membrane association, but that helix A is crucial for efficient membrane association and normal 1a localization to perinuclear ER membranes. While other aa such as the positively charged lysines contribute, the leucines on the hydrophobic side of helix A are the most important residues for effective association of 1a with ER membranes.
We previously showed that, in the absence of 2aPol or other viral components, 1a targets itself to perinuclear ER membranes and induces spherular invaginations that by EM are indistinguishable from those that replicate BMV RNA when 1a is expressed together with low 2aPol levels expressed from the yeast ADH1 promoter [13]. Examples of such spherules are shown in Fig. 6, top left panel. In contrast, 1a plus high 2aPol levels expressed by the strong yeast GAL1 promoter shift the predominant viral-induced membrane rearrangements from spherules to large, karmellae-like, multilayer stacks of double membrane layers surrounding the nucleus (Fig. 6, top right panel). Although dramatically different in organization, such membrane layers support BMV RNA replication as efficiently as spherules [28].
Fig. 6 shows that deleting helix A (1aΔ35H, 1aΔ18H) abrogated 1a's ability to induce either type of ER membrane rearrangement. Likewise, mutating the hydrophobic face of helix A in triple mutant 1a L396/400/407A or reversing the positive charge of the two lysines in double mutant 1a K403/406E abolished 1a's ability to form either membrane rearrangement, whether expressed alone or together with GAL1-promoter -driven 2aPol (Fig. 6A). Arginine substitution of the single negatively charged amino acid E405 maintained a wt phenotype, although alanine substitution at this position resulted in a >30-fold reduction in the number of spherules, showing the importance of a charged, hydrophilic amino acid at this position (data not shown).
In surprising contrast, double mutant 1a K403/406R showed an entirely different phenotype. This mutant, which as described earlier maintained full flotation efficiency (Fig. 5A), was revealed by EM analysis to form dramatically more, and somewhat smaller, membrane-bound spherules than wt 1a (Fig. 6B). To specify which of the two amino acid changes contributed to this phenotype, single mutants 1a K403R and 1a K406R were generated and expressed in yeast cells. In keeping with the DSA/membrane interaction of K403 but not K406 (Fig. 3C), Fig. 6B shows that 1a K403R maintained this mutant phenotype while 1aK406R induced spherules with the frequency and size of wt 1a. Moreover, 1a K403R induced high frequency, smaller spherules even in the presence of high levels of GAL1-promoter-driven 2aPol expression (Fig. 6B), conditions under which wt 1a preferentially induces ER membrane layers rather than spherules (Fig. 6A). These results show both that 1a-ER membrane association through helix A is crucial for 1a-induced membrane rearrangements, and that additional characteristics of helix A have important roles in determining the type of membrane rearrangement and the extent of membrane curvature.
Hereafter, we will refer to helix A mutants that have lost all membrane-rearranging capacity, like triple mutant L396/400/407A, as Class I mutants, and to mutants with the hyper- abundant, smaller spherule phenotype, like K403R, as Class II mutants. To evaluate the possible role of other helix A amino acids in Class I or Class II phenotypes, we first made alanine substitutions at the other residues besides L396/400/407 in the major membrane interacting face of helix A, i.e., F394, T397, Y401, Y404 and T408 (Fig. 3C, bottom view; see also Fig. 2). Strikingly, EM analysis showed that 1a T397A, 1a Y401A and 1a Y404A all were Class II mutants, inducing a plethora of small spherules like 1a K403R (Fig. 7A). Flotation analyses showed that all four of these Class II mutants also maintained wt 1a levels of membrane association (Fig. 7B). The F394A substitution, positioned on the same side of helix A as the above Class II mutants but at the N-terminal end of helix A (Fig. 2 and Fig. 3C), had a partial Class II phenotype of producing spherules at normal frequency but slightly smaller diameter than wt 1a. An alanine substitution at K403 resulted in a similar phenotype. By contrast, spherules of wt frequency and size were produced by 1a T408A, at the C-terminal end of helix A, by 1a bearing alanine substitutions at residues on the upper face of helix A (Fig. 3C), i.e., V392, L398, N399, and Q402. and by 1a A395S (results not shown).
To more accurately and precisely describe the Class II mutant phenotypes, we measured the abundance and diameter of spherules in the subset of cells that were sectioned through their nuclei among a total of 200 cells for each mutant. As shown in Table 3, spherule abundance in yeast cells expressing Class II mutants was 5- to 7- fold higher than in cells expressing wt 1a. Moreover, the average spherule diameter in cells expressing wt 1a was ∼66 nm, but was only ∼40–55 nm in cells expressing Class II mutants.
As mentioned earlier, when wt 1a and high levels of 2aPol are co-expressed, only 15–25% of cell sections with BMV-induced, perinuclear membrane rearrangements show spherules, while 75–85% bear double-membrane layers that support efficient RNA replication [28]. Even under such conditions of high 2aPol expression, the four Class II mutants induced ∼6- to 8-fold more spherules than wt 1a and reduced the frequency of cells with double membrane layers by >3- to 10-fold (Table 3). Thus, helix A mutations not only alter 1a's intrinsic functions for ER membrane rearrangement, but also the ability of 2aPol to modulate the type of 1a-induced ER membrane rearrangements.
In addition to mediating its own membrane association, wt 1a also recruits 2aPol to the RNA replication complex, mediated at least in part by a direct interaction between 1a's C-terminus and the N-terminus of 2aPol [19],[29]. In conjunction with such recruitment in this and previous studies [19], co-expressing wt 1a increased 2aPol accumulation by approximately two-fold (Fig. 8A). Accordingly, we measured 2aPol accumulation in the presence of the various 1a mutants to determine to what extent this 1a function depended on sequences in helix A. All mutant 1a proteins accumulated to levels similar to wt 1a, but Class I and Class II mutants showed directly opposite effects on 2aPol accumulation (Fig. 8B and 8C). Class I 1a mutants that lack the ability to induce ER invaginations not only retained the ability to stimulate 2aPol accumulation, but did so to nearly double the level of wt 1a (Fig. 8B). In contrast, Class II 1a mutants that form more numerous, smaller spherules, lost the ability to stimulate 2aPol levels over those in cells expressing 2aPol alone (Fig. 8C). Thus, 1a-mediated stimulation of 2aPol accumulation was inversely correlated with the capacity of 1a to induce ER membrane invaginations.
The localization of wt 1a and selected representatives of the Class I and II 1a derivatives in cells co-expressing 2aPol is shown in Fig. 8D (see also Fig. 5B for localization of the 1a derivatives without 2aPol). For each 1a mutant class, similar results were obtained with all members, and representative results are shown in Fig. 8D. In these studies we used a replication-competent GFP-2aPol fusion protein to allow direct fluorescence microscopy detection rather than immunofluorescence, which is often compromised by low 2aPol detection sensitivity [19].
As seen previously [19], GFP-2aPol fluorescence in the absence of 1a was mostly faint and diffusely cytoplasmic with a few punctate dots. When co-expressed with wt 1a, GFP-2aPol co-localized with 1a in typical partial to almost complete ring-like perinuclear ER structures (Fig. 8D), consistent with prior observations [19]. Although Class II 1a mutants failed to significantly stimulate 2aPol accumulation, the GFP-2aPol that accumulated in cells expressing Class II mutants co-localized with the mutant 1a in perinuclear rings similar to wt 1a (Fig. 8D, right two columns). By contrast, in the presence of the reduced membrane affinity Class I 1a mutants, GFP-2aPol accumulated in large cytoplasmic clusters also containing a significant fraction of the mutant 1a, while the remaining 1a was distributed diffusely over the cytoplasm (Fig. 8D, third and fourth columns), as when these Class I mutants were expressed without 2aPol (Fig. 5B).
As another assessment of membrane association, flotation efficiency of 1a or its helix A mutants remained unaffected when co-expressed with 2aPol (compare Fig. 8E with Fig. 5A). In the presence of wt 1a, 2aPol accumulation was stimulated and essentially all 2aPol became membrane-associated (Fig. 8E). Likewise, 2aPol was recruited to membranes by the Class II 1a mutants with ∼98% efficiency, but without any increased accumulation (Fig. 8E). When co-expressed with any Class I 1a mutants, the efficiency of 2aPol flotation with membranes was only 50–60% (Fig. 8E), slightly higher than without 1a and similar to the reduced membrane-association of Class I mutant 1a proteins themselves, with or without 2aPol (Fig. 8E and 5A).
In yeast cells, the half-life of RNA3 increases from 5–10 min in the absence of 1a to more than 3 hours in the presence of 1a, which is reflected in a marked increase in RNA3 accumulation [22]. Accordingly, as shown in Fig. 9A, lanes 1 and 2, GAL1 promoter-driven wt 1a increased RNA3 accumulation ∼20-fold. Strikingly, the effects of the Class I and Class II mutations on 1a stimulation of RNA3 accumulation were opposite to each other and opposite to the effects of each mutant on 2aPol. Co-expressing class II 1a mutants stimulated RNA3 accumulation ∼40-fold, or double the stimulation by wt 1a (Fig. 9A), in parallel with the increased frequency of spherule formation by these mutants (Table 3). In contrast, Class I 1a mutants showed no ability to stimulate RNA3 accumulation, so that RNA3 levels in cells expressing Class I 1a mutants were similar to those in cells lacking 1a (Fig. 9A).
Wild type 1a recruits RNA3 into a membrane-associated, nuclease-resistant state [13]. To define the state of RNA3 in the presence of the Class I and Class II 1a mutants, we assayed RNA3's membrane flotation efficiency, sedimentation, and nuclease sensitivity when co-expressed with these mutants (Fig. 9B and 9C). Without wt 1a, RNA3 remained at the bottom of flotation gradients, indicative of a complete lack of membrane-association. In sedimentation assays, RNA3 from cells lacking 1a was mainly detected in the membrane-depleted supernatant and readily degraded with micrococcal nuclease (Fig. 9C). In the presence of wt 1a or its Class II mutants, at least 80% of RNA3 segregated with the membrane fraction in the top gradient fractions or the membrane-enriched pellet fraction in sedimentation assays, and became highly nuclease-resistant, while Class I mutants failed to induce RNA 3 membrane association or nuclease resistance (Fig 9B and 9C). Thus, the loss or enhancement by Class I or II 1a helix A mutants of wt 1a's ability to stimulate RNA3 accumulation in vivo was closely linked with RNA3's acquisition of a membrane-associated, nuclease-resistant state, and with the capacity of each 1a mutant's ability to induce membrane invaginations.
In cells expressing 1a and 2aPol, RNA3 transcripts are recruited into 1a- and 2aPol-containing replication complexes to serve as templates for synthesis of negative-strand RNA3, which in turn becomes the template for synthesis of progeny positive-strand RNA3 and subgenomic positive-strand RNA4 [8],[21]. Since the Class I and Class II mutants in 1a helix A show opposite effects on 1a-associated intracellular localization, ER membrane rearrangements and stimulation of 2aPol and RNA3 accumulation (Table 3) we compared how these mutants affect BMV RNA replication in yeast and in a natural plant host of BMV, barley.
Yeast cells expressing wt 1a, 2aPol and RNA3 supported efficient viral RNA replication (Fig. 10A). In contrast, in cells expressing Class I 1a mutants, 2aPol and RNA3, only weak RNA3 signals were detected, similar to the levels of RNA3 derived entirely from plasmid-based transcription in cells lacking 1a (Fig. 10A). In cells expressing Class II 1a mutants, 2aPol and RNA3, positive-strand RNA3 accumulated to levels intermediate between those in cells with and without wt 1a (Fig. 10A), consistent with the ability of class II 1a mutants to mediate RNA3 recruitment to a membrane-protected state (Fig. 9). However, positive-strand RNA4 and negative-strand strand RNA3, which are only synthesized as products of viral RNA replication, were undetectable in cells expressing any of the Class I 1a mutants, and reached only 5–10% of wt levels in cells expressing most Class II 1a mutants (Fig. 10A). The only exception was Class II mutant 1aT397A, which weakly stimulated 2aPol accumulation (Fig. 8C) and retained ∼25% of wt 1a replication levels (Fig. 10A). Thus, BMV RNA replication was severely inhibited by the helix A mutations in both classes.
To compare the replication competence of the 1a helix A mutants in yeast to that in BMV's natural plant host, 7-day old leaves of barley plants were inoculated with in vitro transcribed wt or mutant RNA1 transcripts and equal amounts of RNA2 and RNA3 transcripts. Seven to nine days post inoculation with wt BMV RNAs, even leaves that were not inoculated but rather depended on systemic viral spread for infection contained abundant levels of RNA1, 2, 3 and RNA4 (Fig. 10B, lane 1). However, none of the RNA 1 mutants, including 1aT397A, supported detectable systemic infection (Fig. 10B). Thus, 1a mutations in helix A that abolish or severely inhibit BMV RNA replication in yeast also render the virus severely replication-deficient in its natural host.
Positive-strand RNA virus RNA replication occurs exclusively on intracellular membranes. Thus, the interactions by which viral replication proteins target specific membranes, recruit other viral proteins and viral RNA templates, and reorganize their target membranes to accommodate active RNA replication compartments are crucial to understanding replication complex assembly and function [12],[23]. In the case of BMV, the multifunctional replication protein 1a directs replication complex targeting and ER membrane-association, in addition to providing all viral enzymatic functions for RNA replication other than polymerase activity. Previously, we mapped the major 1a ER membrane association-mediating sequences between aa 368 and 478 (region E), and membrane association enhancing sequences in upstream region D [24]. Additional contributions to 1a membrane association were mapped to the 158 N-terminal amino acids of 1a (region A&B) [24]. However, we found that region E was sufficient for ER targeting, whereas the auxiliary sequences in region A, B, and D were not. Here, we have used genetic, biochemical and NMR analyses to identify a small amphipathic α-helix within BMV 1a region E, helix A, that is not only critically involved in 1a-induced membrane association and rearrangement, but also in 1a-mediated recruitment of viral RNA templates and RNA polymerase, and subsequent assembly and function of active replication complexes.
NMR structure analyses showed that at a minimum, the core twelve amino acids of helix A are in an α-helical configuration (Fig. 3). Mutational analyses (Fig. 4 and Fig. 5) and SDS micelle-based NMR and DSA-16 contact data (Fig. 3C and 3D and Fig. 11) show that the primary membrane association function of helix A resides in a hydrophobic face comprised primarily of three leucines at aa positions 396, 400, and 407. These leucines show significant conservation among sequenced bromoviruses (Fig. 1A), and mutation of these leucines to alanines, in effect removing their side chains, greatly diminished helix A- and full-length 1a-mediated membrane association (Fig. 4 and Fig. 5). Insertions of alanines immediately adjacent to the leucines, which disrupts their correct spacing and the amphipathic characteristics of helix A, likewise reduces 1a membrane association efficiency to that of complete helix A deletion mutants (Fig. 5). When lysines at positions 403 and 406, which can potentially interact with the negatively charged polar head groups of the lipid bilayer, were mutated to glutamic acids to change positive charge to negative charge, membrane association was also affected, but to a lesser extent (Fig. 4 and Fig. 5).
Immediately adjacent to the triple leucine hydrophobic face of helix A, T397, Y401, and Y404 form a polar and uncharged side (Fig. 3C, bottom view and 11). Polar residues Y and T are common targets for phosphorylation, which would add negative charge. However, computer-assisted predictions [30] do not support the likelihood of phosphorylation at these residues, due to lack of flanking phosphorylation site consensus sequences. Indeed, neither mutations T397D, Y401E and Y404E (intended to mimic phosphorylation), nor mutations T397R, Y401R and Y404R (that added positive charge) affected spherule formation, although they did render 1a defective in RNA replication (data not shown). Strikingly, however, alanine substitutions in this same T-Y-Y face of helix A revealed an entirely new 1a mutant phenotype. Unlike the L396/400/407 Class I mutants, these Class II mutants not only retained efficient ER membrane association, but dramatically increased the frequency of 1a-induced membrane invaginations 5- to 8-fold (Table 3).
Additional characterization further extended the opposing nature of the Class I and Class II 1a mutant phenotypes to the regulation of BMV 1a-mediated recruitment of 2aPol and viral RNA templates into the membrane-associated replication complex. Wildtype 1a directs cytosolic 2aPol to ER membranes via interaction of its C-terminal sequences with the N-terminal sequences of 2aPol [19], simultaneously stimulating 2aPol accumulation by ∼2-fold (Fig. 8A). Remarkably, Class I 1a mutations, which significantly inhibit 1a membrane affinity and abolish the capability to rearrange membranes, nearly doubled the ability of 1a to stimulate 2aPol accumulation (Fig. 8B). These 1a mutants decreased 1a membrane affinity to about 50% and, although interacting more efficiently with 2aPol, they did not recruit 2aPol to the typical perinuclear ER location. Instead, co-expressing these Class I 1a mutants with 2aPol induced both to concentrate into large cytoplasmic clusters (Fig. 8D). In sharp contrast, stimulation of 2aPol accumulation by 1a was completely abolished when co-expressed with the Class II 1a mutants that induced dramatically more abundant spherules than wt 1a (Fig. 7).
Analysis of the Class I and II mutants also showed that RNA3 recruitment and protection by 1a, unlike 2aPol recruitment, strongly correlated with 1a-induced membrane invagination. Class I 1a mutants that did not induce ER membrane invaginations failed to mediate significant recruitment of template RNA3, while Class II 1a mutants that form hyper-abundant spherules enhance RNA3 accumulation to even higher levels than wt 1a (Fig. 9 and Table 3). Along with prior results [13], this implies that the membrane-associated, nuclease-resistant state associated with RNA3 recruitment (Fig. 9C) represents the spherule interior.
The Class I and II 1a mutant phenotypes reveal significant insights into the pathways by which BMV RNA replication complexes assemble (Fig. 12). Immunogold electron microscopy and stoichiometric calculations of the various viral components in wt BMV replication complexes indicate that each spherule replication complex contains ∼200–400 BMV 1a molecules [13]. Calculations of spherule surface area and the predicted size of the 1a protein, 1a's strong affinity for the cytoplasmic face of the ER membrane [24], 1a self-interaction [29], and other results all imply that 1a forms an inner shell inside the spherules, explaining the formation and maintenance of these high-energy membrane deformations [12],[13]. Similar conclusions, based on electron microscope tomography and multiple other approaches, recently emerged for the role of transmembrane viral replication protein A in spherule RNA replication complexes formed by flock house nodavirus on mitochondrial membranes [23]. The observation that the Class II cluster of helix A mutations alters the size of the induced membrane spherules (Table 3) suggests that this part of helix A affects 1a-membrane and/or 1a-1a self-interactions that determine the diameter of the inner protein shell. Such altered interactions, together with the ∼3-fold reduced volume of Class II spherules, explain how Class II 1a mutants produce significantly more spherules than wt 1a (Table 3) from a similar or only slightly increased number of 1a proteins (Fig. 3C).
Since 1a mutants in the triple-leucine Class I cluster fail to induce ER invaginations and have reduced membrane association, these mutant 1a proteins must remain exposed at the cytoplasmic surface of the ER membrane or dissociate entirely from the membrane (Fig. 12). Such Class I 1a mutants were over twice as effective as wt 1a at interacting with, stabilizing and recruiting 2aPol to membranes (Fig. 8 and Table 3). This implies that 1a-2aPol interactions occur most efficiently and perhaps exclusively prior to spherule formation (Fig. 12). Consistent with these findings, Class II mutants, which are hyper-active in spherule formation, were markedly defective in 2aPol stabilization (Fig. 8 and Table 3). These results suggest that 1a-2aPol interaction and 1a-induced membrane invagination are sequential and perhaps antagonistic functions. Spherule formation might interfere with 1a-2aPol interaction by sequestering 1a in the spherule interior, by inducing conformational changes in 1a, or both. In either case, such interference could help to explain how spherules regulate accumulation of 2aPol to catalytic amounts of only one 2aPol for every 20 1a molecules [13]. Since Class I mutants with reduced membrane association enhanced 1a-2aPol interaction and nascent 2aPol can be efficiently recruited by 1a from cytosolic, translating polysomes [31], 1a-2aPol interaction may preferentially occur prior to 1a-membrane association.
In contrast to their 2aPol recruitment phenotypes, the Class I and Class II 1a mutant phenotypes as noted above showed strong correlation between spherule formation and recruiting and protecting RNA3 templates (Fig. 9 and Table 3), which likely extends to the mechanistically very similar recruitment of RNA1 and RNA2 templates [20],[32],[33]. This implies that RNA template recruitment is either closely linked to or subsequent to spherule formation ([21] and Fig. 12). The resulting ordered progression of 2aPol recruitment, replication complex assembly and RNA template recruitment seems tailored to satisfy the virus's crucial survival need to effectively use the limited number of viral genomic RNAs - potentially one - present during early phases of infection. It also is consistent with the dual function of the viral RNA genome to serve as a template for replication only after it has been translated and sufficient amounts of viral replication proteins have accumulated.
In summary, we find that helix A has crucial roles in directing and/or regulating multiple essential 1a functions in RNA replication complex assembly and function, including binding to membranes, inducing membrane curvature, and interacting with itself, 2aPol and viral RNA templates. In addition, the fact that RNA replication was abolished or severely inhibited by all Class II mutations (Fig. 10), which preserved membrane interaction, invagination and RNA recruitment, suggests that helix A may affect one or more additional 1a functions required for RNA synthesis, such as the enzymatic functions of the 1a RNA capping or NTPase/helicase domains (Fig. 1). Similar to the central role of helix A in 1a, amphipathic α-helices are also essential for the peripheral membrane association and function of some other positive-strand RNA virus replication factors, such as the Flaviviridae NS5A membrane anchor [34] Semliki Forest virus nsP1 RNA capping protein [4], and picornavirus 2C protein [6]. As with the possible role of helix A in modulating 1a enzymatic activities, the RNA capping activity of nsP1 is dependent on membrane association by its short amphipathic helix [35]. Moreover, like 1a, picornavirus 2C not only associates with membranes through an amphipathic helix, but induces membrane rearrangements, has NTPase activity and conserved helicase motifs required for RNA replication, and is implicated in amphipathic helix-modulated interactions with other viral RNA replication proteins [6]. Such emerging commonalities suggest that the membrane interaction and function of such amphipathic helices may embody common principles extending across important virus groups.
Yeast strain YPH500 and culture conditions were as described previously [8]. BMV 1a and mutant derivatives and 2aPol were expressed under control of the GAL1 promoter, using pB1YT3 [16] or derivatives and pB2YT5 [36], respectively. BMV RNA3 was expressed from pB3MS82, a GAL1 promoter expression plasmid of an RNA3 derivative with a four-nucleotide insertion in the coat protein gene has, abolishing expression of the coat protein [36]. The Sec63-GFP fusion protein was expressed from plasmid pWSECG, a derivative of pJK59 (gift from P. Silver, Department of Biological Chemistry and Molecular Pharmacology, Harvard University). The yeast-enhanced version of GFP and GFP-2aPol were expressed from pGFP and pGFP-2aPol, respectively, both based on pB2YT5 [19].
Ten OD600 units of yeast cells grown to mid-logarithmic phase were spheroplasted [37] and resuspended in 350 µl buffer TNT (50 mM Tris-HCl [pH 7.4], 150 mM NaCl, 5 mM EDTA, 5 mM benzamidine, 1 mM PMSF, and 10 µg/ml each aprotinin, leupeptin, and pepstatin A). Spheroplasts were lysed via 25 passes through a 22 gauge, 4 cm long needle. Total lysates were centrifuged for 5 minutes at 4°C at 500×g to remove cell debris, and 250 µl of supernatants were mixed with 500 µl of 60% Optiprep (Axis-Shield, Oslo, Norway). Density gradient centrifugation was performed for 5 hours at 55,000 rpm in a Beckman TLS55 rotor using 600 µl of each sample overlaid by 1.4 ml of 30% Optiprep and 100 µl of lysis buffer [21] After centrifugation, 6 fractions were collected from top to bottom of the gradient. For protein detection, samples were boiled in SDS loading buffer prior to SDS-PAGE and western blotting. For RNA analysis, RNA was isolated and prepared by the hot phenol method [38], and northern blotting was performed as described previously [21].
Spheroplasts were lysed in 100 µl lysis buffer (50 mM Tris-Cl pH 8.0, 2.5 mM EDTA, 1 mM PMSF, 5 µg/ml pepstain, 10 µg/ml leupeptin, 10 µg/ml aprotinin, 10 mM benzamidine) and centrifuged 5 min at 4°C at 2000×g to yield pellet and supernatant fractions. For RNase treatment, 1 U of micrococcal nuclease was added to 100 µl of supernatant and pellet fractions, incubated at 30°C for 15 minutes and inactivated by addition of 2 µl of 0.5 M EGTA (pH 8.0). RNA was isolated and prepared by the hot phenol method [38], and northern blotting was performed as previously described [21].
BMV RNA1 or its mutants, RNA2, and RNA3 were in vitro transcribed and capped (Ambion, Austin, TX) from EcoRI-linearized plasmid pB1TP3 or its derivatives, pB2TP5 and pB3TP8, respectively [39]. Seven-day-old barley leaves were inoculated with the resulting in vitro transcripts [40] and viral RNA was isolated seven to nine days post inoculation using a Qiagen RNeasy Mini kit. Northern blotting was performed as previously described [21].
The helix A peptide, representing amino acids 392–409 of BMV 1a (GenBank accession number ABF83485), was synthesized on an Applied Biosystems 432A synthesizer using standard Fmoc chemistry with HBTU/HoBT coupling [41]. Except for phenylalanine, all Fmoc-15N, (U)-13C labeled amino acids were purchased from Cambridge Isotope Laboratories. Labeled Fmoc-phenylalanine was purchased from Isotech. Wang resin was loaded with Fmoc-15N, (U)-13C-alanine using N, N-Diisopropylcarbodiimide and 4-diththylaminopyridine. The synthesis was carried out at a scale of 12.5 micromoles with a 3-fold excess of each amino acid. Coupling times for the first three and final 5 couplings were fixed at one hour each. The remaining 6 couplings were programmed as extended couplings. The cleaved and deprotected peptide was purified by HPLC using a C18 Vydac column (250×10 mm). Mass confirmation was done using a Bruker Biflex III MALDI-TOF.
NMR spectra were collected from a solution of 400 mM peptide in 100 mM SDS and 5% 2H2O using a Varian VNMRS 600 MHz spectrometer equipped with a 5 mm cryogenic triple resonance probe. DSA (4,4-dimethyl-4-silapentane-1-ammonium trifluoroacetate) and deuterated SDS were purchased from Aldrich. The 3D data were collected using HIFI, a rapid methodology for collection of multidimensional NMR spectra [42]. Spectra collected for assignments were: 1H{15N}HSQC, HNCO, CBCA(CO)NH, HNCACB, HNCA, HN(CO)CA, HN(CA)CO, HN(CA)CB. All experiments were standard Varian Biopack pulse sequences modified for the HIFI method [42].These sequences are available from the National Magnetic Resonance Facility at Madison. Automatically generated peak lists from HIFI were used as input to the automated assignment package suite (PISTACHIO [43], LACS [44], and PECAN [45]. 3D 15N-edited 1H-1H NOESY and 3D 13C-edited 1H-1H NOESY spectra were used as input for the ATNOS/CANDID/CYANA suite of programs [46],[47],[48]. The Protein Structure Validation Software suite of programs was used to assess the quality of the computed structure [49]. Images were rendered using PyMOL molecular graphics software (DeLano Scientific LCC http://www.pymol.org). The NMR and structural data described have been deposited in BioMagResBank (http://www.bmrb.wisc.edu) under BMRB accession number 20027.
Confocal microscopy was as described [21],[27]. Briefly, yeast cells co-expressing either wt 1a or 1a mutants and Sec63-GFP or GFP-2aPol were fixed with 5% formaldehyde, spheroplasted with lyticase, and permeabilized with 0.1% Triton X-100. Spheroplasts were then stained by using rabbit anti-1a serum, followed by Texas red-conjugated donkey anti-rabbit antibodies. Fluorescent images were acquired with a Bio-Rad 1042 double-channel confocal microscope system.
Samples were prepared for electron microscopy as described [13]. In brief, yeast cells were fixed for 1 hr with 2% glutaraldehyde and 4% paraformaldehyde, washed, and post-fixed for 1 hr with 1% OsO4 and 1% uranyl acetate. Cells then were dehydrated via a series of step-wise increasing ethanol concentrations ranging from 50% to 100%, and infiltrated and embedded with Spurrs resin. Samples were sectioned and placed on nickel grids, washed, incubated in 15 min in 2% glutaraldehyde, poststained with 8% uranyl acetate and Reynold's lead citrate, and viewed with a Philips CM120 microscope.
|
10.1371/journal.ppat.1003186 | Epstein - Barr Virus Transforming Protein LMP-1 Alters B Cells Gene Expression by Promoting Accumulation of the Oncoprotein ΔNp73α | Many studies have proved that oncogenic viruses develop redundant mechanisms to alter the functions of the tumor suppressor p53. Here we show that Epstein-Barr virus (EBV), via the oncoprotein LMP-1, induces the expression of ΔNp73α, a strong antagonist of p53. This phenomenon is mediated by the LMP-1 dependent activation of c-Jun NH2-terminal kinase 1 (JNK-1) which in turn favours the recruitment of p73 to ΔNp73α promoter. A specific chemical inhibitor of JNK-1 or silencing JNK-1 expression strongly down-regulated ΔNp73α mRNA levels in LMP-1-containing cells. Accordingly, LMP-1 mutants deficient to activate JNK-1 did not induce ΔNp73α accumulation. The recruitment of p73 to the ΔNp73α promoter correlated with the displacement of the histone-lysine N-methyltransferase EZH2 which is part of the transcriptional repressive polycomb 2 complex. Inhibition of ΔNp73α expression in lymphoblastoid cells (LCLs) led to the stimulation of apoptosis and up-regulation of a large number of cellular genes as determined by whole transcriptome shotgun sequencing (RNA-seq). In particular, the expression of genes encoding products known to play anti-proliferative/pro-apoptotic functions, as well as genes known to be deregulated in different B cells malignancy, was altered by ΔNp73α down-regulation. Together, these findings reveal a novel EBV mechanism that appears to play an important role in the transformation of primary B cells.
| Approximately 20% of worldwide human cancers have been associated with viral infections. Many oncogenic viruses exert their transforming properties by inactivating the products of tumour suppressor genes. One of the best characterized events induced by ongocenic viruses is the inactivation of the transcriptional factors p53. The mucosal high-risk HPV types, EBV, HTLV-1 and KSHV, via their viral proteins, are able to target p53 by distinct mechanisms. We have recently described a novel p53 inactivation mechanism of some cutaneous beta HPV types which have been suggested to be associated with skin carcinogenesis. Beta HPV38 induces accumulation of the p53 antagonist, ΔNp73α which in turn silences the expression of the p53-regulated genes. Here we report that also EBV, via the oncoprotein LMP-1, induces the expression of ΔNp73α which is dependent on the recruitment of p73 on ΔNp73 promoter and the activation of JNK-1. The recruitment of p73 to the ΔNp73 promoter correlated with the displacement of the histone-lysine N-methyltransferase EZH2 which is part of a transcriptional repressive polycomb 2 complex. We also show that ΔNp73α plays an important role in transformation of primary human B cells and regulates the expression of a large number of cellular genes that encode proteins linked to cancer development, including lymphomagenesis.
| Epstein-Barr virus, also known as human herpesvirus 4 (HHV4), belongs to the gammaherpesvirus family and is largely spread as it can be detected in 90% of the worldwide population. EBV infects B cells and, in most cases, does not lead to any clinical manifestations. However, when EBV infection occurs during adolescence or young adulthood, it may cause infectious mononucleosis, a benign lymphoproliferative disease. A minority of EBV infections result in the development of several types of human B cell malignancies, including Burkitt's lymphoma (BL), Hodgkin and non-Hodgkin lymphomas [1]. In addition, EBV has been clearly associated with epithelial cancers, i.e. nasopharyngeal carcinoma (NPC) and a sub-set of gastric carcinoma [1]. The risk of developing EBV-induced malignancies is significantly increased in immuno-compromised individuals, such as AIDS patients and organ-transplant recipients. In vitro EBV efficiently infects human resting B cells and transforms them into proliferating lymphoblastoid cell lines (LCLs) [2]. Similar to other herpesviruses, the EBV life cycle includes a latent and non-productive phase, as well as a lytic phase leading to the production of the virus progeny. After primary infection, EBV persists lifelong in a latent state in a sub-population of resting memory B cells [3]. Recent studies led to a model of EBV persistence whereby different viral transcription programs were used within the context of the normal biology of B lymphocytes in order to carry out its life cycle [4], [5]. Eleven genes can be expressed in the latency phases, namely the EBV nuclear antigens (EBNA) 1, 2, 3A, 3B, 3C, LP, the latent membrane proteins (LMP) 1, 2A, 2B, the untranslated EBER-1 and EBER-2 RNAs, as well as multiple microRNAs [2]. Based on the expression pattern of the different latency genes, four latency phases have been identified so far. Type I latency is normally present in Burkitt's lymphoma and is associated with the expression of EBNA-1 as well as EBERs and miRNAs. Type II latency is frequently detected in Hodgkin's lymphoma and nasopharyngeal carcinoma, and is linked to the expression of EBNA-1, LMP-1, LMP-2A, LMP-2B, EBERs and miRNAs. Type III latency is characterized by the expression of all 11 latency genes and is mainly found in lymphoproliferative diseases in immunocompromised individuals and in in vitro EBV-transformed LCLs. Finally, type IV latency is associated with the infectious mononucleosis and is less well defined, since the expression pattern of the latency genes may differ in different patients [6].
LMP-1 is the major EBV oncoprotein and displays transforming activities in in vitro and in vivo models [2]. It is an integral membrane protein composed of a short cytoplasmic amino-terminal domain, six hydrophobic transmembrane domains, and a cytoplasmic carboxy-terminal domain [2]. LMP-1 exerts its transforming properties by functioning as a member of the tumour necrosis factor receptor (TNFR) superfamily leading to the a constitutive activation of several cellular signaling pathways [7]–[9]. In particular, LMP-1 activates the nuclear factor-kappa B (NF-κB) signaling pathway, thus promoting cell growth and inhibition of apoptosis. In addition to LMP-1, other EBV latent proteins, i.e. EBNA-2, EBNA-3A and EBNA-3C, are involved in the immortalization of primary B cells. The role of EBNA-2 is mainly mediated by its ability to modulate the transcription of host and viral genes, while EBNA-3C plays a direct role in cellular transformation by inactivating the products of tumor suppressor genes, such as retinoblastoma (pRb) and p53 [10]–[14]. Interestingly, as for other oncogenic viruses, e.g. human papillomavirus type 16 (HPV16) [15], EBV developed multiple and redundant mechanisms to inactivate p53-regulated pathways. Indeed, EBNA-3C is able to alter the p53 transcription activity via direct binding as well as by inducing stabilization of p53 inhibitors, such as mdm2 and Gemin3 [10], [12], [14].
p73 is a closely p53-related transcription factor that shows functional similarity to p53 [16], [17]. The impact of EBV on p73 has been poorly investigated so far, although initial findings indicate that, similarly to p53, p73 is targeted by EBV. Indeed, p73 expression was found down-regulated in EBV-positive gastric carcinoma by heavy methylation of CpG islands within its promoter [18]. In addition, it has been recently shown that EBNA-3C attenuates p73 expression in LCLs by targeting the transcription factor E2F-1 [19]. However, it is likely that, in line with the previous findings on p53, EBV has developed multiple mechanisms to alter p73 function. In addition, due to the variability in the expression pattern of the latent genes in EBV-positive cancer cells, it is possible that more than one viral protein has the ability to target the p53/p73 pathway.
We have previously shown that cutaneous HPV38 which belongs to an HPV subgroup potentially associated with the development of non-melanoma skin cancer (NMSC), is able to induce the accumulation of ΔNp73α which in turn alters the transcriptional functions of p53 and p73 [20], [21]. ΔNp73α is a p73 isoform that lacks transactivation (TA) domain and is accumulated in several tumors [22]. Most importantly, increase in ΔNp73α protein levels correlates with poor outcome of the disease and bad response to therapy [23]–[25].
Here we show that EBV LMP-1 activates ΔNp73α expression in B cells by favoring the recruitment of p73 to ΔNp73 promoter. This phenomenon appeared to be mediated by c-Jun NH2-terminal kinase 1 (JNK-1), and resulted in the inhibition of p53-regulated genes encoding key anti-proliferative regulators.
Several isoforms of ΔNp73 have been identified, that can be generated by alternative splicing at the 5′ region of the p73 mRNA or by transcriptional initiation from a promoter (p2) within the p73 gene [26], [27]. We have previously shown that HPV38 induces the accumulation of ΔNp73α mRNA generated by the p2 promoter [20]. We therefore evaluated whether the levels of this specific ΔNp73 transcript were induced by EBV. ΔNp73α mRNA levels were determined in EBV-positive and EBV-negative B-cell lines by RT-PCR using specific primers. Six LCLs expressed high levels of ΔNp73α transcript, while no signal was detected in the EBV-negative B lymphoma cell line BJAB (Figure 1A). To assess that the over-expression of ΔNp73α was indeed linked to EBV infection, we infected primary B cells with recombinant EBV and analyzed ΔNp73α mRNA levels by quantitative RT-PCR. As previously shown, ΔNp73α is not expressed in primary B cells (Figure 1B) [28]. In contrast, an increase in ΔNp73 mRNA levels was observed between 12–36 hours post-EBV infection which correlated with LMP1 transcript levels (Figure 1B). ΔNp73 mRNA accumulation was also observed in cancer B-cell lines, RPMI, upon EBV infection cells (Figure 1C). Several C-terminus ΔNp73 isoforms have been characterized, that are generated by alternative splicing (α, β, γ, ε). The isoform alpha plays a key role in altering the p53/p73 functions and is over-expressed in several human cancers [29]. RT-PCR experiments with specific ΔNp73 isoform primers confirmed that alpha, and not beta, ΔNp73 is expressed in EBV-infected B-cells (Figure 1D). Accordingly, immunoblotting with a p73 antibody revealed a 65–70 kD protein band in LCLs that co-migrated with the ΔNp73α ectopically expressed in HEK 293 cells (Figure 1E).
Taken together, these data showed that EBV specifically activates ΔNp73α transcription in primary and cancer B-cells.
Studies on other oncogenic viruses demonstrated that alterations of p53-regulated pathways are normally induced by the viral oncoproteins [15], [30]. Figure 1B showed a correlation between ΔNp73α and LMP-1 expression levels, supporting the possible involvement of the viral oncoprotein in ΔNp73α up-regulation. Therefore, we determined later whether the major EBV transforming protein, LMP-1, was responsible for ΔNp73α accumulation. RPMI cells were transduced with empty (pLXSN) or LMP-1 expressing retrovirus (pLXSN-LMP-1) and ΔNp73α transcript and protein levels were determined by RT-PCR and immunoblotting, respectively. Both ΔNp73α mRNA and protein levels were strongly increased in RPMI/LMP-1 cells in comparison to cells infected with the empty retroviral vector (Figures 2A and B). To further demonstrate the role of LMP-1 in ΔNp73α up-regulation, we infected RPMI cells with a wild-type or mutated EBV, in which the LMP-1 gene was deleted (EBVΔLMP-1). Two RPMI/EBVΔLMP-1 cell lines were generated by two independent infections and ΔNp73α expression levels were compared with the ones of mock infected cells as well as RPMI/EBV. Deletion of LMP-1 gene from the EBV genome abolished ΔNp73α up-regulation, as shown by RT-PCR and immunoblot analyses (Figures 3 A and B). Transduction of RPMI/EBVΔLMP-1 cells with a recombinant retrovirus expressing LMP-1 (pLXSN LMP-1) restored the ability of EBV to promote ΔNp73α mRNA and protein accumulation (Figures 3C and D).
In summary, these data highlight the central role of LMP-1 in EBV-mediated ΔNp73α up-regulation.
The p73 p2 internal promoter contains a p53 responsive element (RE) which can be activated by both p53 and p73 [31]–[33]. Therefore, we evaluated whether p53 and/or p73 are involved in the regulation of p2 promoter in the presence or absence of LMP-1. Chromatin immune precipitation (ChIP) experiments using the p53 null SaOS-2 cells as experimental model showed that LMP-1 over-expression increased p73 binding affinity for the RE within the p2 promoter, while it did not influence p53 recruitment to the same site (Figure 4A). ChIP experiments with anti p73 antibody in RPMI cells transduced with empty (pLXSN) or pLXSN-LMP-1 retrovirus also showed an increased binding of p73 to the p2 promoter in presence of LMP1 (Figure 4B). In addition, DNA pull-down experiments, in which a biotinylated DNA probe containing a region of the p2 promoter encompassing the p53RE was incubated with cellular extracts of RPMI or RPMI LMP-1 cells, showed that LMP-1 increased p73 efficiency in binding DNA (Figure 4C). Silencing of p73 expression in LCL by shRNA correlated with down-regulation of ΔNp73α mRNA levels (Figure 4D). In contrast, targeting p53 with a siRNA in the same cells did not alter ΔNp73α levels (Figure 4D), indicating that p53 is not involved in the transcriptional regulation of ΔNp73α in LCLs.
We conclude from this set of data that LMP-1-mediated ΔNp73α transcriptional activation is partly due to the recruitment of p73 to the p2 promoter.
It is known that LMP-1 stimulates JNK-1, which in turn leads to p73 phosphorylation and increase in its transcriptional activity [34]–[36]. We therefore determined whether JNK-1 is involved in ΔNp73α accumulation mediated by LMP-1. Ectopic levels of JNK-1 in BJAB cells induced ΔNp73α accumulation (Figure 5A). In addition, treating LCLs with a specific inhibitor of JNK (SP600125) led to a time-dependent decrease in ΔNp73α mRNA levels (Figure 5B). JNK-1 down-regulation in LCL by siRNA also led to a decrease in ΔNp73α mRNA and protein levels (Figures 5C and D).
LMP-1 transforming activities lie mostly on two distinct domains in its cytoplasmic C-terminus, namely C-terminal activation region 1 (CTAR1) (amino acids 187–231) and CTAR2 (amino acids 351–386). As CTAR2 LMP-1 mutant (LMP-1/378 stop) is unable to activate JNK-1 [37], we determined whether deletion of CTAR2 affected LMP-1 ability to promote p73 and ΔNp73α accumulation. We first transfected SaOS-2 cells with HA tagged p73 in the presence of wild-type or 378 stop-mutant LMP-1. ChIP experiments performed with an HA antibody showed that only the wild-type LMP-1 increased p73 recruitment to p2 promoter (Figure 5E). According to previous data [35], immunoblotting showed that wild-type LMP-1, but not the LMP-1 378 stop-mutant that is unable to activate JNK-1, induced p73 accumulation (Figure 5F). Finally, ΔNp73α expression was only detected in RPMI/EBVΔLMP-1 cells containing the wild-type LMP-1 and not the LMP-1 378 stop-mutant (Figure 5G).
LMP-1 is also able, via the CTAR1 and CTAR2, to activate the cellular kinase p38 which in turn activates p73 [38]–[40]. However, p38 inhibition in LCLs by a chemical inhibitor did not result in a decrease of mRNA levels of ΔNp73α (data not shown), indicating that a different mechanism is involved in the event.
Together, these data highlight a crucial role of JNK-1 in LMP-1-mediated accumulation of p73 and ΔNp73α.
Emerging lines of evidence show that epigenetic changes play an important role in carcinogenesis [41]. Accordingly, several oncogenic viruses, including EBV, are able to hijack the epigenetic machinery in order to de-regulate cellular gene expression, to persist in the host cell and complete the viral cycle [42]. EZH2 is a component of the Polycomb 2 complex and is able to methylate the Histone H3 on Lysine 27, leading to chromatin condensation and gene silencing [43]. Interestingly, it has been recently shown that, upon interferon alpha treatment, EZH2 inhibits ΔNp73α expression in hepato-cellular carcinoma cells (HCC) by direct binding to the p2 promoter [44]. Therefore, we next determined whether in primary B cells and LCL the observed alterations in ΔNp73α expression may be ascribed to changes in the EZH2 recruitment to p2 promoter. ChIP experiments showed that EZH2 binds the p53 RE within the p2 promoter only in primary B cells, but not in LCL (Figure 6A). A similar pattern was observed in ChIP experiments performed with an antibody that specifically recognized H3K27 methylated form (Figure 6A). In contrast, acetylation on lysine 9 of the Histone H3, a marker of transcriptionally active chromatin, was increased at the p53/p73 RE in LCLs in comparison to primary B cells (Figure 6A). In addition, p73 was more efficiently recruited to p53/p73 RE in LCLs than primary B Cells (Figure 6B). Loss of EZH2 at ΔNp73α promoter appeared to be dependent on LMP-1 expression in LCLs and RPMI cells (Figure 6C and D).
According to these results, histone H4 hyperacetylation, another event associated with active transcription, is strongly enhanced at ΔNp73α promoter in LCLs in comparison to primary human B cells (Figure 6E). JNK can induce H4 hyperacetylation to regulate gene expression [45], accordingly inhibition of JNK-1 by siRNA or chemical inhibitor in LCLs resulted in the decrease of histone H4 hyperacetylation at ΔNp73α promoter (Figure 6E and F). In addition, the recruitment of p73 to the ΔNp73α promoter in LCL is strongly reduced upon JNK-1 inhibition (Figure 6F).
Immunoblotting showed that EZH2 is weakly detected in primary B cells, while its protein levels are considerably elevated in LCLs (Figure 6G and data not shown). Similarly, expression of LMP-1 in RPMI cells infected with EBV-ΔLMP-1 led to a substantial increase in EZH2 protein levels (Figure 6H). To determine whether the increase of EZH2 levels and decrease of its recruitment to ΔNp73 promoter in LMP-1 cells was due to changes in its localization, we performed cellular fractionation experiments followed by immunoblotting. LMP-1 expression did not alter EZH2 cellular localization, which appeared to be exclusively nuclear in primary B cells and LCLs (Figure 6I). However, immuno-fluorescence experiments with an anti-EZH2 antibody highlighted a different pattern of staining in primary B cells and LCLs, indicating that LMP-1 induced a redistribution of EZH2 in the nucleus (Figure 6J). Indeed, EZH2 staining appears to be punctuated in primary B cells, while it is more diffuse in LCLs (Figure 6J).
Together the data show that ΔNp73α expression mediated by EBV LMP-1 correlates with the release of EZH2 from the ΔNp73 promoter, which results in the opening of the chromatin and in an increased access to transcription factors.
To investigate the role of ΔNp73α in EBV infected cells, we down-regulated its protein levels by expressing an anti ΔNp73α anti sense oligo-nucleotide (AS). The sense oligo-nucleotide (S) was used as negative control. Figure 7A shows that ΔNp73α was efficiently down-regulated in a dose-dependent manner by AS in LCLs. Most importantly, ΔNp73α down-regulation led to PARP cleavage, indicating that ΔNp73α is involved in the inhibition of apoptosis in EBV infected cells (Figure 7A). FACS analysis confirmed that transfection with AS promoted cellular death (Figure 7B). To gain more insight into the biological significance of ΔNp73α expression in EBV infected cells, we compared the transcriptome profiling by RNA-seq of LCLs transfected by S or AS at two different concentrations. The expression levels of approximately 253 genes were found differentially expressed (214 up-regulated, and 39 down-regulated, p value<0.01) in cells transfected with S and AS (Tables S2 and S3, supplementary material). Functional analysis of the data was conducted as explained in Materials and Methods. It has been previously shown that ΔNp73α, but not ΔNp63α, plays a role in nervous system development and in the prevention of nerve growth factor-induced p53-mediated apoptosis [46]. Accordingly, we observed that the decrease in ΔNp73α levels by AS in LCL also led to the deregulation of several genes encoding products that are involved in development, e.g. in neurogenesis and in the regulation of apoptosis in neurons (Tables S2 and S3, supplementary material), indicating the specificity of our approach and validating our RNA seq analysis. Most importantly, AS ΔNp73α up-regulated genes encoding proteins, which play crucial roles in cellular transformation processes, such as apoptosis, cell cycle, DNA-repair and signaling pathways, i.e. NF-κB, Notch, RAS, and toll-like receptors (TLRs) (Table S2 and S3, supplementary material). In agreement with the functions of ΔNp73α as an antagonist of p53, several p53-regulated pro-apoptotic genes were found up-regulated in LCLs transfected by AS (Figure 7C and Tables S2 and S3, supplementary material). Interestingly, the inhibition of ΔNp73α by AS in LCLs resulted in a strong increase of PLK2 mRNA levels. PLK2 is a p53-regulated gene and encodes a serine threonine kinase which is involved in cell cycle regulation and cellular response to stresses. Its expression has been often found silenced by promoter methylation in Burkitt's Lymphomas [47]. ChIP experiments in SaOS-2 cells expressing HA-tagged-ΔNp73α demonstrated its binding to the promoter of PLK2 to the p53 binding site 1, which was further increased in the presence of LMP-1, while p73 recruitment to the same promoter was not influenced by the viral oncoprotein (Figure 7D). To corroborate the RNA-seq data, we have down-regulated ΔNp73α and determined PLK2 mRNA levels by quantitative PCR. Figure 7E shows that PLK2 and Pig3 are down-regulated in LCL in comparison to primary B cells. In addition down-regulation of ΔNp73α resulted in a rescue of their expression confirming the RNA seq data. Another gene that was found up-regulated in ΔNp73α AS-transfected LCLs is KLHDC8B which also appears to be associated with lymphomagenesis and is often found mutated in familiar and sporadic Hodgkin lymphomas [48]. KLHDC8B mRNA levels were increased by 3.66 and 18.44 folds in LCLs transfected with low and high doses of ΔNp73α AS, respectively (p value = 0.001693). Taken together, these data show that in LCL, ΔNp73α plays a key role in regulating cellular genes, the products of which exert important functions in cellular transformation, including two genes, PLK2 and KLHDC8B, that have been previously reported to be associated with lymphomagenesis.
Oncogenic viruses share the ability to target key pathways involved in preventing cellular transformation, considerably increasing the probability of an infected cell to evolve towards malignancy. One of the best characterized mechanisms of oncogenic viruses is the ability to inhibit the function of the tumor suppressor p53, a transcription factor that can trigger cell cycle arrest or apoptosis in response to stress or DNA damage [49]. Many oncogenic viruses, such as HR mucosal HPV types [15], EBV [10], [12], [14], Human T-cell Lymphotropic Virus (HTLV-1) [50], [51], Kaposi's sarcoma-associated herpesvirus (KSHV) [52]–[54] have developed strategies to inactivate p53. We have recently described a novel mechanism of deregulation of p53 transcriptional functions by the beta cutaneous HPV38 which appears, together with other beta HPV types, to be linked to skin carcinogenesis. HPV38 E7 oncoprotein promotes accumulation of ΔNp73α increasing its transcription and protein half-life ([20], [21] and our unpublished data). In turn, ΔNp73α competes with p53 for binding to p53 RE elements, preventing the activation of p53-regulated genes. Although the involvement of HPV38 in NMSC is still under debate, the demonstration that also other well-established oncogenic viruses promote ΔNp73α accumulation will further highlight the importance of the event and corroborate the potential role of HPV38 in human carcinogenesis. In this study, we show that the oncoprotein LMP-1 from EBV activates the transcription of ΔNp73, favoring the recruitment of p73 to cis p53 element of ΔNp73 p2 promoter. We also demonstrated that LMP-1-mediated up-regulation of ΔNp73α transcription is dependent on JNK-1, a kinase strongly activated by LMP-1. JNK-1 inhibition by different means strongly decreased ΔNp73α expression in EBV-infected cells. In addition, expression of ectopic levels of JNK-1 in the EBV-negative B-lymphoma cell line, BJAB, resulted in the activation of ΔNp73α transcription. Accordingly, LMP-1 mutants lacking the JNK-1 activating domain (CTAR2) did not influence the ΔNp73α expression levels. It is well established that several amino acid residues of p73 are phosphorylated by JNK-1 [34]. Therefore, it is likely that p73 recruitment to the ΔNp73 promoter is mediated by its JNK-1-dependent phosphorylation. Additional experiments are required to confirm this hypothesis and establish whether the p73 affinity for ΔNp73 promoter is determined by phosphorylation of one or more specific amino acids. Previous studies have shown that inhibition of JNK-1 in LMP-1 expressing cells led to decrease of cdc2 levels and cell cycle arrest [55]. In our experimental model the inhibition of cdc2 in LCLs by the chemical inhibitor roscovitine slightly affected ΔNp73α levels (data not shown).
JNK-1 could also induce ΔNp73 transcription by an alternative mechanism via activation of the proto-oncogene c-Jun. It has been shown that p73 acts in synergistic manner with c-Jun in promoting cellular survival [56]. This event is well explained by their cooperative ability to activate the transcription of specific subsets of cellular genes. ChIP-seq experiments have revealed the presence of AP1-binding motifs in close proximity to the p73 cis elements in promoters of genes encoding proteins with anti-apoptotic functions [57]. Brigati et al. have shown that TPA treatment of Germinal Center B cells, able to induce ΔNp73α expression, also leads to binding of c-Jun to an AP1 site which was located on the promoter of ΔNp73α, just upstream the p53/p73 RE [28]. According to these findings, we observed that in LCLs c-Jun is recruited to an AP1 cis element closely located to the p53/p73RE of ΔNp73 promoter (our unpublished data).
ChIP experiments in primary and EBV-immortalized B cells showed that activation of ΔNp73 promoter by the recruitment of p73 correlated with the displacement of the polycomb 2 complex component EZH2 and epigenetic changes. The apparently paradoxical finding that EBV infected B cells, despite the increased intracellular levels of EZH2, show reduced amount of EZH2 and lower levels in H3K27 methylation on the promoter of ΔNp73, recalls the scenario observed in HPV16 E6/E7 expressing cells [58]. Hyland et al. observed increased levels of EZH2 in the presence of HPV16 E6 and E7 proteins, which correlated with a decrease of H3K27 methylation. The authors explained that this phenomenon was due to an increase in KDM6A and KDM6B levels, two demethylase enzymes, and a decrease in BMI1, a Polycomb1 protein which stabilizes Polycomb 2-mediated methylation. According to this model, EBV is able to trigger accumulation of KDM6B via LMP-1 ([59] and our unpublished data) as well as a reduction of BMI1 levels (our unpublished data). Accumulation of EZH2 in cells expressing LMP-1 could be a consequence of post-translational modifications that negatively regulate its enzymatic activity. Accordingly, it has been previously shown that phosphorylation of EZH2 by AKT on serine 21 suppresses methylation of lysine 27 in Histone 3 [60]. It has been reported that EBV LMP-1 triggers the AKT pathway [61] which is often found activated in NPC and Hodgkin's lymphomas [62], [63]. Based on these findings, we could speculate that the loss of EZH2 recruitment to ΔNp73 promoter is due to serine 21 phosphorylation. We are currently assessing this hypothesis.
High levels of p73 and ΔNp73 have been observed in B cell chronic lymphocytic Leukemia [64]. Although resting B cells do not express ΔNp73, epigenetic changes leading to ΔNp73 up-regulation were observed in the activated B cells compartment of the germinative center of the tonsil. Thus, it is likely that the mechanisms characterized in EBV-infected cells in this study may also occur in different scenarios independently of the presence of the viral oncoprotein.
To evaluate the biological significance of EBV-mediated ΔNp73α over-expression in the transformation of B cell, we down-regulated ΔNp73α expression by AS in LCL, and compared the cellular expression profiling with one of the S transfected LCL. Decrease in ΔNp73α levels led to the alteration of the expression of cellular genes linked to neurogenesis as well as to the regulation of apoptosis in neurons. These results are consistent with the known in vivo functions of ΔNp73 in brain development and in the prevention of nerve growth factor induced p53-mediated apoptosis [46]. An additional cluster of genes that appeared down-regulated by ΔNp73α in LCL is a group of Homeobox genes. To our knowledge, this is the first time that ΔNp73α has been shown to be implicated in the regulation of HOX genes. Since both HOX genes and ΔNp73α are aberrantly expressed in cancer cells, these findings, if confirmed, could further contribute to the understanding of the events associated with carcinogenesis [22], [65].
Loss of PLK2 expression by promoter CpG island methylation is one of the most common epigenetic events in B-cell lymphomas [47]. It is worth noting that we found PLK2 gene strongly up-regulated upon inhibition of ΔNp73α in LCL. It is well known that PLK2 promoter is positively regulated by p53. Thus, it is highly likely that ΔNp73α induces PLK2 down-regulation by altering the p53 transcriptional function. For the first time, our data provide evidence for a link between PLK2 expression silencing and ΔNp73α in EBV-infected cells. Our findings also suggest that LMP-1 increases the affinity of ΔNp73α for PLK2 promoter without altering its protein levels (HA-ΔNp73α: Figure 7D and our unpublished data). It is possible that the viral oncoprotein, independently of its ability to positively regulate the ΔNp73α transcription, may increase ΔNp73α affinity for PLK2 promoter by promoting post-translational modifications.
Similarly to PLK2, KLHDC8B has been linked to lymphomagenesis. Multiple cases of Hodgkin's lymphoma with translocations or polymorphisms affecting KLHDC8B have been reported in the same family [66], indicating that inhibition of its expression plays a role in B cell transformation. Our data show that the expression of KLHDC8B was restored upon ΔNp73α down-regulation. According to the well characterized ΔNp73α ability to act as an inhibitor of p53, the expression levels of different p53 responsive genes (MDM2, APAF1, GADD45, BAX, CCND1, etc.) increased significantly after ΔNp73 down-regulation, leading to apoptosis. Other subgroups of genes that resulted to be regulated by ΔNp73α are the ones involved in lymphocyte migration and proliferation, cytokine production and innate immune response. As a whole, the RNA seq data indicate that in EBV infected cells ΔNp73α may contribute to the development of EBV-associated disease in several ways, e.g. by inhibiting apoptosis, promoting B cell growth, as well as by modulating host defence machinery allowing EBV persistence.
In summary, our data underline the important function of ΔNp73α in EBV-induced cellular transformation, unveiling novel links between its accumulation and deregulation of the expression of many cellular genes. However, the degree to which ΔNp73α oncogenenic effects exceed tumor suppressor effects of p73 activation remains to be better determined.
Cellular and viral genes were expressed using the retroviral vector pLXSN (Clontech, Palo Alto, CA) or the expression vector pcDNA-3 (Invitrogen). The pLXSN-LMP-1 construct has been previously described (68). The constructs pLXSN-LMP-1 mutants (LMP-1AxAxA, LMP-1 378 STOP, and LMP-1AxAxA/378STOP) were generated in this study using standard molecular biology techniques. The constructs pcDNA3 HA-ΔNp73α, pcDNA3 HA-p73α were previously described [20].
RPMI 8226 cells (harbouring mutated p53: Glu 285 to Lys) (RPMI) were kindly provided by Dr Christophe Caux (Centre Léon Bérard, Lyon, France). RPMI pLXSN or pLXSN-LMP-1 were generated as described in Fathallah et al. [67]. RPMI EBV and EBVΔLMP-1 were obtained as in [68]. Expression of LMP-1 wild-type or LMP-1 378 stop mutant in RPMI EBVΔLMP-1 was achieved by transduction with recombinant retroviruses [67]. The EBV-immortalized lymphoblastoid cell lines (LCLs), ATM14, ATM11, 48513F, 48513M, 2095, 2145 and EBV negative immortalized B cells, BJAB (with mutated p53: His 193 to Arg), were previously described [69] (or generated at IARC, Lyon France). Several LCLs were generated in this study by infecting primary B cells isolated from different donors as previously described [67]. Primary and immortalized B cells were cultured in RPMI 1640 medium (GIBCO; Invitrogen life Technologies, Cergy-Pontoise, France) supplemented with 10% FBS, 100 U/ml penicillin G, 100 mg/ml streptomycin, 2 mM L-glutamine, and 1 mM sodium pyruvate (PAA, Pasching, Austria). SaOS-2 and human embryonic kidney cells (HEK293) were cultured in foetal calf serum (FCS) and Dulbecco's modified Eagle medium (DMEM) (Gibco) using standard culturing conditions. Cells were treated with JNK inhibitor SP600125 at the final concentration of 20 µM. DNA plasmids were transiently transfected in cells by using FuGENE6 or Xtreme gene 9 reagents (Roche) according to the purchased protocol.
Immuno-fluorescence in primary and immortalized B cells was performed as described in Fathallah et al. [67].
For FACS staining, cells were collected and washed twice in PBS, then stained with Propidium iodide (PI) at the final concentration of 5 µg/ml. Subsequently, cells were analyzed for the % of dead cells by FACS CANTO (Becton Dickinson).
Gene silencing of SAPK/JNK was performed using SignalSilence SAPK/JNK siRNAI (6269 cell signalling). Cells (8×105) were transfected with siRNA to the final concentration of 100 nM by oligofectamine (invitrogene) according to the manufacture protocol. p53 gene silencing was performed as in Accardi et al. [20]. ΔNp73α levels were down-regulated by electroporating 1.5×106 cells with either 0.5 or 2 µg of AS or S oligos (for AS and S oligos sequences please see [20]). Cells electroporation was performed by Neon Transfection System, using a pulse voltage of 1350 v and a pulse width of 30 ms. To specifically silence p73 isoform we used the target sequence: 5′-CAGACAGCACCTACTTCGA-3′ spanning from +71 to +90 bp downstream of the transcription start codon was cloned in OmicsLink shRNA Expression system containing a puromycin selection marker (HSH018180-6-HIVmH1, OS395979, GeneCopoeia). As negative control, a scrambled shRNA (sH1) was used. Lentivirus production was performed as previously described [70]. Lentiviral suspension was added to 1.5×106 LCLs and selection with puromycin (0.8 µg/ml) was initiated 24 hours later. One week post infection cells were collected and processed for the different experiments.
Total cellular RNA was extracted from cells using the Absolutely RNA Miniprep kit (Stratagene). RNA Reverse transcription to cDNA was carried out by RevertAid H Minus M-MuLV Reverse Transcriptase (MBI Fermentas) according to manufacturer's protocol. Quantitative PCR (Q-PCR) was performed in duplicate in each experiment as previously described [67]. The primer sequences used for RT and Q-PCR are indicated in Table S1A (supplementary material).
Preparation of whole or fractioned cell lysates extracts, sodium dodecyl sulfate -polyacrylamide gel electrophoresis (PAGE), and immunoblotting (IB) were performed as previously described [20], [21]. The following antibodies were used for IB: β-actin (C4; MP Biomedicals), human p53 (NCL-CM1; Novocastra Laboratories Ltd.), p73 (anti-p73 Ab-1; Calbiochem), hemagglutinin (HA)-peroxidase-high affinity (3F10; Roche), anti-LMP-1 Ab (S12; a gift from Georges Mosialos, Alexander Fleming Institute, Varkiza, Greece), PARP (9542; Cell signalling), SAP/JNK1(56G8; Cell signalling), anti-EZH2 (AC22; Cell signalling).
A biotinylated fragment of ΔNp73 promoter was generated by PCR using the genomic DNA as template and a biotinylated primer Fw 5′-Btndt CTGGTGGGTTTAATTA-3′ and a non-biotinylated primer Rev 5′- AGGAGCCGAGGATGCTGG-3′ (Sigma). Cells were re-suspended in lysis buffer HKMG (10 mM HEPES, pH7.9, 100 mM KCl, 5 mM MgCl2, 10% Glycerol, 1 mM DTT and 0.5% NP-40) incubated in ice for 10 min, then lysed by sonication (25% Amp, 1 min). One mg of total cellular extracts was pre-cleared with 40 µl of streptavidine-agarose beads (Amersham Bioscience) for 1 hour at 4°C, and then incubated with 2 µg of purified DNA biotinylated probe for 16 hours at 4°C. Poly dI-dC (40 µg) was added to the reaction to avoid unspecific binding. DNA bound proteins were recovered by incubating with 60 µl of streptavidine-agarose beads for 1 hour at 4°C and washed several times with HKMG buffer [71]. Beads were resuspended in 1× SDS-PAGE loading buffer and analyzed by immunoblotting.
Chromatin immunoprecipitation (ChIP) was performed with Diagenode Shearing ChIP and OneDay ChIP kits or with LowCell ChIP Kit according to the manufacturers' protocols, using the following antibodies: EZH2 (AC22; Cell signalling), Histone 3 Lysine 27 Trimethyl polyclonal antibody H3K27 (Epigentek), Histon 3 acetylated lysine 9 antibody H3K9Ac (9649S; cell signalling), Acetyl-Histone H4 (17–630; Millipore), HA hemagglutinin (HA) high affinity (3F10; Roche), and p73 (anti-p73 Ab-1; Calbiochem). The eluted DNA was used as template for Quantitative PCR as previously described [21]. Primers for Quantitative ChIP are listed in Table S1B (supplementary material).
RNA integrity and quantification of the total cellular RNA from LCLs tranfected with ΔNp73α AS and S oligo-nucleotide were characterized by measuring the 28s/18s rRNA ratio and RIN using the Agilent 2100 bioanalyzer instrument, and the Agilent RNA 6000 Nano kit. 5 µg of total cellular RNAs was depleted from rRNA using the Invitrogen RiboMinus Eukaryote kit according to manufacturer's standard protocol. The absence of 28s/18s rRNA was checked on the Agilent 2100 bioanalyzer instrument. Five hundred ng of each sample were enzymatically fragmented using 1 unit of RNase III provided in the SOLiD Total RNA-seq Kit, incubated at 37°C for 10 minutes and cleaned up using the Invitrogen RiboMinus Concentration Module according to manufacturer's standard protocol. RNA yield and size distribution were assessed with the Agilent 2100 Bioanalyzer instrument and the Agilent RNA 6000 Pico kit. The amplified whole transcriptome library for each sample was constructed according to Lifetechnologies's SOLiD Total RNA-seq Kit protocol (PN 4452437 Rev.B). To summarize, adaptors were hybridised and ligated to 100 ng of fragmented rRNA-depleted total RNAs followed by the construction and subsequent purifications of cDNAs using successively the SOLiD Total RNA-seq Kit and the Agencourt AMPure beads. cDNAs were then barcoded, amplified with 15 cycles of PCR and purified using the Invitrogen PureLink PCR Micro Kit. Yield and size distribution of the amplified DNA libraries were assessed with the Agilent 2100 Bioanalyzer instrument and the Agilent DNA 1000 kit. After minimizing the DNA in the 25–200 bp range, 0.4 pM of each barcoded libraries were pooled at equimolar concentrations prior to template bead preparation, in which the pooled library is clonally amplified by emulsion PCR following the Lifetechnologies's SOLiD EZ bead E80 protocols (PN 4441486 Rev. D, 4443494 Rev. D, 4443496 Rev. D). Two hundred and forty µl of emPCR beads were 3′ modified and deposited on 4 lanes FlowChip before being incubated for 60 minutes at 37°C. The forward 50 bp reads sequencing chemistry was applied.
The secondary and tertiary analyses was done with LifeScope software v. 2.5.1 from Life Technologies (Build ID:LifeScope-v2.5.1-r0_102906_20120406100430)
The raw data (xsq files) from each lane were grouped per sample (based on the barcodes) before launching the standard RNA seq workflow on the 3 samples (EBV_sense, EBV_antisens1, EBV_antisens2). We kept all the standard parameters as advised by Life Technologies.
This workflow includes 3 modules: the “mapping analysis” for which we used hg19 as reference genome, the “coverage analysis” and the “count known genes and exons analysis”.
After reads mapping, the R/Bioconductor package edgeR (empirical analysis of digital gene expression data in R) was used to study differential gene expression [72]. After fitting a negative binomial model, data obtained from antisense samples were grouped before applying the “common dispersion” function in edgeR. Next, differential gene expression was determined using the exact test. Heatmaps and gene set expression comparisons were performed with BRB-ArrayTools software Version 4.2.1. To this end, reads were RPKM (Reads Per Kilobase of exon model per Million mapped reads) normalized and corresponding gene lists were filtered for selected pathways (Table S3, supplementary material).
Statistical significance was determined by Student T test. The p value of each experiment is indicated in the corresponding Figure legend. Error bars in the graphs represent the standard deviation.
|
10.1371/journal.pcbi.1005283 | The Multilayer Connectome of Caenorhabditis elegans | Connectomics has focused primarily on the mapping of synaptic links in the brain; yet it is well established that extrasynaptic volume transmission, especially via monoamines and neuropeptides, is also critical to brain function and occurs primarily outside the synaptic connectome. We have mapped the putative monoamine connections, as well as a subset of neuropeptide connections, in C. elegans based on new and published gene expression data. The monoamine and neuropeptide networks exhibit distinct topological properties, with the monoamine network displaying a highly disassortative star-like structure with a rich-club of interconnected broadcasting hubs, and the neuropeptide network showing a more recurrent, highly clustered topology. Despite the low degree of overlap between the extrasynaptic (or wireless) and synaptic (or wired) connectomes, we find highly significant multilink motifs of interaction, pinpointing locations in the network where aminergic and neuropeptide signalling modulate synaptic activity. Thus, the C. elegans connectome can be mapped as a multiplex network with synaptic, gap junction, and neuromodulator layers representing alternative modes of interaction between neurons. This provides a new topological plan for understanding how aminergic and peptidergic modulation of behaviour is achieved by specific motifs and loci of integration between hard-wired synaptic or junctional circuits and extrasynaptic signals wirelessly broadcast from a small number of modulatory neurons.
| Connectomics represents an effort to map brain structure at the level of individual neurons and their synaptic connections. However, neural circuits also depend on other types of signalling between neurons, such as extrasynaptic modulation by monoamines and peptides. Here we present a draft monoamine connectome, along with a partial neuropeptide connectome, for the nematode C. elegans, based on new and published expression data for biosynthetic genes and receptors. We describe the structural properties of these "wireless" networks, including their topological features and modes of interaction with the wired synaptic and gap junction connectomes. This multilayer connectome of C. elegans can serve as a prototype for understanding the multiplex networks comprising larger nervous systems, including the human brain.
| The new field of connectomics seeks to understand the brain by comprehensively mapping the anatomical and functional links between all its constituent neurons or larger scale brain regions [1]. The C. elegans nervous system has served as a prototype for analytical studies of connectome networks, since the synaptic connections made by each of its 302 neurons have been completely mapped at the level of electron microscopy [2, 3]. Through this approach, the C. elegans nervous system has been found to share a number of topological features in common with most other real-world networks, from human brain networks through social networks to the internet [1, 4, 5]. One well-known example is the small-world phenomenon, whereby networks are simultaneously highly clustered (nodes that are connected to each other are also likely to have many nearest neighbours in common) and highly efficient (the average path length between a pair of nodes is short) [6, 7]. Another characteristic feature of real-world networks which has attracted much attention is the existence of hubs or high-degree nodes, with many more connections to the rest of the network than expected in a random graph [8]. As in other networks, these topological features of the C. elegans connectome are thought to reflect the functional needs of the system [9, 10]. For example hubs are known to play a privileged role in coordinating functions across a distributed network [11], while the short path lengths (often mediated by the hubs) help increase the efficiency of information transfer across the network [6].
Although connectomics has primarily focused on mapping the synaptic links between neurons, it is well established that chemical synapses are only one of several modes of interaction between neurons. For example, gap junctions, which mediate fast, potentially bidirectional electrical coupling between cells, are widespread in all nervous systems. Likewise, volume transmission and neurohumoral signalling provide means for local or long-range communication between neurons unconnected by synapses. As neuromodulators released through these routes can have profound effects on neural activity and behaviour [12–14], a full understanding of neural connectivity requires a detailed mapping of these extrasynaptic pathways.
In C. elegans, as in many animals, one important route of neuromodulation is through monoamine signalling. Monoamines are widespread throughout phyla, with evidence that they are one of the oldest signalling systems, evolving at least 1 billion years ago [15]. In both humans and C. elegans, many neurons expressing aminergic receptors are not post-synaptic to releasing neurons, indicating that a significant amount of monoamine signalling occurs outside the wired connectome [16]. Monoamines are known to be essential for normal brain function, with abnormal signalling being implicated in numerous neurological and psychiatric conditions [17]. In C. elegans, these monoaminergic systems play similarly diverse roles in regulating locomotion, reproduction, feeding states, sensory adaptation, and learning [16]. Clearly, if the goal of connectomics is to understand behaviourally relevant communication within the brain, extrasynaptic monoamine interactions must also be mapped, not just the network of wired chemical synapses and gap junctions.
In addition to monoamines, neuropeptides are also widely used as neuromodulators in the C. elegans nervous system. C. elegans contains over 250 known or predicted neuropeptides synthesized from at least 122 precursor genes, and over 100 putative peptide receptors [18, 19]. These include homologues of several well-known vertebrate neuropeptide receptors, including those for oxytocin/vasopressin (NTR-1), neuropeptide Y (NPR-1) and cholecystokinin (CKR-2) [19]. As in other animals, neuropeptide signalling is critical for nervous system function, and frequently involves hormonal or other extrasynaptic mechanisms.
This study describes a draft connectome of extrasynaptic monoamine signalling in C. elegans, as well as a partial network of neuropeptide signalling, based on new and published gene expression data. We find that the extrasynaptic connectomes exhibit topological properties distinct from one another as well as from the wired connectome. Overall, the neuronal connectome can be modelled as a multiplex network with structurally distinct synaptic, gap junction, and extrasynaptic (neuromodulatory) layers representing neuronal interactions with different dynamics and polarity, and with critical interaction points allowing communication between layers. This network represents a prototype for understanding how neuromodulators interact with wired circuitry in larger nervous systems and for understanding the organisational principles of multiplex networks.
To investigate the extent of extrasynaptic signalling in C. elegans monoamine systems, we systematically compared the expression patterns of monoamine receptors with the postsynaptic targets of aminergic neurons. Monoamine-producing cells were identified based on the published expression patterns of appropriate biosynthetic enzymes and vesicular transporters (see Methods). The expression patterns for each of five serotonin receptors (ser-1, ser-4, ser-5, ser-7 and mod-1), three octopamine receptors (octr-1, ser-3 and ser-6), four tyramine receptors (ser-2, tyra-2, tyra-3 and lgc-55), and four dopamine receptors (dop-1, dop-2, dop-3 and dop-4) were compiled from published data (see S1–S7 Tables). Since these receptors are either ion channels or serpentine receptors predicted to couple to pan-neuronal G-proteins, we therefore assumed all neurons expressing monoamine receptors are potential monoamine-responding cells.
Three additional genes encode known or candidate monoamine receptors but have missing or incomplete expression data. Specifically, a ligand-gated chloride channel, lgc-53, has been shown to be activated by dopamine [20], but its expression pattern and biological function have not been characterized. Additional expression profiling using a transgenic lgc-53 reporter line crossed to a series of known reference strains indicated that lgc-53 is expressed in a small subset of neurons in the head, body and tail (Fig 1). Together with the published dop-1, dop-2, dop-3 and dop-4-expressing cells, these were inferred to make up the domain of dopamine-responding neurons. In addition, two G-protein coupled receptors, dop-5 and dop-6, have been hypothesized based on sequence similarity to dop-3 to be dopamine receptors. Using the same approach used for lgc-53, we identified most of the cells with clear expression of dop-5 and dop-6 reporters (Fig 1). These cells were included in a broader provisional dopamine network, the analysis of which is presented in the supplemental material (S1 Fig, S3 Fig).
Receptor expression patterns suggest that a remarkably high fraction of monoamine signalling must be extrasynaptic. For example, the two tyraminergic neurons, RIML and RIMR, are presynaptic to a total of 20 neurons. Yet of the 114 neurons that express reporters for one or more of the four tyramine (TA) receptors, only 7 are postsynaptic to a tyraminergic neuron (Fig 2A; Table 1). Thus, approximately 94% of tyramine-responsive neurons must respond only to extrasynaptic TA. Similar analyses of the other monoamine systems yield comparable results: 100% of neurons expressing octopamine receptors receive no synaptic input from octopamine-releasing neurons (Fig 2B), while 82% of neurons expressing dopamine receptors, and 76% of neurons expressing serotonin receptors receive no synaptic input from neurons expressing the cognate monoamine ligand (Table 1). Thus, most neuronal monoamine signalling in C. elegans appears to occur extrasynaptically, outside the wired synaptic connectome. The prevalence of extrasynaptic monoamine signalling between neurons unconnected by synapses or gap junctions implies the existence of a large wireless component to the functional C. elegans connectome, the properties of which have not previously been studied.
Using the gene expression data, a directed graph representing a draft aminergic connectome was constructed with edges linking putative monoamine releasing cells (expressing monoamines, biosynthetic enzymes, or transporters) to those cells expressing a paired receptor (Fig 2C; Table 2; S1 Dataset). Since biologically-relevant long-distance signalling (e.g. from releasing cells in the head to tail motoneurons) has been experimentally demonstrated in C. elegans for both dopamine and serotonin [21, 22]–while tyramine and octopamine are each released from a single neuronal class [16]–edges were not restricted based on the physical distance between nodes. For the serotonin network, only those neurons with strong, consistent expression of serotonin biosynthetic markers such as tryptophan hydroxylase were included (NSM, HSN and ADF). Additional neurons (AIM, RIH, VC4/5) that appear to take up serotonin but not synthesize it [23][24] were not included in the network, since they may function primarily in the homeostatic clearing of serotonin. We also did not include the ASG neurons, which produce serotonin only under hypoxic conditions [25], though they are likely to participate conditionally in the serotonin signalling networks.
With the inclusion of the monoamine systems, the full C. elegans connectome can be considered as a multiplex or multilayer network [26], with each node representing a neuron and each layer of connections–synaptic, gap junction, and monoamine–characterized by distinct edge properties (Fig 2D). For example, chemical synapses represent unidirectional, wired connections that signal on a fast (ms) time scale, while gap junctions generate reciprocal electrical connections that function on an even faster time scale. In contrast, monoamine connections are wireless (with a single sending cell broadcasting to multiple receivers), slow (acting on a time scale of seconds or longer) and unidirectional [22, 27]. Conceptually, additional modes of signalling between neurons, such as peptide neuromodulation, could represent additional layers.
Prior studies of multiplex networks in non-biological systems–such as communication networks–have tended to find a large degree of overlap between the links observed in distinct layers, implying that they may not be truly independent channels of interaction [28]. In contrast, we observe that out of 1940 monoamine connections only 80 overlap with chemical or electrical synapses, meaning 96% of the monoamine connections are unique to the monoamine layer (Fig 2C; Table 1). Reducibility analysis [28], which clusters the different network layers based on their redundancy or degree of overlap, provides further support that the monoamine networks have a unique structure. Considered either separately or in the aggregate, the monoamines form a distinct cluster separate from the wired synaptic and gap junction networks (Fig 3A and 3B). This shows that the monoamine networks overlap less with the synaptic and gap junction networks than the synaptic and gap junction networks do with each other.
Similarly, in many previously-described multiplex networks, the high-degree hubs in each layer are often co-located, unequivocally highlighting certain nodes as key controllers of information flow in the system [26]. While the synaptic and gap junction layers of the worm connectome are observed to follow this trend, with the same high-degree neurons in both systems (Fig 3C), the extrasynaptic monoamine network exhibits a vastly different structure. While the synaptic and gap junction degrees of individual nodes show high positive correlation (R = .594), no significant degree-degree correlation is observed between the wired and extrasynaptic monoamine layers, indicating that the hubs of the monoamine system are distinct. These analyses suggest two distinct interpretations for the dissimilarity to the wired network layers. Firstly, monoamines may be functioning as an independent network, with little relation to the faster wired network. Secondly, the dissimilarity between layers might indicate that monoamines have a complementary function that is nevertheless coupled to that of the synaptic and gap junction connections.
To address these possibilities, we investigated whether the isolated C. elegans monoamine network displays the structural organisation required for information processing. Considered separately, the monoamine networks of C. elegans consist of only a few topologically central neurons that broadcast signals to a large number of peripheral neurons. These monoamine-releasing cells are mostly sensory and motor neurons, with the downstream receptors being distributed throughout the worm (Fig 3D). In total, 18 of the 302 neurons in the adult hermaphrodite release monoamines, while 251 neurons (83%) were found to express one or more monoamine receptors. This gives the network a star-like topology, which can be directly observed in all of the separate monoamine networks (Fig 4A, S1 Fig). As a consequence, the monoamine network exhibits a heavy tailed distribution containing a small number of high-degree hubs (Fig 3C). This structure is also reflected in other topological network measures, with the monoamine network exhibiting high disassortativity characteristic of star networks (Fig 4B). Disassortativity is known to be relevant in the organisation of collective network dynamics, such as synchronisation [29] and cooperation behaviour [30, 31], and is widely observed in other biological and technological networks [32]. The star-like structure of the monoamine layer was also confirmed by three-neuron motif analysis, which revealed the enrichment of a motif consisting of a hub node signalling to two spokes (S2 Fig).
The inclusion of these additional monoamine connections into the connectome has a number of effects on the aggregate network. For one, it greatly reduces the overall path length of the network (Fig 4C), increasing the efficiency of integrative information processing by providing paths between more segregated subgraphs of the wired network [33]. In particular, monoamine signalling provides a direct route of communication between sensory neurons and motor neurons (Fig 3D), bypassing the premotor interneurons that play a prominent role in the synaptic and gap junction systems [11]. Together, these observations suggest that the monoamines provide efficient global connections for coordinating behaviour throughout the entire organism due to the presence of highly connected hubs directly linking many disparate parts of the network. This is a useful feature given the role of monoamines in signalling physiologically important states relevant to the entire organism, such as food availability [27]. The increased connectivity provided by the monoamines also results in a reduction in the aggregate network's modular structure, a consequence of increasing the number of connections between functionally segregated units (Fig 4D). The network is, however, still more modular than random, with the monoamine layer also exhibiting greater-than-random modularity compared to null models that rewire the network edges while preserving degree distribution (see Methods). This is expected given the monoamine layer's composition from separate signalling systems; indeed the individual monoamine networks considered on their own show very low modularity (S1 Fig).
Despite the hub-and-spoke structure of the extrasynaptic network, the monoamine layer exhibits a significant level of global clustering (measured here as transitivity) (Fig 4E). This observation is explained by two factors. Firstly, the expression of monoamine receptors by releasing neurons creates a central cluster of hub neurons in the network; secondly, as many neurons also express more than one monoamine receptor, triangles are formed in the network with a receiving neuron as one vertex, and two transmitting neurons as the others. Indeed, three-neuron motif analysis confirmed that this configuration is overrepresented in all the monoamine networks save tyramine (S2 Fig). This structure provides a method of dual lateral inhibition, where a releasing neuron can inhibit antagonistic signals from another hub neuron while simultaneously negating the downstream effects of those signals, a pattern previously observed in the OA/TA and 5-HT systems between RIC/RIM & NSM in the aminergic control of feeding behaviours [34]. Similar patterns also exist within individual monoamine layers; for example, the ventral cord motor neurons express both excitatory (dop-1) and inhibitory (dop-3) dopamine receptors [35], while the expression of an inhibitory receptor (dop-2) in dopamine-releasing neurons suggests that the hubs mutually suppress one another to regulate dopamine release.
Many neural and brain networks have been shown to exhibit rich-club organisation [36–39] in which the most highly-connected nodes are more connected to one another than expected by chance [40]. It was previously shown that the C. elegans wired connectome includes a rich-club consisting primarily of a small number of premotor interneurons, controlling forward and backward locomotion [11]. Subjecting the monoamine connectome to similar analysis, it was found that this network also contains a distinct rich-club (Fig 5A and 5B; Table 3), consisting of dopamine, serotonin, and tyramine-releasing neurons. The rich-club property stems from the fact that most serotonergic neurons contain receptors for both tyramine and dopamine, while dopaminergic and tyraminergic neurons likewise express receptors for the other two aminergic transmitters (Fig 5B), suggesting that the different monoamines coordinate their actions. This rich-club structure is also reflected in the 3-neuron motif analysis, in which the fully-connected motif was overrepresented in the aggregate monoamine layer (S2 Fig). Interestingly, in contrast to the wired rich-club, all of whose members are interneurons, the monoamine rich-club consists of sensory neurons and motor neurons (Fig 5C, Table 3).
We next investigated the structure of the signalling network for neuropeptides. The receptors for many neuropeptides, and the ligands for many neuropeptide receptors, remain unknown; moreover, the distance over which signalling can occur is uncharacterized for most neuropeptide systems. Despite these caveats, we reasoned that a partial and provisional neuropeptide network could provide useful insight into the differences between peptide signalling networks and synaptic, gap junction and monoamine networks. We focused on 12 neuropeptide receptors with well-established ligands (with biologically-plausible EC50 values in in vitro assays) and precisely-characterized expression patterns for both receptor and peptide precursor genes (S8 Table, S9 Table). Networks were classified by receptor, allowing many-to-many relationships between neuropeptides and receptors. Even for this partial network, 239 neurons are seen to be involved in neuropeptide signalling (out of 302 possible) with 7035 connections between them, providing greater connectivity than either the synaptic or monoamine layers. Of the receptor-expressing neurons, almost 60% received no synaptic input from neurons expressing one of their ligands, suggesting that at least for this partial network, neuropeptide signalling, like monoamine signalling, is largely extrasynaptic. Likewise, the majority of edges in the neuropeptide network do not overlap with synapses (97% non-overlapping), again consistent with a largely extrasynaptic mode of signalling (Fig 6A).
The neuropeptide network, like the monoamine network, exhibits a structure distinct from the wired connectome. No significant degree correlation was observed between the partial neuropeptide network and the synaptic, gap junction, or monoamine networks, indicating that neuropeptide hubs are distinct from those in other layers (Fig 3C). Likewise, reducibility analysis shows low overlap between the neuropeptide edges and those in the monoamine, synaptic and gap junction layers (Fig 3A). Interestingly, some individual neuropeptide systems, in particular CKR-2, overlap significantly with the networks of monoamine systems, while others, including the neuropeptide F/Y receptors NPR-1/2/5/11, show little overlap with either the wired or other extrasynaptic networks (Fig 6B).
Examining the network measures for the neuropeptide network reveal it to have some topological properties in common with the monoamine network, but also crucial differences. For example, both networks have a shorter characteristic path length and lower modularity than the wired networks (Fig 4C and 4D). On the other hand, the neuropeptide network has much higher clustering than any other connectome layer (Fig 4E), and is significantly less disassortative (Fig 4B) than the monoamine network. In part, this is an expected consequence of the large number of connections in the neuropeptide network; however, the observed clustering in the neuropeptide network was significantly higher even than null models with the same edge density. In addition, the neuropeptide network shows much higher reciprocity than the monoamine network (Fig 4F), with the individual neuropeptide systems generally lacking the star-like topology characteristic of the monoamines (Fig 6C).
Despite the distinct structures and topologies of the different neuronal connectome layers, they are likely to interact in functionally significant ways. For example, although the wired and monoamine rich-clubs do not overlap, there are significant links between them (Fig 5C). To systematically identify neurons that have a role in linking all of the layers, neurons were first ordered according to the product of their degree-rank across the synaptic, gap junction and monoamine layers (Table 4). We observe that the highest ranking neurons, which have the highest participation across all layers, include three from the monoamine rich-club (RIML, RIMR, and ADEL) and two from the wired rich-club (RIBL and DVA). Indeed, the premotor interneuron DVA is a receiver for serotonin, tyramine and (provisionally) dopamine signalling, while the tyraminergic RIMs are highly connected to the premotor interneurons of the wired rich-club. As one might expect from their topological role in linking the monoamine and wired network layers, the RIMs have been shown in a number of studies to play a central role in the modulation of sensory pathways in response to feeding states as well as the control of downstream locomotion motor programs [41–43]. Similarly RIB, which expresses receptors for serotonin and dopamine, is thought to integrate numerous sensory signals [44, 45] and has been demonstrated to influence reorientation in foraging behaviour [46].
Multilink motif analysis provides another approach for investigating the interactions between the synaptic, gap junction and monoamine layers [47]. Since each layer contains the same set of nodes but a different pattern of edges, the frequencies with which different combinations of links co-occur between pairs of nodes throughout the multiplex network can be determined. Of the 20 possible multilink motifs, seven were found to be overrepresented and four underrepresented compared to networks composed from randomized layers (Fig 7). Many of these do not involve monoamines; for example, three overrepresented motifs–reciprocal chemical synapses (motif 3) and the co-occurrence of a gap junction with a single or reciprocal chemical synapse (motifs 5 & 6)–have been reported in an earlier analysis of the wired network [5]. These also align with results from the degree-degree correlation and reducibility (Fig 3A, 3B and 3C) indicating that synapses and gap junctions frequently overlap. This is mirrored in the underrepresentation of motifs 2 & 4 corresponding to synapses or gap junctions alone; conversely, the underrepresentation of these single link motifs leads to an overrepresentation of unlinked pairs (motif 1).
Although the overlap between monoamine and wired connectivity is low, multilink motif analysis revealed a few overrepresented motifs involving monoamines. The most interesting (and statistically significant) of these corresponds to a unidirectional monoamine link coincident with reciprocal synaptic connections (motif 10). The structure of this motif is well-suited to provide positive or negative feedback in response to experience, suggesting that this may be a functionally important aspect of monoamine activity within the wider network. Indeed, connections of this type (Table 6) have been implicated in a number of C. elegans behaviours; for example, motif 10 connections between ADF and AIY have been shown to be important for the learning of pathogen avoidance [48] and connections between RIM and RMD are important for the suppression of head movements during escape behaviour [49]. Putative motif 10 connections between PDE and DVA are also thought to play a role in controlling neuropeptide release [50].
Intriguingly, most examples of motif 10 (all except RIMR-RMDR) involve either serotonin or dopamine as the monoamine transmitter. Indeed, when we considered the monoamine networks separately (e.g. Syn-Gap-DA or Syn-Gap-TA multilink), motif 10 was overrepresented for multilink containing either serotonin and dopamine (Fig 7C and 7D, S3 Fig), but not for tyramine or octopamine (Fig 7E, S3 Fig). Interestingly, two different motifs were found to be overrepresented in the 3-layer octopamine network (Fig 7E, S11 Table), motif 9 (a unidirectional synaptic connection coincident with an octopamine connection in the opposite direction) and motif 11 (a unidirectional octopamine link coincident with a gap junction). (Presumably these were not overrepresented in the aggregate network because the octopamine network is much smaller than the networks for the other monoamines). These motifs might serve similar functions to motif 10 for dopamine and serotonin in providing feedback to modulate wired connections.
Interestingly, although the neuropeptide network showed little structural overlap with the monoamine network, its modes of interaction with the wired connectome showed striking parallels. When the neuropeptide network was included in the multiplex participation analysis, we observed that the RIM and DVA neurons continue to play central roles in linking the four network layers (S12 Table). Likewise, multilink motif analysis, this time using the neuropeptide and wired layers, again identified motif 10 (a unidirectional neuromodulatory connection coincident with a reciprocal synaptic connection) as significantly overrepresented, further supporting the notion that this motif plays a key role in extrasynaptic modulation of synaptic computation (S3 Fig; S13 Table). Even more highly overrepresented relative to expectation was motif 20, reciprocal neuropeptide and synaptic connections coincident with a gap junction. This motif was not overrepresented in the multilink analysis for monoamines, perhaps because of the low reciprocity of the monoamine network. Interestingly, several of the motif 20 multilink (S14 Table) are components of the RMG hub and spoke network, which has been implicated in the control of various behaviours including locomotion, aggregation, and pheromone response [51, 52].
This study has analysed the properties of an expanded C. elegans neuronal connectome, which incorporates newly-compiled networks of extrasynaptic monoamine and neuropeptide signalling. Analyses reveal that these extrasynaptic networks have structures distinct from the synaptic network, and from one another. The monoamine network has a highly disassortative, star-like topology, with a small number of high-degree broadcasting hubs interconnected to form a rich-club core. The monoamine systems are thus well-suited to broadly coordinate global neural and behavioural states across the connectome. Although the partial neuropeptide network we analyse is only a small sample of the complete network, it shows a different, highly clustered topology with higher reciprocity, suggesting the importance of these neuropeptide systems in the cohesion of the nervous system. While these extrasynaptic networks are separate and non-overlapping with the wired connectome, the hubs of both the wired and wireless networks are interconnected, with multilink motifs showing interaction between the systems at specific points in the network. This suggests that the extrasynaptic networks function both independently–coordinating for example through the monoamine rich-club–and in unison with the synaptic network through multilayer hubs such as RIM and through overrepresented multilink motifs.
The low degree of overlap between the monoamine and synaptic networks occurs not only because many neurons expressing monoamine receptors are not postsynaptic to aminergic neurons, but also because many postsynaptic targets of aminergic neurons to not appear to express monoamine receptors. Some of these synapses could be explained by cotransmission; in particular, tyraminergic and serotonergic neurons also express either cholinergic or glutaminergic markers, and thus classical transmitters could be used in these wired synapses. However, the dopaminergic and octopaminergic neurons of C. elegans are not known to coexpress any classical neurotransmitter. A second possibility is that these synapses could utilize synaptically-released peptides as neurotransmitters. A third possibility is that the postsynaptic cells might express either an unknown monoamine receptor, or a known one at levels too low to be detected using existing reporters. Finally, it is possible that these putative synapses, which have been identified on the basis of electron micrographs, are not really functional synapses. Further work will be necessary to resolve this puzzling question.
The importance of extrasynaptic neuromodulation to the function of neural circuits is clearly established, for example from work on crustacean stomatogastric circuits [13]. However, systematic attempts to map whole-organism connectomes have focused primarily on chemical synapses, with even gap junctions being difficult to identify using high-throughput electron microscopy approaches [53]. The incorporation of extrasynaptic neuromodulatory interactions, inferred here from gene expression data, adds a large number of new links largely non-overlapping with those of the wired connectome. Although the valence and strength of these inferred neuromodulatory links are largely unknown (information also lacking for much of the synaptic connectome), the monoamine and neuropeptide networks described here nonetheless provide a far more complete picture of potential pathways of communication between different parts of the C. elegans nervous system.
Although monoamine and neuropeptide signalling both occur extrasynaptically and act on similar timescales, the monoamine and neuropeptide networks show distinct topologies, perhaps reflecting differences in biological function. As noted previously, the monoamine network has a star-like architecture that is qualitatively different to the other network layers. This structure is reflected in the network's high disassortativity and in the low number of recurrent connections. In addition, we observed that the monoamine network contains a rich-club of highly interconnected high-degree releasing neurons, whose members are distinct from (though linked to) the rich-club of the wired connectome. Together, this structure is well-suited to the organisation of collective network dynamics, and is a useful feature given the role of monoamines in signalling physiologically important states relevant to the entire organism, such as food availability.
Despite enormous differences in scale, the monoamine systems of C. elegans and mammals share a number of common properties suggestive of common network topology. As in C. elegans, mammalian brains contain a relatively small number of monoamine-releasing neurons that project widely to diverse brain regions; for example, in humans serotonin is produced by less than 100,000 cells in the raphe nuclei, or one millionth of all brain neurons [54]. Moreover, extrasynaptic volume transmission is thought to account for much, if not most, monoamine signalling throughout the mammalian brain [55, 56]. Parallels between monoamine systems in C. elegans and larger nervous systems are not exact; for example, in C. elegans, most if not all aminergic neurons appear capable of long-distance signalling, whereas monoamines in larger nervous systems can be restricted by glial diffusion barriers [57]. Nonetheless, mammalian monoamine-releasing neurons, like their C. elegans counterparts, appear to function as high-degree broadcasting hubs with functionally and spatially diverse targets [54]. Thus, understanding how such hubs act within the context of the completely mapped wired circuitry of C. elegans, may provide useful insights into the currently unknown structures of multilayer neuronal networks in larger animals.
Although the neuropeptide network has been only partially characterized, the partial network analysed here suggests it may differ in important ways from the other connectome layers, including the monoamine network. In particular, the neuropeptide layer shows strikingly high clustering, even taking into account its high density of connections, and higher reciprocity than the monoamine network. These properties suggest the neuropeptide networks are important for cohesiveness within the nervous system. Multilink analysis also identified differences between the extrasynaptic monoamine and neuropeptide networks. In both cases, a unidirectional extrasynaptic connection coincident with a reciprocal synaptic connection (motif 10) was overrepresented in the multiplex connectome. This motif is well-suited to provide feedback between linked nodes, and occurs in several microcircuits implicated in learning and memory. For neuropeptides, a second multilink motif, involving reciprocal neuromodulatory and synaptic connections coincident with a gap junction (motif 20) was even more highly overrepresented. This motif occurs in several places in the RMG-centred hub-and-spoke circuit that plays a key role in control of aggregation and arousal. As more neuropeptide systems become characterized, it is reasonable to expect additional examples of this motif will be identified; these may likewise have important computational roles in key neural circuits.
While network theory has occasionally provided novel insights in C. elegans biology, more often the C. elegans wired connectome has provided a useful test-bed for validating new network theoretical concepts or their application to larger mammalian brains [10]. In recent years, multilayer complex systems have become an area of intense focus within network science, with a large number of papers dedicated to extending classical network metrics to the multilayer case and to developing new frameworks to understand the dynamical properties of multilayer systems [58].
By definition, multilayer networks contain much more information than simple monoplex networks, leading to significant data-collection challenges. In social networks, for example, large monoplex datasets have been collected describing various types of interactions between people, but these are typically disparate datasets based on different populations. Multiplex datasets combining various edge types into a number of layers are often restricted in size (the number of nodes for which data are collected) or in the choice of edges it is possible to consider (interaction types constrained by data availability) [58].
The multiplex connectome of C. elegans has the potential to emerge as a gold standard in the study of multilayer networks, much like the wired C. elegans connectome has for the study of simple monoplex networks over the last 15 years. The synaptic, gap junction, and monoamine layers already represent a relatively reliable and complete mapping of three distinct connection types. The lack of degree-degree correlation between some of these layers suggests that they are not just different facets of one true underlying network (such that each edge is essentially duplicated across all layers). Rather, it suggests that the wired and wireless layers provide distinct channels of communication with differing functional roles. We therefore expect wired and wireless connections to be coupled in functionally relevant 3-node and 4-node motif structures [59, 60], such as (for example) monoamine-based feedback loops or monoamine-regulated wired interactions. The different time-scales on which each of the layers operates are also likely to allow the emergence of interesting dynamical phenomena. Finally, the large number of distinct extrasynaptic interactions offers the scope for a more refined dataset, each aligned to the same complete set of 302 nodes.
How feasible is it to obtain a complete multiplex neuronal connectome? Although the neuropeptide network described here represents only a sample of the total network, the monoamine network already represents a reasonable draft of a complete monoamine connectome. Since expression patterns for amine receptors have been based on reporter coexpression with well-characterized markers, the rate of false positives (i.e. neurons falsely identified as monoamine receptor expressing) is probably very low. In contrast, the false-negative rate (monoamine receptor-expressing cells not included in the network) is almost certainly somewhat higher. In some cases (e.g. dop-4 and dop-3 in ASH [27, 61]), reporter transgenes appear to underreport full functional expression domains; in others (e.g. ser-5) only a subset of cells expressing a particular reporter have been identified [62]. With recently developed marker strains [60, 63], it should be possible to revisit cell identification and fill in at least some of these missing gaps. In addition, other monoamines (e.g. melatonin [64]) might function as neuromodulators in C. elegans, and some of the currently uncharacterized orphan receptors in the worm genome [19] might respond to monoamines. Potentially, some of these receptors might be expressed in postsynaptic targets of aminergic neurons (in particular, those of dopaminergic and octopaminergic neurons, which are not known to express classical neurotransmitters). However, the existence of additional monoamine receptor-expressing cells also means that non-synaptic edges are almost certainly undercounted in the network. Thus, the high degree of monoamine releasing hubs–and their importance for intraneuronal signalling outside the wired connectome–is if anything understated by the current findings.
In the future, it should be possible to expand the scope of the multilayer connectome to gain a more complete picture of intraneuronal functional connectivity. Obtaining extrasynaptic connectomes for larger brains, especially those of mammals, will likely be vastly more complicated than for C. elegans, due not only to the increase in size, but also the existence of additional structural and dynamical properties, such as glial barriers, cellular swelling, and arterial pulsations, all of which dynamically alter extracellular diffusion [65, 66]. In contrast, reanalysis of reporters for monoamine receptors using recently developed reference strains [60, 63] could provide a largely complete monoamine signalling network for C. elegans. A greater challenge would be to obtain a complete neuropeptide network; this would require comprehensive de-orphanization of neuropeptide GPCRs as well as expression patterns for hundreds of receptor and peptide genes. Additional layers of neuronal connectivity also remain unmapped, such as extrasynaptic signalling by insulin-like peptides, purines, and classical neurotransmitters such as acetylcholine and GABA [67–69]. Obtaining this information, while difficult, is uniquely feasible in C. elegans given the small size and precise cellular characterisation of its nervous system. Such a comprehensive multilayer connectome could serve as a prototype for understanding how different modes of signalling interact in the context of neuronal circuitry.
The synaptic and gap junction networks used in this work were based on the full hermaphrodite C. elegans connectome, containing all 302 neurons. This network was composed from the somatic connectome of White et al [2], updated and released by the Chklovskii lab [5, 70]; and the pharyngeal network of Albertson and Thomson [3], made available by the Cybernetic Caenorhabditis elegans Program (CCeP) (http://ims.dse.ibaraki.ac.jp/ccep/) [71]. The functional classifications referred to in the text (i.e. sensory neuron, interneuron, motorneuron) are based on the classification scheme used in WormAtlas [72]. The gap junction network was modelled as an undirected network with bidirectional electrical synapses; note however that some gap junctions might be rectifying and thus exhibit directionality.
To map the aminergic signalling networks of C. elegans, a literature search was first performed to identify genes known to be receptors, transporters or synthetic enzymes of monoamines. A further search was performed to collect cell-level expression data for the monoamine associated genes identified in the previous step. This search was assisted with the curated expression databases of WormBase (Version: WS248; http://www.wormbase.org/) [73] and WormWeb (Version date: 2014-11-16)[74]. A summary of these data is in Supplemental S1–S7 Tables. Neurons expressing multiple receptors for a single monoamine receive a single edge from each sending neuron. Reciprocal connections between nodes are considered as two separate unidirectional connections. Edge lists for individual network are provided in S1 Dataset.
The neuropeptide network was constructed from published expression data for peptides and receptors, using an approach similar to that used for the monoamines. Only those systems were included for which sufficient expression and ligand-receptor interaction data existed in the literature, with interactions being limited to those with biologically plausible peptide-receptor EC50 values (Supplemental S8–S10 Tables). In total, 15 neuropeptides and 12 receptors were matched and included in the network. Networks were classified by receptor, allowing a many-to-many relationship between neuropeptides and receptors.
The expression patterns of the dopamine receptors were determined using the reporter strains DA1646 lin-15B & lin-15A(n765) X; adEx1646 [lin-15(+) T02E9.3(dop-5)::GFP], BC13771 dpy-5(e907) I; sEX13771 [rCesC24A8.1(dop-6)::GFP + pCeh361], and FQ78 wzIs26 [lgc-53::gfp; lin-15(+)];lin-15B & lin-15A(n765) (kindly provided by Niels Ringstad).
The neurons expressing the receptors were identified based on the position and shape of the cell bodies and in most cases co-labelling with other markers. The reporter strains were all crossed with the cholinergic reporter [60] OH13646 pha-1(e2123) III; him-5(e1490) V; otIs544 [cho-1(fosmid)::SL2::mCherry::H2B + pha-1(+)] and the glutamatergic reporter [63]OH13645 pha-1(e2123) III; him-5(e1490) V; otIs518 [eat-4(fosmid)::SL2::mCherry::H2B + pha-1(+)] (both kindly provided by Oliver Hobert), and dye-filled with DiI using standard procedures. Strains were also crossed to AQ3072 ljEx540[cat-1::mcherry] and PT2351 him-5(e1490) V; myEx741 [pdfr-1(3kb)::NLS::RFP + unc-122::GFP], which label cells expressing the vesicular monoamine transporter and the PDFR-1 receptor, respectively. When ambiguous, reporter strains were crossed with additional strains, as listed below.
Reporter expression in individual neurons was confirmed with the following crosses:
For dop-5:
AIM and ADF were confirmed based on coexpression with cat-1. URX, PVC, RIF, RIB, AIY, M5, and DVA were identified based on position and coexpression with cho-1[60]. MI, DVC, ASE (previously identified in [75]) and ADA were confirmed based on position and coexpression with eat-4 [63]. ASI, PHA and PHB were confirmed based on costaining with DiI. PVT, RMG and BDU were identified based on cell body position and shape alone.
For dop-6:
RIH and ADF were confirmed based on coexpression with cat-1[24]. ASI and PHA were confirmed based on costaining with DiI. AQ3499 ljEx805 [sra-6::mcherry + PRF4] was used to confirm expression in PVQ. AQ3682 ljEx921[flp-8::mcherry + unc-122::gfp] was used to confirm expression in URX and AUA. IL2, RIB, RMD and URA were identified based on position and coexpression with cho-1. AVF was identified based on position and failure to coexpress eat-4 and cho-1. RID was identified based on position relative to URX and morphology.
For lgc-53:
AIM was confirmed based on coexpression of cat-1. AVF was confirmed based on coexpression with pdfr-1 and failure to coexpress eat-4 and cho-1. URY was confirmed based on position, coexpression with eat-4, and lack of coexpression with ocr-4. AQ3526 ljEx822 [klp-6::mcherry + pRF4] was used to confirm IL2 expression. AQ3535 ljEx828 [unc-4::mcherry + pRF4] was used to confirm VA expression. FLP was confirmed based on position, morphology, and coexpression with eat-4. HSN, CAN and PVD expression were identified based on position and morphology.
Strains were examined using a Zeiss Axioskop. Images were taken using a Zeiss LSM780 confocal microscope. Worms were immobilized on 3% agarose pads with 2.5mM levamisole. Image stacks were acquired with the Zen 2010 software and processed with Image J.
Edge counts, adjacency matrices and reducibility clusters were all computed using binary directed versions of the networks. The same networks, excluding self-connections (i.e. setting all diagonal elements to 0), were used to compute all other measures.
Network measures are compared to 100 null model networks (shown in the boxplots) generated using the degree-preserving edge swap procedure. This is performed by selecting a pair of edges (A→B) (C→D) and swapping them to give (A→D)(C→B). If the resulting edges already exist in the network, another pair of edges is selected instead. Each edge was swapped 10 times to ensure full randomisation. To compute the multilink motif z-scores, the null model was constructed by randomizing each layer independently.
To identify neurons with high-participation in all of the network layers, the normalized degree-rank product was used. This is computed by ranking neurons in each network layer by their degree in descending order, and scaling to the range [0, 1]. The product is then taken of the ranked degrees in each layer. Thus, if a neuron had the highest degree in each of the network layers, it would have a degree product of 1.
The measure of clustering described here is the global clustering, also known as transitivity, given in [76–78], which measures the ratio of triangles to triples (where a triple is a single node with edges running to an unordered pair of others, and a triangle is a fully-connected triple). For a directed network, this is equivalent to:
T=∑i∈Nti∑i∈N[(kiout+kiin)(kiout+kiin−1)−2∑j∈NAijAji]
where A is the adjacency matrix, N is the number of nodes, kout and kin are the out-degree and in-degree, and ti is the number of triangles around a node:
ti=12∑j,h∈N(Aij+Aji)(Aih+Ahi)(Ajh+Ahj)
To obtain the characteristic path length of a network, the geodesic (i.e. minimum) distance, d, between each pair of nodes i, j, is first computed:
dij=∑Auv∈g(i,j)Auv
where g(i,j) returns the geodesic path between nodes i and j. The characteristic path length is then given:
L=1n∑i∈N∑j∈N,i≠jdijn−1
The modularity Q is determined by first subdividing the network into non-overlapping modules c to maximise within-module connectivity and minimise between-module connectivity [79]. The modularity then gives the proportion of edges that connect to nodes within the same module:
Q=1M∑i,j∈N(Aij−kiinkjoutM)δ(ci,cj)
where ci, cj are the modules respectively containing nodes i, j; M is the number of edges, and δ is the Kronecker delta function:
δ(x,y)={1ifx=y0ifx≠y
The assortativity of a network is the correlation between the degrees of nodes on either side of a link. This is given by Newman [80] as:
R=M−1∑ij∈Ekioutkjin−[M−1∑ij∈E12(kiout+kjin)]2M−1∑ij∈E12([(kiout)2+(kjin)2])−[M−1∑ij∈E12(kiout+kjin)]2
Structural reducibility measures the uniqueness of layers by comparing the relative Von Neumann entropies. The larger the relative entropy, the more distinguishable the layer. Formally, the Von Neumann entropy for a layer is given:
H=−∑iNλi[α]log2λi[α]
where λi[α] are the eigenvalues of the Laplacian matrix associated to layer A[α]. To visualise layer similarity, hierarchical clustering was performed using the Jensen-Shannon distance [28] and the Ward hierarchical clustering method [81].
Reciprocity is the fraction of reciprocal edges in the network:
r=|E↔|M
where M is the number of edges, and |E↔| is the number of reciprocal edges:
|E↔|=∑i≠jAijAji
The rich-club phenomenon is the tendency for high-degree nodes in a network to form highly-interconnected communities [40, 82]. Such communities can be identified by creating subnetworks for each degree level k, where nodes with a degree ≤ k are removed, and computing the rich-club coefficient Φ(k) for each subnetwork. This is the ratio of remaining connections Mk to the maximum possible number of connections. For a directed network with no self-connections, where Nk is the number of remaining nodes, this is given by:
Φ(k)=MkNk(Nk−1)
Thus, a fully-connected subnetwork at a given degree k has a rich-club coefficient Φ(k) = 1. To normalise the rich-club coefficient, we computed the average values for 100 random networks ⟨Φrandom(k)⟩:
Φnorm(k)=Φ(k)⟨Φrandom(k)⟩
We used the same threshold previously used in determining the wired rich-club of C. elegans [11], defining a rich-club to exist where Φnorm(k) ≥ 1 + 1σ, where σ is the Standard Deviation of Φrandom(k).
Multilink motif analysis considers the full range of possible link combinations that can exist between any two nodes across all layers of a network, and is based on the concept of multilink as described in [47, 83, 84]. Due to the conceptual and structural similarity between monoamine layers (see reducibility), we limited our analysis to three layers: synaptic, gap junction, and monoamine (see SI for neuropeptides), giving a total of 20 possible multilink motifs. Instances of each motif were recorded by simultaneously traversing the three network layers. This was also conducted for 100 randomized three-layer networks, generated by rewiring each of the real networks individually using the same randomisation procedure described above. These random networks were used to calculate motif z-scores and p-values for the actual network.
Network measures were computed in MATLAB (v8.5, The MathWorks Inc., Natick, MA) using the Brain Connectivity Toolbox [77] and MATLAB/Octave Networks Toolbox [85]. Reducibility analysis, clustering, and multilayer plots were computed in MuxViz [86]. Reducibility is based on the algorithm described in [28], and layer similarity was visualized using the Ward hierarchical clustering method [81]. Hive plots were generated using the custom hiveplotter function written in Python (Python Software Foundation. Python Language Reference, v3.5). 3-node network motifs were computed using FANMOD [87]. Additional network visualisations were created using Cytoscape [88] and Dia (https://wiki.gnome.org/Apps/Dia/).
|
10.1371/journal.pntd.0005340 | The significant scale up and success of Transmission Assessment Surveys 'TAS' for endgame surveillance of lymphatic filariasis in Bangladesh: One step closer to the elimination goal of 2020 | Bangladesh had one of the highest burdens of lymphatic filariasis (LF) at the start of the Global Programme to Eliminate Lymphatic Filariasis (GPELF) with an estimated 70 million people at risk of infection across 34 districts. In total 19 districts required mass drug administration (MDA) to interrupt transmission, and 15 districts were considered low endemic. Since 2001, the National LF Programme has implemented MDA, reduced prevalence, and been able to scale up the WHO standard Transmission Assessment Survey (TAS) across all endemic districts as part of its endgame surveillance strategy. This paper presents TAS results, highlighting the momentous geographical reduction in risk of LF and its contribution to the global elimination target of 2020.
The TAS assessed primary school children for the presence of LF antigenaemia in each district (known as an evaluation unit—EU), using a defined critical cut-off threshold (or ‘pass’) that indicates interruption of transmission. Since 2011, a total of 59 TAS have been conducted in 26 EUs across the 19 endemic MDA districts (99,148 students tested from 1,801 schools), and 22 TAS in the 15 low endemic non-MDA districts (36,932 students tested from 663 schools). All endemic MDA districts passed TAS, except in Rangpur which required two further rounds of MDA. In total 112 students (male n = 59; female n = 53), predominately from the northern region of the country were found to be antigenaemia positive, indicating a recent or current infection. However, the distribution was geographically sparse, with only two small focal areas showing potential evidence of persistent transmission.
This is the largest scale up of TAS surveillance activities reported in any of the 73 LF endemic countries in the world. Bangladesh is now considered to have very low or no risk of LF infection after 15 years of programmatic activities, and is on track to meet elimination targets. However, it will be essential that the LF Programme continues to develop and maintain a comprehensive surveillance strategy that is integrated into the health infrastructure and ongoing programmes to ensure cost-effectiveness and sustainability.
| Lymphatic filariasis (LF) was highly endemic in Bangladesh at the start of the Global Programme to Eliminate Lymphatic Filariasis (GPELF) in 2000, with approximately 70 million people at risk. To address this burden, the National LF Programme implemented mass drug administration (MDA) in 19 highly endemic districts to interrupt transmission, and conducted sentinel site assessments in 15 low endemic districts. In 2011, as part of the LF Programme’s endgame surveillance strategy, the standard WHO Transmission Assessment Survey (TAS) was used to show the reduction below transmission thresholds in order to stop MDA in all 34 endemic districts by testing a total of 136,080 primary school-aged children in 2,464 schools using rapid diagnostic tests. The data show that LF transmission has been interrupted in all districts except one, with the latter requiring two further two rounds of MDA. Bangladesh can now be considered, after 15 years of LF programmatic activities, as having very low or no risk of LF infection and is therefore on track to meet National and Global elimination targets of 2020.
| Bangladesh is a remarkable example in terms of the progress it has made in the elimination of lymphatic filariasis (LF), following the launch of the Global Programme to Eliminate LF (GPELF) by the World Health Organization (WHO) in 2000 [1]. Bangladesh was one of the first countries in the South-East Asia Region to start the elimination process with mass drug administration (MDA) to interrupt transmission in endemic areas [2,3], and one of the first countries to begin the elimination verification process using the new WHO guidelines of the Transmission Assessment Survey (TAS) on a large scale [4,5]. Bangladesh was considered to be widely endemic for LF, caused by the parasite Wuchereria bancrofti and transmitted by Culex sp. mosquitoes [2,3,6]. An estimated 70 million people (approximately half the total population) were considered to be at risk of LF infection, with tens of thousands of people suffering from various forms of clinical presentation, including limb lymphoedema/elephantiasis and hydrocele [7,8].
Fortunately, the Directorate General of Health Services in Bangladesh recognised the immense burden of LF, and responded positively to the new GPELF initiative [1]. In 2001, the Ministry of Health and Family Welfare launched the National LF Elimination Programme with an aim to eliminate the disease as a public health problem by 2020, with MDA and morbidity management as its core components [2,3]. The decision to classify districts as endemic and implement MDA was made in the early phase of the Global and National LF Elimination Programmes, and not always straightforward. Therefore, the Bangladesh LF Programme adopted a conservative approach and used a combination of the following three parameters to inform MDA eligibility; i) microfilaria (Mf) prevalence, ii) antigenaemia (Ag) prevalence, and iii) frequency of clinical cases.
Baseline prevalence mapping and historical data indicated that the disease was endemic in 19 of the 64 districts and were considered eligible for MDA due to Ag and Mf rates of between 1% and 17% and evidence of clinical cases [9–11]. Some endemic districts, however, showed <1.0% Mf rate, but had some localised areas with chronic disease patients. Hence, it was decided to implement MDA, as a very conservative approach. Overall, three broad regions of the country required MDA, which included districts from the north (Rangpur Division) central west (Rajshahi/Khulna Divisions) and the south (Barisal Division) as shown in Fig 1A. The burden is highest in Rangpur Division, where 23% Mf prevalence and up to 10% chronic disease have been reported [3,12]. An additional 15 districts across the central and southern regions were considered to have low endemicity and not requiring MDA (Fig 1A), though some districts had considerable Ag positivity, their Mf rates were 0% and clinical cases were rare. Hence the programme decided not to implement MDA.
In 2001, the Bangladesh LF Elimination Programme began MDA in a single district using the drugs albendazole (procured from GSK through WHO), and diethylcarbamazine (DEC; procured locally using the standard government drug policy) and has since steadily scaled up to reach full geographical coverage i.e. all 19 endemic districts (Fig 2), targeting approximately 35 million people with door-to-door distribution [3,13]. Each endemic district was considered to be an implementation unit (IU) for MDA with implementation conducted in November every year using health workers and community volunteers with a ratio of one volunteer to approximately 1000 people. By 2010 the LF Programme had successfully distributed at least three rounds of MDA to all 19 endemic districts, with 12 districts receiving more than six rounds of MDA accounting for more than 150 million treatments to the 35 million target population [13,14]. Overall reported treatment rates have been high, which were confirmed by independent coverage surveys, in accordance with the WHO guidelines i.e. at least five rounds of MDA and 65% coverage of the total population in an endemic area [15]. This is considered to be sufficient to interrupt transmission, and provides the basis to commence pre-TAS and TAS activities.
Ongoing assessments through sentinel site surveillance have been crucial to measuring the impact of MDA, and determining when to stop this aspect of the programme [2]. In 2010, the LF Programme identified 10 districts that could potentially stop MDA based on Mf rates of <1%, which coincided with the development of the new TAS guidelines [15]. Bangladesh was one of the first countries to consider the new WHO TAS protocol to verify the interruption of LF transmission following MDA [13]. The Bangladesh Health Services with the support from two major international donors were able to support steadily scale up TAS following MDA in the 19 endemic districts between 2010 and 2015 (Table 1,Fig 2). Further, the LF Programme has taken the initiative to assess the 15 low endemic districts (not requiring MDA) using the TAS method following the recommendation of the WHO South East Asian Regional Programme Review Group, with activities starting in 2014 [14,16]. The TAS protocol was used as there is currently no recommended strategy for assessing low endemic districts, which remains an important issue for the GPELF to address as many countries make good progress towards the endgame [17].
Since the publication of the WHO TAS guidelines, a number of international workshops have been conducted [18], and more than one third of endemic countries have started to use them, with mixed results and implications [5,19,20]. Given that Bangladesh was one of the first countries to implement this TAS protocol and has been successful at expanding it on such a large scale, this paper presents the results of TAS across both the endemic and low endemic districts, highlighting the significant geographical reduction in risk, reasons for success and its contribution to the national, South-East Asia regional and global elimination targets of 2020 [2–4,21].
The endemic districts from the northern region include Dinajpur, Kurigram, Lalmonirhat, Nilphamari, Panchagarh, Rangpur and Thakurgaon with baseline Mf prevalence rates ranging from 4.8% to 16%, and number of MDA rounds from 5 to 12 (Table 2). The majority of these districts were the first to start MDA (before 2005) and last to start TAS (after 2011). This contrasts to the central western and southern regions where prevalence rates were lower and districts received fewer number of MDA rounds. The endemic districts from the central west include Chuadanga, Kushtia, Meherpur, Chapainawabganj, Pabna, Rajshahi and Sirajganj with baseline Mf prevalence rates ranging from 0.2% to 8.4%, and number of MDA rounds from 5 to 9. The endemic districts from the south include Barguna, Barisal, Jhalokathi, Patuakhali and Pirojpur with baseline Mf prevalence rates ranging from 1.0% to 2.2%, and five MDA rounds conducted for all districts. An annual timetable of the pre-TAS and TAS schedule is outlined in Table 1, and the TAS status in 2015 mapped in Fig 1B and baseline prevalence rates in Fig 3, highlighting the three different regions.
The 15 low endemic districts, which did not require MDA due to the overall low endemicity evident from historical data and found during Ag sentinel site mapping activities in 2002–2004 included Bagerhat, Bandarban, Bogra, Feni, Gazipur, Gopalganj, Habiganj, Jamalpur, Jhenaidaha, Laxmipur, Munshiganj, Narayanganj, Narail, and Narshingdi, and peri-urban areas of Dhaka. The TAS in these low endemic districts were conducted throughout 2014 and 2015. The distribution of these districts is shown in Fig 1B.
The TAS activities for both the 19 endemic districts and 15 low endemic districts were implemented according to the guidelines of the WHO [15]. Briefly, the survey design, number of schools and number of students to be sampled, and the critical cut-off point (i.e. the number of positive cases found that determined if an evaluation unit (EU) passed or failed the TAS) were determined using the Survey Sample Builder (SSB;http://www.ntdsupport.org/resources/transmission-assessment-survey-sample-builder), which is specifically designed for TAS.
The SSB identified a cluster survey design as the most appropriate method for Bangladesh given its very high enrolment (≥ 90%) of students in primary school. The target population included students from the 1st and 2nd grades, most of whom were 6 to 7 years of age. The SSB generated random numbers in order to select the survey schools, as well as the students within each of the selected survey schools. The inputs for the SSB included the total number of primary schools for each EU, the total number of students enrolled in the 1st and 2nd grades in EU and the expected non-response rate. The data on the list of schools and numbers of students were obtained from the Directorate of Primary Education, Dhaka.
All the schools within the district were listed and ordered geographically, and then allocated serial numbers. The schools bearing the numbers that matched the random numbers (generated by SSB) were selected for the TAS. The list of selected survey schools was prepared and forwarded to the Bangladesh Ministry of Health Services and Ministry of Education, along with a request to the latter to write to each of the selected schools asking them to consent and participate in the survey, and inform parents of the upcoming activity. The Health Services then proceeded to forward the list of the survey schools to each of the district and sub-district managers and medical officers of the local Public Health Centres to notify them of the upcoming TAS activity.
The TAS activities are part of routine LF programme activities conducted by the Ministry of Health and Family Welfare and permit the use of oral consent for all surveys. Approval from the Ministry of Education, local school authorities and Head Masters of the selected school was obtained prior to the surveys. Parents and children were verbally informed of the procedures and those who did not want to participate in the survey were excluded. Ethical clearance was obtained from the Liverpool School of Tropical Medicine Research Ethics Committee (Research Protocol 11.89R).
A sensitization and preparatory meeting was conducted about one week prior to the start of the survey. District and sub-district managers of both health and education sectors, and primary health care staff attended the meeting. Details of the objectives and procedures of the surveys were shared in this meeting, and a suitable time schedule was developed in consultation with both departments. Prior to visiting the school and starting the survey, the team visited the Public Health Centres to brief the medical officer and staff, who also took part in and supported the surveys.
Each year, field teams were selected to conduct the TAS activities, and they visited all the districts and completed all the surveys. Each team consisted of two members from the LF Programme, and two members from the local Public Health Centres. Taking into consideration the distance between the selected schools and logistics, the teams visited two to four schools per day depending on the number of teams sent from Dhaka. The two teams used the same vehicle and the two schools nearest each other were selected to make it convenient and more cost effective. On average a TAS activity took one week to complete and cost approximately US$6000 per EU.
On the day of survey, the team visited the ‘selected’ school, re-briefed the head master and other teachers of the school about the purpose and objectives of the TAS. A room within the school was then selected and set up to conduct the survey. A teacher worked with the team to organise students to be tested and to provide support to the TAS team.
All the 1st and 2nd grade students were seated in their class rooms and given serial numbers on a laminated piece of paper. The students bearing the numbers that matched the random numbers (generated by SSB), were selected and guided for blood sampling. Each student was gently explained the purpose of the blood sampling in simple language. Those students who were very fearful of blood sampling, even after explanation, were excluded from the survey. If any random number fell on a student of higher age (>7years) of the same grade the respective student was also excluded. However, the proportion of such students was negligible. In his/her place, the child with the next number was selected for sampling. Students were called one by one for finger pricking blood sampling using a disposable needle. Blood was collected into the capillary tube and then put directly on the immunochromatographic test (ICT) card to detect the presence of W. bancrofti antigen. The ICT card was manufactured by Binax NOW (Alere Inc., Scarborough, ME; http://www.alere.com/en/home/product-details/binaxnow-filariasis.html), and a positive antigen control was used to test the validity before the start of TAS. Each child was asked to rest for 5 minutes after giving the blood sample and then proceed back to the class.
For each student, the name, sex and age were recorded on the ICT card and on a data form. The ICT test and reading of the result was done strictly in accordance with the guidelines provided by the manufacturer. The results of the test were noted against each child’s name and directly on the ICT card with a permanent marker pen. ICT positive students, their families and 20 households in close proximity were treated with single dose of albendazole and DEC (diagrammatically shown in Fig 4), and provided with information on the transmission and associated risk factors of LF. As there were no formal guidelines, the decision for the LF Programme to treat 20 households was arbitrary and based on the i) high probability of transmission within the positive household and neighbouring households, and ii) tendency for the main Culex sp. mosquitoes to feed and breed in close proximity. The details of the positive student and their family were kept confidential during this localised follow-up MDA activity. The TAS data from each school were entered into an electronic database. All results were reported by the LF Programme to the local Public Health Centres, the surveyed schools and Ministry of Education.
For each endemic district a summary of the population size, baseline Mf prevalence, MDA start year, number of MDA rounds, reported and verified coverage rates and information related to the SSB calculations were collated and summarised. For each low endemic district, the population size and baseline Mf prevalence were collated. The TAS results were tabulated and the distribution of ICT positive students in endemic and low endemic districts mapped at sub-district level using geographical information software ArcGIS 10.2 (ESRI, Redlands, CA), to better understand the geographical patterns of potential residual infection or persistent transmission at a finer scale. In Bangladesh, the main sub-district administrative units are upazilas and unions. In total there are 492 upazilas (approximately 7.68 upazilas per district) and 4,501 unions (approximately 70.33 unions per district). As only about 30 schools per EU are sampled under TAS, these schools only represent a proportion of the unions, which will vary between EUs, and also limited sub-district level spatial analysis.
A summary of the TAS results for endemic districts is found in Table 2, and highlights the overall higher baseline prevalence rates and number of MDA rounds in the northern Rangpur Division. The range of reported and independently verified MDA coverage rates is also summarised and indicates overall high coverage across all districts in most years (See S1 File for details). For the first five districts to conduct TAS in 2011, the two districts of Dinajpur and Rajshahi (IUs for MDA implementation) were divided into two EUs each as the population exceeded the recommended 2 million. This resulted in seven EUs for the first TAS and these same divisions were also used for second and third TAS–known as TAS 2 and TAS 3. Similarly, for the TAS activities conducted in 2012, 2013, 2014 and 2016, the large population districts were divided into two EUs each (Sirajganj and Pabna in 2012; Barisal in 2013, and Kurigram and Rangpur in 2014 and 2016), which resulted in seven, six and six EUs being assessed in these years respectively.
Since 2011, a total of 59 TAS activities (TAS 1 in 26 EUs; TAS 1 repeated in 2 EUs; TAS 2 in 24 EUs; TAS 3 in 7 EUs) have been conducted in 26 EUs across the 19 endemic districts that had received MDA. A total of 99,148 students from 1,801 schools have been tested. The number of primary schools, and number of 1st and 2nd grade students in each EU ranged from 328 to 1,109, and from 35,137 to 93,500 respectively. Based on the SSB calculations, the number of schools selected in each EU ranged from 30 to 36, with sample sizes of 1,692 students and a critical cut-off of 20 for every EU, except for Meherpur and Jhalokathi where 1,556 students were tested and had a critical cut-off of 18 each (Table 2). All EUs passed the critical cut-off with the exception of Rangpur-B in 2014, which repeated two more rounds of MDA together with adjacent Rangpur-A (details are described below in the Rangpur District TAS overview section). In total, 112 students were found to be ICT positive across the three TAS activities (Table 3). The majority of ICT positive students were found in the highly endemic northern Rangpur Division (n = 94; 83.9%), compared with the central western Rajshahi Division (n = 13; 11.6%) and the southern Barisal Division (n = 5; 4.5%). A significant positive correlation was found between district baseline Mf measures and the number of ICT positive students found in TAS 1 (Pearson’s correlation coefficient r = 0.464; p = 0.045). There was no significant difference between the number of male (n = 59) and female (n = 53) students found to be ICT positive.
For TAS 1, there were 24 EUs from 18 endemic districts (IUs) that passed and stopped MDA. Of these EUs, a total of 40,336 students from 733 schools were tested, with 41 students found to be ICT positive (Table 2). The majority of ICT positive students were from Nilphamari (n = 9 students; 5 schools), Lalmonirhat (n = 8 students; 6 schools), Dinajpur-B (n = 6 students; 5 schools), Chapainawabganj (n = 4 students; 4 schools), Pabna-B (n = 3 students, 1 school), and Thakurgaon (n = 3 students, 3 schools). There was no difference between males (n = 20) and females (n = 21), and most were aged 6 years (n = 19) or 7 years (n = 20) with two aged 8 years. TAS 1 endemic districts requiring MDA with the related upazilas that reported ICT positive students are shown in Fig 5A. Close-up maps of the ICT positive upazilas and unions in northern region are shown in Fig 5B and Fig 5C respectively, and highlight that the majority of upazilas and unions show few or no positive students, except Rangpur and Gangachara upazilas. The union map specifically highlights the geographically sparse distribution and reduction in endemicity with only two unions reporting more than two positive cases in Rangpur.
For TAS 2, there were 24 EUs from 18 districts that passed with results below the critical cut-off. A total of 40,336 students from 733 schools were tested, with 25 students found to be ICT positive. The majority of ICT positive students were from Dinajpur-A (n = 5 students; 4 schools), Dinajpur-B (n = 5 students, 3 schools), and Nilphamari (n = 5 students; 4 schools). Of the 25 positive students, 18 were males and 7 females, and were mainly aged 6 years (n = 14), or 7 years (n = 8) with two aged 5 years and one aged 8 years. For TAS 3, there were seven EUs from five districts that passed with results below the critical cut-off. A total of 11,708 students from 215 schools were tested, with only two students found to be ICT positive, who were from Dinajpur-A (n = 1; 7 year old male student, 1 school) and Dinajpur-B (n = 1; 6 year old female student, 1 school). A map of the TAS 2 and TAS 3 districts and positive upazilas are shown in Fig 5D. Close-up maps of the ICT positive upazilas and unions in northern region are shown in Fig 5E and Fig 5F respectively, and further highlight the geographically sparse distribution and reduction in endemicity with no union reporting more than two positive students across the region.
Rangpur district had a 10% Mf baseline prevalence rate, and since 2005 had received nine rounds of MDA with coverage rates ranging from 66.3% to 95.9% and pre-TAS measures from 0.05–0.1% (Table 2). The EU Rangpur-B failed TAS 1 with 27 of 1,692 students tested from 30 schools found to be ICT positive (males = 12; females = 15). Most cases were found in Gangachara (n = 14) and Rangpur (n = 10) upazilas. The adjoining EU Rangpur-A, which found 11 ICT positive students (males = 4; female = 7), primarily in Badarganj (n = 4) and Mithapukur (n = 5) upazilas, and although passed TAS, was recommended to conduct two further rounds of MDA together with Rangpur-B. The Rangpur EUs and the upazila boundaries are shown in Fig 6A. The number of ICT positive students by upazila for this failed TAS 1 is shown in Fig 6B, highlighting the proximity of Gangachara and Rangpur upazilas. The distribution of ICT positive students by union are shown in Fig 6C, highlighting two focal locations where seven ICT positive students each were reported in Barabil and Mominpur unions, which are adjacent to the large urban areas of Rangpur upazila. Since the failed TAS 1, MDA campaigns with enhanced social mobilisation activities have been conducted in 2014 and 2015. The TAS 1 was repeated in Rangpur-A and Rangpur-B in November 2016 (1,692 students tested from 30 schools per EU), with three ICT positive students found in each EU. In Rangpur-A, ICT positive students were from Mithapukur (n = 2); and Badargani (n = 1) upazilas, and in Rangpur-B from Rangpur (n = 3) and Kaligonj (n = 1) upazilas. All positive students were from different schools.
Of the 15 low endemic districts, seven districts—Bogra, Narayanganj, Narshingdi, Gazipur, Jamalpur and Habiganj, and peri-urban Dhaka–had large populations and were divided into two EUs each. This resulted in a total of 22 EUs that were assessed in 2014 and 2015 (Table 4). The number of primary schools, and number of 1st and 2nd grade students for each of the EUs ranged from 387 to 4,071, and from 21,069 to 108,393 respectively. Based on the SSB calculations, the number of selected schools visited ranged from 30 to 33. The sample size was 1,692 (critical cut-off = 20) for all EUs, except for Bandarban where it was 1,552 (critical cut-off = 18), Narail where it was 1,556 (critical cut-off = 18), and Gazipur (A and B) where it was 1,684 (critical cut-off = 20) (Table 4).
All EUs from the low endemic districts passed the TAS critical cut-off. In total 36,932 students from 663 schools were tested across the 22 EUs, with only four students found to be ICT positive. The students testing ICT positive were from four different EUs including Bagerhat, Narail, Laxmipur and Bogra-A. Of the four ICT positive students, there was one male and three females, with one aged 6 years, one aged 10 years, and two aged 7 years.
This paper presents results from the largest scale up of post-MDA surveillance activities reported in any of the 73 LF endemic countries in the world [22]. The TAS results and the maps presented in this paper have shown a significant reduction in prevalence and geographical distribution of infection since the inception of MDA, with approximately 70 million people across Bangladesh now considered to be at very low or no risk of LF infection. This is a major step towards the national elimination goal of 2020, and key contribution to the efforts of GPELF and its specific goal to interrupt transmission [1,23]. Further, the paper presents a detailed account of the WHO TAS implementation [15], and highlights its potential use in low endemic areas, which may be applicable for other countries in certain epidemiological situations. Addressing low endemic areas or re-evaluating endemicity using a decision making prevalence survey based on probability sampling [24], is becoming increasingly important as countries need to provide nationwide evidence that LF is no longer a public health problem as part of the elimination validation and dossier requirements.
The overall success of the LF Programme in implementing MDA, reducing prevalence and scaling up TAS activities as an endgame surveillance strategy may be attributed to many key determinants of success [25,26]. Factors highlighted in the review by Kyelem et al. [25] are grouped as biological/ epidemiological/therapeutic, economic/political/social and programmatic operational effectiveness, and the most prominent factors identified that relate to Bangladesh include i) the initial level of LF endemicity ii) MDA drug regime and iii) population compliance. Bangladesh had generally low transmission levels at baseline (i.e. pre-MDA) with the majority of Mf and ICT prevalence rates <15%, further it used the drug combination of albendazole and DEC, which is considered highly effective against the W. bancrofti parasite [27,28], and the programme was able to achieve high MDA coverage rates >70% across the majority of endemic districts.
Other positive influencing factors for Bangladesh include good administrative development and health system infrastructure, relative political stability, strong political commitment and financial support, strong programme management leadership, heightened awareness of morbidity in the endemic areas, which helped to increase drug compliance and importance of disease elimination [13]. The LF Programme has also been operationally effective in many standard practices including advocacy, training drug distributors, involving community leaders in social mobilization, coordinating drug logistics with predictable MDA schedules. The LF Programme is integrated with the National Soil Transmitted Helminth (STH) Programme that implements MDA twice yearly to all primary school-children, which has also helped to increase the community knowledge and awareness of helminths across the country.
Furthermore, the Bangladesh LF programme managers have collaborated well with partnering organisations, who have been able to provide key financial and technical support. This has helped to address challenges and maintain momentum towards the elimination goal. For example, the extensive scale up of TAS activities have been achieved by two dedicated TAS teams who have worked year round with the financial support from international donors. In addition, specific technical support has been provided to develop detailed maps of prevalence data and all TAS results to highlight potential ‘hotspots’ of persistent transmission that need targeting with further investigations, interventions and surveillance. The LF Programme has also engaged in operational research to enhance the standard monitoring and evaluation activities with key findings and success stories presented internationally, including surveillance strategies and the use of TAS for low endemic areas [16,29,30].
Notwithstanding the positive factors outlined above, the LF Programme has had several challenges, particularly in the northern region where several districts received many more MDA rounds than recommended. The reason for the high number of MDA rounds was related to two main factors. First, the lack of a formal strategy with defined thresholds of ‘when to stop MDA’, meant the programme kept implementing until the TAS guidelines were released in 2010 [15]. Second, sentinel site data showed evidence of persistent infection in a number of districts up until 2013, and the recent failure of TAS 1 in Rangpur district in 2014. The reason for the persistence is not clear, but may be related to localised demographic or environmental characteristics such as poverty [31], peri-urbanisation [32], presence of Culex spp. mosquito breeding sites [6], slightly lower coverage rates, which are currently being investigated in more detail to better prepare for future surveillance. Rangpur district failing TAS was a setback for the LF programme with the elimination target being realigned, and a further two rounds of MDA implemented in November 2014 and 2015. To increase coverage during these MDA rounds, mapping of coverage rates and TAS results at sub-district level helped to identify and target problem areas, which helped the repeated TAS 1 to succeed and pass the critical cut-off in November 2016.
The TAS maps presented in this paper highlight that the persistent areas are quite focal with only a small number of upazilas and unions reporting a higher number of ICT positive students, with no specific gender or primary school identified to be at increased risk. While the distribution of positive students was found to be relatively geographically sparse across each district, an overall positive relationship between baseline Mf measures and the number of ICT positive students during the first TAS was found. This is despite the different diagnostics used in different surveyed populations at different spatial resolutions. A broad regional association was also evident with the highest number of TAS ICT positive students found in the northern Rangpur Division, which reported the highest historical and baseline Mf rates (i.e. pre-MDA)[7,12].
These results support the notion that EUs in higher risk regions, may be more likely to fail TAS and/or have areas of persistent residual infection i.e. ‘hotspots’. These and other key associations such as MDA frequency, coverage and adherence patterns as well as vector control, are being examined with modelling methods in Asia and Africa [33–37] and these may help predict hotspots. They may also help to determine appropriate diagnostic tools and surveillance strategies, with assessments of the TAS EU size also being considered as countries move closer to the elimination goal [38–41]. This is important as all districts in Bangladesh need to pass three TAS activities over a 5–6 year period, and will now be using the new rapid diagnostic Alere Filariasis Test Strip (FTS) Alere, Scarborough, ME, United States, which is more sensitive to detect Ag and as a consequence may detect more positive students [42].
The most significant activity and potential challenge for the LF programme now is to develop and maintain a sustainable long-term surveillance system that is integrated into existing health infrastructure and other ongoing programmatic activities [43]. It will be crucial that the system is tailored to detect the risk of recrudescence, which may be determined from a combination of demographic information, programmatic data such as baseline prevalence, sentinel sites, TAS results as well as the local health infrastructure capacity. For Bangladesh the northern Rangpur Division is clearly a priority surveillance area with vigilance also required for migrant populations, in international border areas and adjacent non- or low endemic districts as they are vulnerable if there is high movement of untreated people [29,30]. Ramaiah et al. [44] highlights key categories of migration which may have implications for LF elimination efforts, including i) migration from endemic areas to non-endemic areas, ii) migration from endemic rural areas to endemic urban areas, iii) migration from endemic areas to the areas that achieved control/elimination of LF, and iv) trans-border migration.
The LF Programme aims to integrate seroprevalence surveillance with its current morbidity management and disability prevent activities, which are scaling up to address the needs of approximately 40,000 people with clinical manifestations, most of whom are in the northern areas. The programme also plans to link with the STH programme as it expands to include biannual treatments to all secondary school children. A passive detection method in hospital laboratories across the country is also being assessed. Work is underway to refine this comprehensive surveillance system so that the programme is able to readily respond to any potential hotspots detected in the future, and also to provide evidence that Bangladesh has successfully addressed the essential national, South-East regional and GPELF requirements and eliminated LF as a public health problem [2–4,21,29].
|
10.1371/journal.pcbi.1000870 | Calcium Signals Driven by Single Channel Noise | Usually, the occurrence of random cell behavior is appointed to small copy numbers of molecules involved in the stochastic process. Recently, we demonstrated for a variety of cell types that intracellular Ca2+ oscillations are sequences of random spikes despite the involvement of many molecules in spike generation. This randomness arises from the stochastic state transitions of individual Ca2+ release channels and does not average out due to the existence of steep concentration gradients. The system is hierarchical due to the structural levels channel - channel cluster - cell and a corresponding strength of coupling. Concentration gradients introduce microdomains which couple channels of a cluster strongly. But they couple clusters only weakly; too weak to establish deterministic behavior on cell level. Here, we present a multi-scale modelling concept for stochastic hierarchical systems. It simulates active molecules individually as Markov chains and their coupling by deterministic diffusion. Thus, we are able to follow the consequences of random single molecule state changes up to the signal on cell level. To demonstrate the potential of the method, we simulate a variety of experiments. Comparisons of simulated and experimental data of spontaneous oscillations in astrocytes emphasize the role of spatial concentration gradients in Ca2+ signalling. Analysis of extensive simulations indicates that frequency encoding described by the relation between average and standard deviation of interspike intervals is surprisingly robust. This robustness is a property of the random spiking mechanism and not a result of control.
| The number of proteins organizing cellular processes is huge. The challenge for systems biology is to connect the properties of all these proteins to cellular behavior. Do individual state changes of molecules matter for cell behavior despite these large numbers? Recently, we have experimentally shown for four cell types that intracellular Ca2+ signalling is driven by single channel dynamics. Molecular fluctuations are used constructively for a stochastic spike generation mechanism. The hierarchical structure of Ca2+ signalling prevents averaging of fluctuations and, consequently, the sequence of global spikes still reflects this molecular noise. Here we present a stochastic 3-D multiscale modelling tool living up to these findings by following the consequences of individual channel state changes up to cell level. We simulate the variety of cell responses in different experiments. The stochastic spike generation mechanism is surprisingly robust, providing new insights into the relation of function and robustness. The modelling concept can be applied to a large class of reaction-diffusion processes including other pathways like cAMP.
| Cellular behavior is the dynamics emerging out of molecular properties and molecular interactions. Hence, cells are indispensably subject to intrinsic noise due to the randomness of diffusion and molecule state transitions in gene expression [1], [2], signaling pathways and control mechanisms. It drives noise induced cell differentiation [3], cell-to-cell variability of cloned cells [4] or second messenger dynamics [5]. While noise in gene expression can be attributed to small molecule numbers, we consider here noise in signalling pathways which occurs even in systems with large molecule numbers.
Molecular interactions create nonlinear feedback like substrate depletion and allosteric regulation in enzyme kinetics or mutual activation of ion channels in membrane potential dynamics. They also couple active molecules inside cells spatially by diffusion of product and substrate or electric currents. If this coupling is strong enough, cells respond spatially homogeneous. Otherwise, we observe dynamic spatial structures formed by concentrations of molecules in specific states. These structures are often called microdomains [6]–[9].
The existence of these dynamic structures determines in some systems whether the cell obeys deterministic or stochastic mechanisms. The dynamic compartmentalization of the cell by concentration gradients may prevent the establishment of deterministic dynamics by the law of large numbers even if the total number of molecules in the cell would suggest it otherwise. Microdomains are too small to behave deterministically. Not even the whole ensemble of microdomains will behave deterministically, if they are only weakly coupled or if there are only a few of them. Consequently, noise is not averaged out on cell level.
To determine whether we deal with a deterministic or stochastic system is important since these regimes may exhibit very different dependencies of behavior on system parameters [10]. For instance, repetitive spiking in intracellular signalling would be restricted to parameter values providing oscillatory dynamics with a deterministic mechanism [11], [12]. It may occur with a stochastic system also for parameters which would lead to bistable or excitable dynamics in the deterministic limit, i.e. for larger or different parameter ranges [13]. In the non-oscillatory parameter ranges, the mechanism creating almost regular spike sequences can be coherence resonance [14]–[16] rather than the existence of a limit cycle in phase space of the local dynamics. Noisy systems with gradients usually show also a dependency of system characteristics on parameters of spatial coupling which spatially homogeneous systems do not exhibit. An example is the dependency of the spiking frequency on diffusion properties (see below and [5]).
In summary, the interaction between noise and gradients determines parameter dependencies and mechanisms. Recent experimental and theoretical studies on intracellular dynamics taught us that cells may indeed work in this regime and may exhibit repetitive spiking with non-oscillatory local dynamics. Functionally relevant gradients are also observed with intracellular cAMP [8], [17]–[19], pH [20] and in phosphorylation/dephosphorylation dynamics [21], [22] suggesting that the lessons learned from dynamics may also apply to other systems.
One of these lessons is that the randomness of single molecule state changes is carried up from the molecular level to cell level [23], [24]. Cellular concentration spikes form random sequences of interspike intervals (ISIs) and that randomness arises from the randomness of single molecule state transitions [5], [25]. Consequently, the fluctuations of cellular signals contain information on single molecule behavior. It is a task for modelling now to establish the relation between these fluctuations and single molecule properties to decode this information.
Systems exhibiting the interaction between noise and gradients require modelling tools which can deal efficiently with the large concentration gradients and with the time scale range from molecular transitions to cell behavior. Here, we present such a modelling concept with the example of intracellular dynamics. It simulates all active molecules as stochastic Markov chains with all the individual state transitions and describes diffusion and some bulk reactions deterministically. Active molecules are those carrying the crucial feedbacks and nonlinearities. That allows for linearization of passive bulk reactions and the application of a multi-component Green's function to solve the partial differential equations in the cell analytically. We combine Green's functions with a local quasi-static approximation for the fast concentration changes and diffusion processes at the location of active molecules. That is possible due to the short diffusion time on the molecular length scale of a few nanometers. Since we use Green's functions for the long range concentration profiles we can restrict the calculation of concentration values to the location of active molecules. That renders this method extremely efficient even in 3 spatial dimensions.
We will apply this concept to intracellular dynamics and compare simulated time dependent concentrations with single cell time series obtained from cultured astrocytes all measured under the same condition without any stimulation. is a ubiquitous second messenger in eukaryotic cells that transmits a variety of extracellular signals to intracellular targets. controls fertilization, cell differentiation, gene expression, learning and memory [26]. It triggers secretion in glands, muscle contractions in the heart and transmits apoptosis signals [27], [28].
A main mechanism to increase the cytosolic concentration is release from intracellular stores, especially from the sarcoplasmic reticulum by ryanodine receptor channels (RyRs) or the endoplasmic reticulum (ER) by inositol 1,4,5-trisphosphate receptor channels (). These channels open in a dependent fashion - a self amplifying effect known as induced release (CICR) [27], [29]. If a single channel opens, is released into the cytosol, diffuses to adjacent channels and increases their open probability. Thus release may spread into the entire cell leading to a global cytosolic concentration spike.
The inositol 1,4,5-trisphosphate () pathway initiates release from the ER in many cell types (including astrocytes [30]), since binding of to the primes them for activation by (Figure 1 in Text S1). The spatial arrangement of in channel clusters leads to a hierarchical system with the structural levels channel, channel cluster and cluster array, which is the cell level. pumps and buffers generate large gradients close to open channel clusters. Thus, channels within a cluster are strongly coupled and the coupling between clusters is only weak - the geometrical hierarchy entails a hierarchy of coupling strengths.
Stochastic binding of and to the binding sites of leads to random opening of a single channel in a cluster [31], [32]. This causes other channels of the same cluster to open also leading to a puff. An individual cluster is stochastic due to the small number of per cluster [33]–[35]. The opening of a single cluster can only be detected by adjacent clusters due to the strong gradients [23], [24], [27], [36], [37]. Since they are again only a few, it remains random whether they are opened by the initial puff. If a supercritical number of puffs arises, release spreads into the whole cell causing a global spike. Thus, due to the hierarchy of coupling strength, randomness is carried up from the channel level to the cell level.
In order to model the hierarchical system, we have to consider the stochastic behavior of individual and the spatial heterogeneity of cells induced by clustering. That leads to a reaction diffusion system (RDS) with local stochastic source terms. For sufficient fast simulations, we decompose the system into local stochastic dynamics comprising channel state transitions and fast local concentration changes and a deterministic global dynamics for which we derive an analytical solution in form of a three component Green's function (Text S1). The solution is driven by stochastic channel behavior described by a hybrid deterministic-stochastic algorithm. We apply the model to a variety of experiments to demonstrate its potential.
Our modelling concept simulates active molecules individually by Markov chains, the concentration dynamics in the range of the molecule locally quasi-statically and the diffusional long range coupling by Green's functions. Simulations are orders of magnitude faster than numerical schemes based on spatial grids. Their efficiency derives from the methods which we apply. The use of hybrid deterministic-stochastic algorithms for the Markov chains allows for time steps much larger than traditional Gillespie algorithms. In between stochastic molecule state transitions, we integrate the concentration dynamics. The local quasi-static approximation reduces clusters to spatial -function sources which turns integrals into sums. It also substantially reduces the number of modes to be used in the Green's function. And finally Green's function enables us to restrict the calculation of concentration values to the locations of active molecules.
dynamics and spatial channel clustering lead to the hierarchical system depicted in Figure 1. channels are tetrameres [38]. A single channel opens and closes in dependence on binding and dissociation of and to the binding sites of its subunits (see below). An open channel conducts a current from the ER into the cytosol which is due to the huge concentration difference of up to 4 orders of magnitude across the ER membrane.
form clusters on the membrane of the ER consisting of 1 to 10 channels [33], [35]. They physically interact within a cluster and are consequently separated by a few nanometers only [35]. The in a cluster are strongly coupled by the large local concentration close to open channels.
Typical inter-cluster distances found experimentally are in the range of 1–7 [39]. Figure 1A shows a representative example of cluster arrangement used in simulations. Due to cytosolic buffers and SERCAs, the local concentrations close to an open channel cluster exhibit large gradients such that coupling between clusters is weak compared to intra-cluster coupling. This leads to the hierarchical organization of signals. Stochastic opening of a single channel (blip) is locally amplified by CICR leading to a puff (Figure 1B and D). The concentration gradients keep the probability for activation of adjacent clusters small and only a fraction of puffs activates several neighboring clusters. Once a supercritical number of open clusters is reached, more of them open forming a global signal. In that way, the triggering random opening of a single is carried up to the macroscopic scale. The mechanism transforms the fast noise of channel state changes on a millisecond time scale into fluctuations of interspike intervals of tens of seconds as shown in Figure 1D.
An early and widely used channel state model is the DeYoung-Keizer model [40], [41]. It assumes independent subunit dynamics and allocates three binding sites to each subunit as shown in Figure 1C. One site for and one for that cooperatively activate the subunit. Another binding site with lower affinity for inhibits the subunit dominantly. These two different affinities lead to a biphasic dependence of the stationary open probability on the concentration (see Figure 1 in Text S1). Only the state out of the 8 possible subunit states corresponds to an active subunit (Figure 1C), where the first index refers to the binding and is 1, if is bound and 0 otherwise. Analogously, the second and third index describe binding to the activating and inhibiting site, respectively. A channel opens, if at least 3 subunits are in the active state.
The 12 possible transitions between the 8 subunit states correspond to transitions in a state scheme forming a cube (Figure 1C). Some of the transition probabilities depend on the local and concentrations (Figure 1 in Text S1). In simulations, the transitions are realized by a hybrid deterministic-stochastic algorithm [42], which uses the local concentrations and the dissociation rates and binding rate constants given in Table 1 in Text S1.
Since within one cluster are close to each other, a cluster can be approximated by one spatial -source for the purpose of simulating the cluster current in the long range cellular dynamics. The current depends on the number of open channels , the time course of which comes out of the stochastic simulation of channel states. It is proportional to the concentration difference across the ER membrane at the location of the channel molecule. Hence, we actually need to solve the complete reaction-diffusion problem to determine it. But the concentration difference at the cluster is not well defined with a -source term. Therefore, we calculate the cluster current using a spatially extended cluster with radius as described in detail in Ref. [43]. The solution of that problem converges within fractions of a millisecond to its stationary state in the range of the channel molecule [43]. That part of the solution is all we need to calculate the current of the th cluster. Using the stationary concentration profiles we obtain:(1)with where denotes the channel flux constant. and are the diffusion coefficients of in the ER and the cytosol. The cluster radius depends on the number of open channels and the single channel radius . The advantage of the approximation is that it takes local ER depletion into account but only depends on the the spatially averaged concentrations and , which form the boundary conditions for the local quasi-static approximation (see [43] for details). If channel distances within a cluster are of the order of magnitude of the diffusion length of free , the internal cluster geometry becomes relevant. In that case, several -functions can be used for one cluster.
The approximation allows as well for determination of the local concentration at an open channel cluster resulting from its own current (1) as(2)the validity of which had been shown for the buffer concentrations used here [43]. Note that the total concentration at a cluster is the sum of the concentration (2) and the concentrations induced by currents of other open channel clusters. After closing, the concentration is determined by the cellular concentration dynamics (see below) 10 nm apart from the release site.
The modelling strategy for the cellular dynamics is based on the separation of two length scales. On the microscopic scale of channel clusters, we use a detailed and stochastic channel model to determine local currents. On the macroscopic scale of the cell, we use a linearized spatial bi-domain model, and Green's function to integrate it. The microscopic scale determines the currents representing the sources of the macroscopic scale. We implement ideas proposed in [43] and use the currents of Eq. (1) as the amplitudes of the spatial -functions representing the cluster source terms in Eqs. (3). A similar approach was taken by Solovey et al. [44]. We circumvent the concentration divergence at -function sources by using Eq. (2) for the value of the local concentration at open clusters. Vice versa, the macroscopic scale affects the concentration values entering the transition rates of the microscopic state schemes.
The ER is a tubular network spreading throughout the cell [45]. Diffusion in such a geometry can be described by diffusion in unrestricted space with a decreased diffusion coefficient [46]. Subsequently, we can superimpose the ER and the cytosol leading to a bi-domain model. Due to the quasi-static approximation (Eq. 1), we do not need to determine the spatially resolved concentration in the ER. Lumenal and cytosolic domains are coupled by a homogeneous leak flux through the ER membrane, re-uptake of the ER by SERCA pumps and by the stochastic channel currents . Within the cytosol we take free , one mobile buffer and one immobile buffer with the total concentrations and into account leading to the reaction diffusion equations(3a)(3b)(3c)where we used buffer conservation and linear pump and leak fluxes with the flux constants and . is the stochastic channel cluster current of the th cluster with strength defined by Equation (1).
Scaling concentrations, space and time with typical values reveals the number of independent parameters. It entails the definitions of Table 2. We linearize Eqs. (3), since we would like to use Green's function to solve them. Our parameter values are in the range of the applicability of the linearization to the buffer dynamics as described by Smith et al. [47] for the stationary profiles. We additionally have linearized the pump dynamics. The linearization does not exhibit saturation, which is relevant for calcium concentrations above , with being the dissociation constant of the pump (Figure 2 in Text S1). These concentrations occur close to open clusters. In that area, the dynamics are dominated by the diffusion term and the channel term, which reduces the relative error due to the linearization of pump and buffer rates substantially. However, if precise knowledge of concentration values close to open channels or clusters is required, the complete non-linear reaction diffusion equations must be solved like e.g. in [42]. The scaled linear reaction diffusion system (Text S1) describes the spatially resolved concentration dynamics by:(4a)(4b)(4c)where the leak flux depends on the average lumenal concentration, only. All the reaction rate constants depend on the resting state concentration , and due to the linearization: , and . For simplicity we subsumed also and under .
The cytosolic concentrations are determined by the 3-component Green's function with clusters localized at (see also Figure 3 in Text S1)(5)with the Bessel function of the first kind and the Legendre polynomial , where is the angle between the source location and the point given by(6)
The are determined by the boundary conditions at the plasma membrane (see Text S1).
The three-component response functions and include the time integration over the source history, i.e. the time dependent channel flux strength , and take the buffer reactions as well as the coupling with the ER into account:(7a)(7b)with the dimensionless cell radius and the normalization factors and given in the Text S1. The coupling between the cytosol and the ER by and as well as the reaction rates of with the two buffers determine the time constants of the response functions (0), which are implicitly given by the roots of the determinant of the coupling matrix(8)
The method allows for spatially resolved concentration dynamics as shown in Figure 2 and in the Video S1 by an iso-concentration surface of 2 . An initially opening cluster increases the open probability of adjacent clusters and release is spreading through the cell until inhibition stops release.
For the global dynamics, the average concentrations are obtained by spatial integration of the analytical solution (9) as(9)where denotes the cell radius. The first component of describes the cytosolic average concentration . With this, the lumenal average concentration in dimensionless units is determined by(10)which takes into account the leak, pump and channel fluxes, and is the volume ratio of the cytosol and the ER. denotes the equilibrium average lumenal concentration at . The difference between the average cytosolic and lumenal concentration − determines the cluster current according to Eq. (1) (see Text S1).
The two main approximations of our method are the local quasi-static approximation and the linearization of the passive bulk processes. These assumptions do not allow for a precise study of the intra-cluster concentration dynamics. That can be done with finite element methods like in ref. [42]. The structure of the Green's function solution enables an elegant parallel algorithm that we call the Green's cell. It is orders of magnitude faster than finite element methods and able to simulate long lasting whole cell dynamics in feasible computing time. In the Green's cell algorithm the actual concentration of each cluster is calculated with the Green's function and local quasi-static approximation in dependence on the source history of all clusters by a single process. The concentrations are sent to the master process which determines the corresponding state transition and reaction time by the hybrid algorithm and also calculates the average concentrations. The transition times are re-distributed to the cluster processes where they are used to update the concentrations. For further details see Figure 4 in Text S1.
Our recent experimental investigation started from the assumption of a random spike generation by wave nucleation followed by a deterministic refractory time. This prediction yields in a linear dependence of the standard deviation on the average period which was also experimentally confirmed [5]. Previous studies report a possible feedback of on PKC activity in glutamate stimulated rat astrocytes [48]–[50]. This may lead to a positive feedback on the level by activation of PLC. The measured relation between standard deviation and average of interspike intervals for spontaneous spiking has a slope equal to 1 [5], demonstrating that spike generation is poissonian and the spike generation probability is constant on the time scale of ISI. Clearly, there is no feedback on that time scale.
To show that the experimental findings are indeed consistent with our ideas of spike generation, we use our modelling tool to study how molecular noise of single channels can be translated into global signals and whether it is sufficient to cause the observed randomness of spike sequences. Figure 3A shows an example of single cell measurements, where the upper panel exhibits the fluorescent signal related to the cytosolic concentration and the lower panel the individual ISIs. It demonstrates the stochasticity of spiking, since variations in ISIs are in the range of their average. Simulations of a cell with 47 clusters each containing a random number of between 4 and 16 exhibit a behavior very similar to experiments showing that single channel noise can lead to time varying ISIs, since there are not any other sources of noise in the simulations (Figure 3B and C). The simulated oscillations exhibit in accordance with experimental observations different flavors ranging from rare and irregular spiking to fast and more periodic spiking. The standard deviation depends linearly on the average period [5]. Recently we have shown that this linear dependence is not a self-evident relation [51]. In particular, it was found that self-sustained oscillatory systems exhibit a different relation than the one observed in spiking experiments. The dependence of on obtained here in simulations is shown in Figure 3D and exhibits a linear dependence with a slope of 1 which was found in experiments for spontaneous oscillations [5], [52]. The offset of the regression line on the -axis of about 20 s is the deterministic recovery time.
The different − data points in Figure 3D result from different combinations of the and base level concentrations, which are both parameters in the model. In vivo the concentration is related to the stimulation level by activation of Phospholipase C and production. The base level is determined by the leak and the pump flux through the ER membrane. In simulations, we adjust the leak flux according to and the pump strength. If both concentrations are rather high in the range of no spiking occurs since channels are activated as soon as they are in the excitable state (Figure 5 in Text S1). We observe fast and regular spiking (Figure 3C,E and F and Figure 5 in Text S1) for intermediate concentrations. The ISIs have a close to the deterministic refractory time, since a new spike is initiated as soon as the recovery time has elapsed. Regular spiking corresponds to cells with small in Figure 3D. A further decrease in one of the concentrations increases and , in a way depending on the other concentration (Figure 3B,E and F). If both concentrations are small, global spiking vanishes and the signal consists of uncorrelated blips.
In the previous analysis of the dependence of oscillations on the concentrations, we have already seen that the modelling tool can generate a large spectrum of signals ranging from stochastic spiking to almost periodic oscillations. Here, we show that the model can produce all known -induced forms of signals in dependence on physiologic parameters. Figure 4 exhibits different experimental signal forms and the corresponding simulation results for a cell with 32 clusters. The variety of signals is achieved by varying cell parameters leading to distinct cell responses as shown by the behavior of open channels (black) and number of inhibited subunits (magenta) as well as by the resulting average concentration in the cytosol (red) and in the ER (blue). Fast and rather regular oscillations occur by the interplay of activation and inhibition leading to array enhanced coherence resonance as was hypothesized before [5]. This can be seen in the behavior of the channel state dynamics. The number of inhibited subunits (magenta) increases dramatically during a spike and finally inhibition terminates it (Figure 4A). In the following the number of inhibited subunits relaxes slowly towards its resting level. Only very few channels open directly after a spike and these openings do not initiate a new spike, since the number of inhibited subunits is still to high (higher than approximately 220). That causes the deterministic time also observed experimentally [5], [52]. But a spike occurs very soon after the number of inhibited subunits has fallen below a critical range since the open probability is rather high with these parameter values. That keeps the stochastic part of the ISI small and spike sequences regular. Moreover, the amplitude of the spike of open channels seems to be smaller, if the spike is initiated at times where the number of inhibited subunits is still high.
We find longer and more irregular ISIs for decreased and base level concentrations, since the probability of a channel opening is decreased. As a consequence, the cell relaxes to a resting state between spikes with only a few inhibited subunits (Figure 4B). The spike amplitudes of both the number of open channels and of the average concentration are slightly increased compared to the regular spiking.
SERCA pumps also shape signals. Recent studies have shown that different phenotypes of cloned cells with regard to signalling occur due to small variations in SERCA expression levels and activity of RyR [4]. Here, we find that a decreased SERCA activity leads to a burst like behavior (Figure 4C), since is removed slower from the cytosol and thus can activate channels which have recovered early from inhibition or channels which have not been activated before.
For even smaller SERCA activity, cells exhibit long lasting plateau signals often with superimposed oscillations (Figure 4D). In these cases, released stays within the cytosol and reactivates again and again. Cooperativeness induced by inhibition leads to superimposed oscillations on the high level. The panels of Fig. 4 provide also an idea of cell variability within one cell type and even within one experiment.
A direct consequence of the diffusion mediated signal mechanism is the dependence on the strength of spatial coupling by diffusion. That coupling strength can be modulated by exogenous buffers, since they reduce the diffusion length of free . We took advantage of this property of buffers to demonstrate the spatial character of oscillations [5]. Note that we used concentrations of buffers much smaller than usually applied in order to suppress any kind of signal. We measured spiking for several minutes to obtain reference values for ISIs, loaded additional buffer and continued measuring (see Figure 5A). The individual ISIs (blue crosses) are increased and exhibit a larger variability after buffer loading.
To understand the experimental observation in more detail, we use simulations to analyze the response to additional buffer. Analogously to the experiment, we simulate a fixed cellular arrangement with different mobile buffer concentrations. Figure 5B shows a representative example, where the red and the blue parts correspond to 25 and 250 EGTA, respectively. Like in the experiment, larger buffer concentration leads to less and more irregular spiking. In the part with the higher buffer concentration, we observe isolated events which do not lead to global waves since coupling of clusters is too weak. These local events are rare in the reference measurements, since a triggering event initiates a global wave very likely.
From population simulations, where individual cells differ in their spatial arrangement of clusters, initial buffer and base level concentrations, we obtain the − relation shown in Figure 5C, where cells are shifted by an increase of 10 in the EGTA concentration. Similar to experiment [52], cells exhibit individual increases of and with a slope of the shift close to 1 comparable with the population slopes for the two measuring periods.
BAPTA and EGTA are common buffers to suppress signals and we have used both in experiments [5]. Cells responded much more sensitive to BAPTA than to EGTA. BAPTA has much larger binding and dissociation rate constants than EGTA (Table 1). A disadvantage of the experiment is that the buffer is loaded into a cell by its esterificated form and the total amount that has entered is unknown and difficult to control. Here, we use modelling to illuminate the influence of the different buffer kinetics and concentrations of EGTA and BAPTA.
Figure 5D shows the dependence of for fixed cell parameters on the buffer concentration in magenta for EGTA and in black for BAPTA, where squares denote simulations with a single channel current of 0.12 pA and the dots correspond to 1.2 pA. The larger current was achieved by an increased lumenal concentration. Cells only differ in the buffer type. We see that increasing BAPTA has a stronger effect than EGTA, which is mainly caused by the larger capture rate. Moreover, we observe a nonlinear dependence of on the buffer concentration. The nonlinearity explains the individual shifts of cells in the − plane shown Figure 5D. The comparison of the two different current strengths for BAPTA (black) indicates the role of spatial coupling. Higher currents lead to stronger coupling, and subsequently increasing buffer concentrations have a smaller effect on .
From the buffer simulations, we can determine the − relation shown in Figure 5E. For the smaller currents, there is no qualitative difference between EGTA and BAPTA. Both exhibit a slope close to 1 as shown by the regression lines and an estimated deterministic time of 20 s. The simulations with higher cluster currents indicate a similar deterministic refractory period but the slope of the − relation decreases to approximately 0.6. This might explain the experimentally found differences between cell types. Larger currents lead to stronger coupling on the macroscopic length scale and hence to smaller variations.
To confirm these findings and to test the dependency of the slope on other parameters, we analyze spiking of cells for the two different single channel currents. In each simulation set the cells have identical properties and differ only with respect to the buffer content leading to the distinct and values in Figure 5E (see also Section 6 in Text S1). From these values we determine the population slopes . Figure 5F shows averaged over different spatial arrangements, concentrations (stimulation levels) and pump strengths (see Figure 6 in Text S1). Analogously, we investigated , and (data not shown). The results are very similar to those with . For smaller single channel current we obtain always a slope close to 1 when varying all 4 cell properties and for the larger current a slope to 0.6. Varying the buffer concentration, spatial arrangement of clusters, concentration or pump strength (within certain limits) does not change the − relation but only the position of the system on it.
We have presented here an efficient modelling concept for dynamics in 3 spatial dimensions. It simulates cell behavior starting from individual channels in full detail. Using Green's function and multiscale techniques allow for taking concentration gradients into account and thus for capturing the hierarchy of coupling strengths. The method can simulate up to 4000 seconds real time within 24 h on 8 CPUs for a cell with 32 clusters and 10 channels per cluster. In comparison to grid-based numerical methods, its main advantage is a gain of computational speed of several orders of magnitude, which enables us to simulate whole spike sequences. We demonstrate the potential of this modelling concept by simulating a variety of experiments. We compare the in silico data with time series obtained from spontaneous oscillations in cultured astrocytes, but several of the results will also apply to other cell types like those analyzed in [5].
These recent experiments showed for 4 different cell types that the sequences of interspike intervals in signalling are random [5]. In line with the ideas on the signalling mechanisms, we assumed single molecule state transitions to be a sufficient source of noise. We confirm this assumptions with our simulations here in which these state transitions are the only source of randomness. The fluctuations are carried up through the structural levels due to the existence of concentration gradients and hierarchies of coupling strength.
With our bottom-up modelling approach, we were able to generate all experimentally known signal types in dependence on physiologic parameters. Spiking exhibits the random ISI sequences observed experimentally with fast regular sequences and slow irregular ones. In particular, the dependency on parameters of spatial coupling observed in experiments is reproduced. We find a sigmoidal response of the concentration upon very strong stimulation or strong spatial coupling, which is well known as over stimulation. We observe also bursting. We do not compare our bursting simulations with specific experiments here, but we would like to mention a general aspect. This signal type is usually ascribed to the existence of a dynamic feedback like store depletion or inhibition of production which terminates bursts. Such a feedback is not required with a stochastic model. The random length of bursts in our stochastic model offers also a simple explanation for the irregular burst length observed in experiments.
With our method we are able to follow the dynamics from the molecular to the cellular level. The single molecule fluctuations determine the global behavior, since they initiate cellular signals. Simultaneously, the local rough channel signal is smoothed on the cell level by the hierarchical system due to diffusion. The universality and variety of signalling cross talks between signalling and other pathways render a potential source of noise in cellular systems. The fluctuations can be used for cell variability [4] with regards to gene regulation [53], [54] and cell differentiation [3] and provides a flexibility to changing external conditions which is needed during evolution [55].
Both the experiments and simulations show a simple linear relation between the standard deviation of ISI and the average ISI . The existence of this linear relation turned out to be surprisingly robust. It survives even an increase of the single channel current by an order of magnitude. This relation describes for each individual cell the response to stimulation changes. Cells shift the spike pattern from slow and irregular to fast and regular along the − relation when we increase stimulation. That supplements the current ideas on frequency encoding [54], [56].
At the same time, the − relation describes the outcome of spiking experiments with a group of cells. In the experiments, we subjected a sample of cells to the same protocol, and we obtained as many different responses as there are cells in the sample [5]. That set of responses is not arbitrarily scattered across the − -plane but aligns along the − relation. All the variability among individual cells with respect to expression levels of pathway components, cell volume, ER volume, shape, ion concentration, etc. does not lead to severe deviations from this − relation. spiking is robust against variability of many pathway components in the sense that the − relation is robust. We learn from the simulations here, that it is rather the stochastic spike generation mechanism than control and regulation which provides for this robustness.
If we call the − relation from a single cell obtained by parameter changes individual relation and that obtained from a sample of cells population relation, we can describe our findings as identity of individual and population relation.
We could reproduce the variability within a population of cells in simulations by varying cluster array geometry, pump strength, stimulation or buffering conditions. Changing these parameter values simply shifted the system on the − relation and did not modify the relation. But changing the single channel current by one order of magnitude did change the slope of the − relation.
That suggests a mathematical definition of robustness which accounts for the fact that cells should be able to execute certain functions (e.g. to spike with a range of ISI), but not necessarily at the same strength of stimulation or normalized values of other parameters. We denote with and two variables describing the function (e.g. and ), and with ,…, and ,…, two sets of parameters (e.g. stimulation strength, temperature, cell volume). The relation between and is robust with respect to value changes of parameters , if it has the structure . The parameters change only the value of while the control also the properties of , i.e. the properties of the pathway. We call this robustness of the function functional robustness (in difference to the robustness of the value of ). If we identify the stimulation strength with , all cells distinguished by the values of only can realize frequency encoding with the same − relation by varying . They can realize this function also by varying another -parameter or several of them: function and functional robustness are closely related.
The statement on robustness can also be interpreted with respect to identity of pathways converging onto spiking. signals can be caused by many different stimuli. The pathways upstream from responding to the stimuli must differ with respect to their value of the , in order to be distinguishable by pathways downstream from .
In summary, cells realize frequency encoding - the function of spiking - by mainly moving up and down the relation between standard deviation and average of ISI and to some degree by modulating the deterministic part of the ISI [52]. The − relation exists for a stochastic process only, since holds for deterministic systems. The − relation turned out to be functionally robust with respect to changes of values of one set of parameters. That set may describe cell variability within one cell type or pathway. Changing substantially another set of parameters modified the − relation. That set appears rather to specify the identity of pathways converging on spiking.
Our model predicts that close proximity of clusters is a prerequisite for a spontaneous response to spread throughout a cell. Indeed, there are types of astrocytes in which responses spread within the cell and those, such as Bergmann glia where this is not observed. Interestingly local, subcellular spontaneous responses have been recorded which represent functional microdomains [57]. Complementary to the functional units, morphological units are described which are separated from each other by fine processes [58]. It is well conceivable that these thin processes separate endoplasmic reticulum between microdomains by more than 2 and according to our model this separation would prevent the spread of a local signal to other parts of the cell. In contrast, in cultured astrocytes, the endoplasmic reticulum is preferentially arranged around the cell center without apparent discontinuity [59] and these cells frequently exhibit spontaneous responses. In situ, spontaneous responses are reported for hippocampal astrocytes and these astrocytes are less polarized as compared to Bergmann glial cells and we would predict that they are less compartimentalized. Indeed, morphological studies indicate that hippocampal astrocytes have five to ten main processes from which smaller extensions branch off [60]. The synchronized activity obviously can spread within the volume of the main processes and soma of hippocampal astrocytes. Moreover, in contrast to culture, the endoplasmic reticulum in astrocytes in hippocampus tissue is preferentially located close to the plasma membrane [59]. These different morphological arrangements result in distinct patterns of responses and as a consequence in different gene expression patterns [53].
The rise of cell imaging during the last decades illustrated the spatial structure of cells and protein localization. Obviously, cells are not homogeneous and active molecules coupled by diffusional transport are very common. Concentration gradients are functionally relevant, if they create microdomains inside which a pathway is in a state different from its state at other locations in the cell. They have been shown to exist for ‘the other’ fast diffusing intracellular messenger cAMP and in phosphorylation/dephosphorylation dynamics.
Hence, the need for spatially resolved cell models exists and we can apply the modelling concept, if all essential non-linearities are in the discrete active molecules or the boundary conditions and we can linearize remaining bulk reactions. The excellent validity of the linearization for the buffer reactions of dynamics has been shown by Smith et al. [47]. We expect a degradation reaction like the cAMP degradation by PDEs also to be linearizable in good approximation. If local concentrations at active molecules should be outside the range of validity of the linearization, that can be fixed by the choice of the local quasi-static approximation of the diffusion process there in many cases. The non-linearities of cAMP production by membrane-bound adenylyl cyclase can be formulated as boundary condition and Green's function must then be used iteratively with an update of the boundary condition in each time step. These remarks illustrate that there is flexibility in the choice of reactions to be linearized which crucially expands the applicability of the concept.
Astrocyte cell cultures were prepared from cortex of newborn NMRI mice [61]. Briefly, brain tissue was freed from blood vessels and meninges, trypsinised and gently triturated with a fire-polished pipette in the presence of 0.05% DNAase (Worthington Biochem. Corp., Freehold, NY, USA). Cells were washed twice and plated directly on poly-L-lysine (PLL; 100 ; Sigma, Deisenhofen, Germany) coated glass coverslips () at densities of 3 to cells/coverslip, kept in -10-cm-dishes using Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (FCS), 2 mM L-glutamine, 100 units/ml penicillin, and 100 streptomycin. One day later, cultures were washed twice with Hank's balanced salt solution (HBSS).
Cells were maintained for at least 4 days and after reaching a subconfluent state, microglial cells and oligodendrocytes as well as their early precursors were dislodged by manual shaking and removed by washing with HBSS. The purity of the astrocytes was routinely determined by immunofluorescence using an antibody against glial fibrillary acidic protein (GFAP, Sigma), a specific astrocytic marker. The cultures typically exhibited more than 90% cells positive for GFAP.
Cultured cells plated on glass coverslips were measured between p4 and p6. Cells were incubated with the indicator dye Fluo-4-acetoxymethyl-ester (Fluo-4 AM, 5 , Molecular Probes, Eugene, USA) for 30 min at room temperature in HEPES buffer (148.9 mM NaCl, 5.4 mM KCl, 1 mM , 10 mM , 10 mM HEPES, 5 mM D-glucose, pH 7.4) containing 0.01% Pluronic-127 (Molecular Probes). Subsequently cells were washed and kept in HEPES buffer for 15–20 min prior to the measurements with the conventional imaging system at a frequency of 0.33 Hz. Cultures were fixed within the microscope chamber of an upright microscope (Axioskop FS, Zeiss, Oberkochen, Germany) equipped with a 20× water immersion objective (UMPlanFl, numeric aperture: 0.5, Olympus, Hamburg, Germany) by a U-shaped platinum wire and superfused with HEPES buffer at 20. Substances were applied by changing the perfusate. Cells were illuminated (495 nm) from a monochromator (T.I.L.L. Photonics) and fluorescent images (515–545 nm) collected every 3 s with a 12 bit camera (SensiCam) on an upright microscope. At this state, no intercellular waves were observed. Single cell time series were extracted from these images with ImagingCellsEasily software.
|
10.1371/journal.ppat.1000850 | Emergence and Pathogenicity of Highly Virulent Cryptococcus gattii Genotypes in the Northwest United States | Cryptococcus gattii causes life-threatening disease in otherwise healthy hosts and to a lesser extent in immunocompromised hosts. The highest incidence for this disease is on Vancouver Island, Canada, where an outbreak is expanding into neighboring regions including mainland British Columbia and the United States. This outbreak is caused predominantly by C. gattii molecular type VGII, specifically VGIIa/major. In addition, a novel genotype, VGIIc, has emerged in Oregon and is now a major source of illness in the region. Through molecular epidemiology and population analysis of MLST and VNTR markers, we show that the VGIIc group is clonal and hypothesize it arose recently. The VGIIa/IIc outbreak lineages are sexually fertile and studies support ongoing recombination in the global VGII population. This illustrates two hallmarks of emerging outbreaks: high clonality and the emergence of novel genotypes via recombination. In macrophage and murine infections, the novel VGIIc genotype and VGIIa/major isolates from the United States are highly virulent compared to similar non-outbreak VGIIa/major-related isolates. Combined MLST-VNTR analysis distinguishes clonal expansion of the VGIIa/major outbreak genotype from related but distinguishable less-virulent genotypes isolated from other geographic regions. Our evidence documents emerging hypervirulent genotypes in the United States that may expand further and provides insight into the possible molecular and geographic origins of the outbreak.
| Emerging and reemerging infectious diseases are increasing worldwide and represent a major public health concern. One class of emerging human and animal diseases is caused by fungi. In this study, we examine the expansion on an outbreak of a fungus, Cryptococcus gattii, in the Pacific Northwest of the United States. This fungus has been considered a tropical fungus, but emerged to cause an outbreak in the temperate climes of Vancouver Island in 1999 that is now causing disease in humans and animals in the United States. In this study we applied a method of sequence bar-coding to determine how the isolates causing disease are related to those on Vancouver Island and elsewhere globally. We also expand on the discovery of a new pathogenic strain recently identified only in Oregon and show that it is highly virulent in immune cell and whole animal virulence experiments. These studies extend our understanding of how diseases emerge in new climates and how they adapt to these regions to cause disease. Our findings suggest further expansion into neighboring regions is likely to occur and aim to increase disease awareness in the region.
| Newly emerging and reemerging diseases have become a major focus of infectious disease research in the 21st century. Reemerging diseases are classified as those that have been previously documented, but are now rapidly increasing in incidence, geographic range, or both [1]. Emerging disease events have been occurring at higher than average rates in the United States due to several factors such as wildlife diversity, environmental change, international travel, and increases in host susceptibility [2], [3]. An additional factor contributing to increases in morbidity and mortality for many infectious diseases involves genetic recombination events or gene/pathogenicity island acquisitions. These events can occur via either horizontal gene transfer or conjugation/introgression, leading to novel pathogenic genotypes. This form of virulence evolution has been well characterized in bacterial, viral, fungal, and parasitic human diseases [4], [5], [6], [7], [8], [9]. The ability to cause damage to mammalian hosts is a common theme among all microbial pathogens, making it a key aspect of host-pathogen studies [10].
In the genomic era, it is now possible to combine conventional epidemiological approaches with newly developed molecular typing techniques to gain insight into the emergence and molecular epidemiology of pathogens. These approaches can improve understanding of population dynamics during an outbreak, and may lead to novel methods for the rapid identification, treatment, and diagnosis of emerging infections [11]. In addition, molecular typing serves as an initial approach to classify isolates into distinct genotypes for analysis. Further investigations may include the examination of virulence and phenotypic traits that may be common or distinct between genotypes [6], [12], [13]. Gaining insights into the molecular epidemiology and virulence of newly emerging diseases has considerable potential for the rapid assessment and management of newly emerging infections.
Over the past decade, Cryptococcus gattii has emerged as a primary pathogen in northwestern North America, including both Canada and the United States [6], [13], [14], [15], [16], [17], [18]. In the past, C. gattii has often been associated with Eucalyptus trees in tropical and subtropical climates, causing disease in immunocompetent hosts at low incidences [19], [20], [21]. C. gattii is distinct from its sibling species Cryptococcus neoformans [22], which more commonly infects immunosuppressed hosts and infects almost one million people annually with over 620,000 attributable mortalities [23], [24], [25]. C. gattii can be classified into four discrete molecular types (VGI-VGIV), which represent cryptic species as no nuclear allelic exchange between groups has been observed [6]. This molecular classification is significant because VGII is responsible for approximately 95% of the Pacific Northwest infections in Canada and the United States [12], [15]. The appearance of C. gattii in North America is alarming because this is the first major emergence in a temperate climate, indicating a possible expansion in the endemic ecology of this pathogen [26], [27].
Several significant questions persist regarding the outbreak and its expansion within the United States. As the global collection of C. gattii isolates expands, the molecular epidemiology of the species has become increasingly informative, particularly through multilocus sequence typing (MLST), which allows data to be readily compared between groups within the research community [6], [15], [28], [29], [30]. The increase in global and regional isolates that have been typed at the molecular level allows detailed analysis of C. gattii. The analysis of both conserved coding regions, and diverse noncoding regions provides insight into the genotypes responsible for the outbreak. A major finding in this study is a level of underlying diversity within the VGIIa/major genotype in the region of expansion and other geographic locales.
Prior studies documented that the C. gattii VGIIa/major genotype isolates from Vancouver Island are highly virulent in experimental murine infection assays [6]. Here we expanded this analysis to examine clinical VGIIa genotype isolates from Vancouver Island, the United States, and Brazil, in addition to an environmental VGIIa isolate from California. Our findings are consistent with recent macrophage intracellular proliferation studies, demonstrating that United States isolates from the recent Pacific NW outbreak exhibit high virulence [31]. The enhanced virulence of isolates from the outbreak region, when compared with those from other regions, suggests that the genotypes circulating in the Pacific NW are inherently increased in their predilection to cause disease in mammalian hosts.
In addition to the detailed examination of the VGIIa/major genotype clade, we report that the novel VGIIc genotype is highly virulent in a murine inhalation model. Moreover, the VGIIc genotype was found to have high intracellular proliferation rates in macrophages and a significantly increased percentage of mitochondria with tubular morphology after macrophage exposure, and thus VGIIc isolates share virulence attributes with the VGIIa/major genotype isolates from the Vancouver Island outbreak. These results extend the molecular and phenotypic understanding of the recently discovered VGIIc/novel genotype and help shed light into its possible geographic and molecular origins.
These studies provide insights into both the evolutionary history and virulence characteristics of this unique and increasingly fatal fungal outbreak in the temperate climate of the North American Pacific Northwest and highlight the importance of a collaborative interdisciplinary approach to the analysis of emerging pathogens. Application of these approaches may increase awareness of disease risks in the expansion zone, lead to more rapid diagnoses and, as a result, accelerate the implementation of appropriate therapy.
Human and veterinary cases of confirmed or suspected C. gattii infections in the states of Washington and Oregon were identified by referring physicians and veterinarians, and subsequently isolates were purified and examined. Melanin production was assayed by growth and dark pigmentation on Staib's niger seed medium, and urease activity was detected by growth and alkaline pH change on Christensen's agar. These tests established that isolates were Cryptococcus (C. neoformans or C. gattii). Isolates were concomitantly examined for resistance to canavanine and utilization of glycine on L-canavanine, glycine, 2-bromothymol blue (CGB) agar. Growth on CGB agar indicates that isolates are canavanine resistant, and able to use glycine as a sole carbon source, triggering a bromothymol blue color reaction indicative of C. gattii, whereas C. neoformans is sensitive to canavanine, and cannot use glycine as a sole carbon source, resulting in no growth or coloration in this selective indicator medium. All CGB positive isolates were then grown under rich culture conditions prior to storage at −80°C in 25% glycerol and genomic DNA extraction. For genomic DNA isolation, a modified protocol of the MasterPure Yeast DNA purification kit from Epicentre Biotechnologies was used. Briefly, 500 µl of glass beads (425–600 nm) were added into the combination of cells and 300 µl cell lysis solution. The rest of the method followed the protocol provided by the manufacturer.
For multilocus sequence typing analysis (MLST) [32], each isolate was analyzed with a minimum of eight and in some cases sixteen loci. For each isolate, genomic regions were PCR amplified (Table S1), purified (ExoSAP-IT), and sequenced. All primers used for the analysis were designed specifically to amplify open reading frame (ORF) gene sequence regions including those with non-coding DNA regions to maximize discriminatory power. Sequences from both forward and reverse strands were assembled, and manually edited using Sequencher version 4.8 (Gene Codes Corporations). Based on BLAST analysis of the GenBank database (NCBI), each allele was assigned a corresponding number. GenBank accession numbers with corresponding allele numbers are listed in the supplementary information (Table S2). To determine that the nine VGIIc/novel isolates are clonally related, given the level of diversity in the loci and the number of isolates that have been examined, we applied an equation to measure the probability of a genotype occurring more than once in the dataset [33], [34]. For the variable number of tandem repeat (VNTR) analysis, the Tandem Repeat Finder (TRF) version 4.00 software package was employed for marker development, using the genomic sequence of C. gattii isolate R265 (http://www.broadinstitute.org/annotation/genome/cryptococcus_neoformans_b.2/Home.html) [35]. The identified tandem repeat sequences and 400 bp of the flanking region were extracted from the genomic sequence and ranked according to the number of total repeats and the size of repeat units using an in-house Perl script (available upon request). Markers were examined for stability and those with high variability and stability were chosen for the analysis. Sequences were assembled and edited using Sequencher version 4.8 (Gene Codes Corporations) and aligned using the Clustal W web based software package (http://www.ebi.ac.uk/Tools/clustalw2/index.html).
Mating analysis was conducted on V8 media (pH 5). Isolates were incubated at room temperature in the dark for 2–4 weeks in dry conditions. All strains were crossed with the VGIII mating type a isolate B4546 and the VGIII mating type α isolate NIH312, both of which are fertile and commonly used for mating studies [36]. Fertility was assessed by microscopic examination for hyphae, fused clamp cells, basidia, and basidiospore formation.
For each VNTR marker, a sequence type was defined as a sequence exhibiting a unique mutation. Each sequence type was confirmed to be unique by BLAST analysis of the NCBI GenBank database [37]. A concatenated VNTR sequence type (CVST) was defined as unique combinations of sequence types from the VNTR markers. A multiple alignment of the sequences was carried out using Clustal W software [38]. Analysis of the sequences was conducted using the Neighbor-Joining and Maximum Parsimony methods within the MEGA 3.1 software [39]. In addition, the use of the maximum likelihood method (PhyML 3.0) with SH-like approximate likelihood-ratio test and HKY85 substitution model was applied [40], [41]. For this purpose, sequences of the selected VNTR markers were concatenated. We additionally concatenated all of the strain-typing markers including the housekeeping genes used in MLST and VNTR loci for clustering analysis. The haplotype mapping analysis was carried out using TCS software version 1.21 (http://darwin.uvigo.es/software/tcs.html) [42].
A proliferation assay was previously developed to monitor the intracellular proliferation rate (IPR) of individual strains for a 64-hour period following phagocytosis [31]. For this assay, J774 macrophage cells were exposed to cryptococcal cells that were opsonized with 18B7 antibody for 2 hr as described previously [43]. Each well was washed with phosphate-buffered saline (PBS) in quadruplicate to remove as many extracellular yeast cells as possible and 1 ml of fresh serum-free DMEM was then added. For time point T = 0, the 1 ml of DMEM was discarded and 200 µl of sterile dH2O was added into wells to lyse macrophage cells. After 30 minutes, the intracellular yeast were released and collected. Another 200 µl dH2O was added to each well to collect the remaining yeast cells. The intracellular yeast were then mixed with Trypan Blue at a 1∶1 ratio and the live yeast cells were counted. For the subsequent five time points (T = 16 hrs, T = 24 hrs, T = 40 hrs, T = 48 hrs and T = 64 hrs), intracellular cryptococcal cells were collected and independently counted with a hemocytometer. For each strain tested, the time course was repeated at least three independent times, using different batches of macrophages. The IPR value was calculated by dividing the maximum intracellular yeast number by the initial intracellular yeast number at T = 0. We confirmed that Trypan Blue stains 100% of the cryptococcal cells in a heat-killed culture, but only approximately 5% of cells from a standard overnight culture. Compared to a conventional colony counting method, this method was shown to be more sensitive in detecting the clustered yeast population or yeast cells undergoing budding. IPR values were used to assess how consistent the different VGII genotype subgroups were. For this statistical analysis the medians of each population were compared with the non-parametric Mann-Whitney U-test and values of p<0.025, after controlling for multiplicity, and were accepted as statistically significant (http://elegans.swmed.edu/~leon/stats/utest.cgi).
The mitochondrial morphology assays were conducted in a similar way to those in previous studies, with modifications [31]. C. gattii cells, grown overnight at 37°C in DMEM in a 5% CO2 incubator without shaking for 24 hr, or isolated from macrophages 24 hr after infection, were harvested, washed with PBS twice and re-suspended in PBS containing the Mito-Tracker Red CMXRos (Invitrogen) at a final concentration of 20 nM. Cells were incubated for 15 min at 37°C. After staining, cells were washed in triplicate and re-suspended in PBS. For each condition, more than 100 yeast cells per replicate for each of the tested strains were chosen randomly and analyzed. For quantifying different mitochondrial morphologies, images were collected using a Zeiss Axiovert 135 TV microscope with a 100× oil immersion Plan-Neofluar objective. Both fluorescence images and phase contrast images were collected simultaneously. Images were captured with identical settings on a QIcam Fast 1394 camera using the QCapture Pro51 version 5.1.1 software. All Images were processed identically in ImageJ and mitochondrial morphologies were analyzed and counted blindly.
Three individual experiments were performed for each condition and the data were tested for normality using the Shapiro-Wilk test. For homogeneity of variances we used the Levene statistic. For statistically significant differences among the mean data we applied a One-Way ANOVA. Multi-comparisons using Tukey Honestly Significant Differences tests were performed to identify statistically significant differences between pairs. A p-value of p<0.05, after controlling for multiplicity, was considered to be statistically significant. Regression analysis was used to measure the correlation between tubular mitochondrial morphology and IPR values; an F-value of P<0.05 was considered to be a significant correlation.
To examine the virulence potential of global VGII isolates, with a specific emphasis on the Pacific NW VGII outbreak genotypes, two independent murine virulence experiments were conducted at two facilities (Duke University Medical Center and the Wadsworth Center). The murine virulence assays at Duke University Medical Center and the Wadsworth Center used a similar protocol to previous C. gattii and C. neoformans experimental infections [6], [44], [45].
At the Duke University Medical Center Animal Facility, virulence was assessed using female A/Jcr mice (NCI, 18–24 g). Strains were cultured in YPD broth for 18–20 h at 30°C, harvested, washed three times with sterile PBS and counted using a hemocytometer to determine cell concentrations. Inocula for both murine experiments were confirmed by plating on YPD and counting colony-forming units (c.f.u.). Nine to ten A/Jcr mice per strain were anesthetized with pentobarbital and infected via intranasal instillation with 5×104 c.f.u. in 50 µl of sterile 1× PBS. Animals that displayed severe morbidity, based on twice-daily examinations, were euthanized. Time to mortality was evaluated for statistical significance using Kaplan–Meier survival curves within the Prism software package (GraphPad Software), and P values were obtained from a log-rank test. Survival data was plotted for graphical analysis using the Prism software package.
At the Wadsworth center animal facility, all assays were conducted using male BALB/c mice (approximately 6 weeks old, 15–20 g, Charles River Laboratories, Inc.). Strains were grown overnight in YPD broth at 30°C with shaking. The cells were harvested, washed in PBS, and counted using a hemocytometer. Five mice per strain were anesthetized with a mixture of xylazine–ketamine, and allowed to inhale 105 (30 µl) cryptococcal cells per mouse, via intranasal instillation. Mice were given food and water ad libitum and monitored twice daily. At the first sign of poor health or discomfort, infected animals were euthanized. Brain and lung tissues from the dead animals were cultured on Niger seed agar for C. gattii recovery to confirm infections were due to this pathogen. Time to mortality was evaluated for statistical significance as described above.
Two animals from each strain assayed in the study conducted at Duke University were selected for histopathology analysis either at the time of sacrifice or at the conclusion of the experiment for the more attenuated isolates. For each animal, lung samples were collected and stored in 10% neutral buffered formalin. Samples were paraffin embedded and hematoxylin and eosin (H&E) stained at the Duke University Research Histology Laboratory. After staining and slide preparation, each sample was examined microscopically for analysis of cryptococcal cell burden and immune responses. Images were captured using an Olympus Vanox microscope (Duke PhotoPath, Duke University Medical Center).
The animal studies conducted at the Wadsworth Center were in full compliance with all of the guidelines set forth by the Wadsworth Center Institutional Animal Care and Use Committee (IACUC) and in full compliance with the United States Animal Welfare Act (Public Law 98–198). The Wadsworth Center IACUC approved all of the vertebrate studies. The studies were conducted in facilities accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC).
The animal studies at Duke University Medical Center were in full compliance with all of the guidelines of the Duke University Medical Center Institutional Animal Care and Use Committee (IACUC) and in full compliance with the United States Animal Welfare Act (Public Law 98–198). The Duke University Medical Center IACUC approved all of the vertebrate studies. The studies were conducted in Division of Laboratory Animal Resources (DLAR) facilities that are accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC).
To examine the C. gattii outbreak isolates collected from 2005 to 2009 (Figure 1), an in-depth stepwise molecular analysis was applied to each isolate, and the genotypes were compared with other global genotypes. In total, 20 markers were selected for analysis. These markers include both coding and noncoding genomic regions and range in size and allelic diversity (Table 1). Additionally, all of the markers are randomly distributed among the chromosomes in the most recent assembly of the reference C. gattii VGI genome, WM276 (Figure 2). Initially, all isolates were sequenced at a total of eight MLST markers, and four variable number of tandem repeats (VNTR) markers (Figure 3, Table 2). Next, global isolates were selected for diversity, and several isolates from each of the primary genotypes in the expansion region were chosen for sequence analysis at eight additional MLST loci, bringing the total number of genetic markers analyzed for these isolates to 20 (Figure 4A). As expected, the MLST markers were less variable and more conserved, while the VNTR markers allowed for higher-resolution differentiation between isolates that appeared identical by MLST analysis. The generated datasets were then concatenated both without and with VNTR data (Figure 4B, Figure 4C).
The combined analysis of the results presented here, and a 30 marker MLST analysis conducted previously [6], [18], reveal several findings of interest in relation to VGII genotypes in the region. From the analysis of 34 markers (30 MLST/4 VNTR), we show that the Vancouver Island VGIIa/major isolates are fully identical at all loci to several recent isolates from Washington and Oregon, as well as a historical clinical isolate (1970's), NIH444, from Seattle. Additionally, the VGIIb/minor isolates from Australia and Vancouver Island are identical at 34 total loci, and also identical to VGIIb/minor isolates from Oregon at 20 loci (16 MLST/4 VNTR). Furthermore, all VGIIc isolates to date are identical across all 20 loci examined (Figure 4A). However, we also are able to discriminate the outbreak VGIIa genotype from an environmental VGIIa isolate from California, CBS7750, and clinical VGIIa isolates CA1014 and ICB107 from California and Brazil, respectively, at one or more MLST/VNTR loci. It is clear from prior studies that the VGIIa/major and VGIIb/minor isolates are clonal lineages [6], [12], [15], [46], and here we confirmed that this is the case for the nine VGIIc/novel isolates, based on 7-loci MLST analysis of the global VGII population (Figure S1) (p<0.0001).
The largest and most comprehensive dataset arose from the combined analysis of seven MLST and four VNTR loci, resulting in a total of 41 sequence types (STs). This dataset was generated from clinical, veterinary, and environmental C. gattii isolates (Figure 3, Figure S1, Table S3). From the analysis, it is clear that the VGIIa/b/c clusters are all related to each other, but also distinct. In addition, the data show that the VGIIa/major clade is closely clustered to VGIIc, further validating prior reports that examined a more limited number of loci [13], [47]. In addition, VGIIc (ST21) shares high sequence identity to ST34, represented by a mating type a clinical isolate from Colombia, suggesting that the VGIIc genotype may have resulted from a-α mating, even though all isolates related to the Pacific NW outbreak are exclusively α mating type. Additionally, Vancouver Island isolates from our collection that had not been fully typed by MLST were sequenced at two loci to determine if any were unrecognized VGIIc isolates (n = 56) (Figure S2). Of these, 51 were found to be VGIIa, five were VGIIb, and none were VGIIc, consistent with previous data from the region. Thus, VGIIc appears to remain exclusive to the United States, specifically Oregon, and has never been reported from Vancouver Island, the mainland of Canada, Washington State, or elsewhere globally.
Within the VGIIa/major cluster, based on the initial MLST analysis of 30 loci, only a single isolate (ICB107) could be distinguished from the other VGIIa isolates, and this was at only one locus [18]. To further investigate this homogeneous population causing the vast majority of the outbreak-related morbidity and mortality, we expanded the molecular analysis to include highly variable regions of the genome. The application of these VNTR markers, in combination with the MLST markers, allowed us to generate five independent STs from within the VGIIa/major genotype and related isolates (Figure 3).
These five sequence types (ST1, ST2, ST3, ST13, ST30) contained a total of 44 isolates (Figure 3, Table S3). The canonical VGIIa/major outbreak genotype, ST1, contained the vast majority of the 44 isolates (n = 38). As expected based on previous models of the C. gattii outbreak expansion [13], ST1 consisted of isolates exclusively from the initial outbreak and expansion zones, including British Columbia, Washington, and Oregon (Table S3). These results further validate the hypothesis that the epicenter of the outbreak was on Vancouver Island, beginning in the late 1990's, with a direct expansion into neighboring mainland British Columbia and subsequently into the United States [13]. The only exception in this dataset is isolate NIH444, an older isolate from the region that was isolated from a patient sputum sample in Seattle in the early 1970's [18], which is also identical at all 34 markers examined. This suggests that the VGIIa/major genotype responsible for most of the outbreak cases may have been circulating in the region prior to the outbreak. The possible travel history of this patient is unknown, and could therefore have involved exposure on Vancouver Island. Overall, this analysis provides increased evidence that the outbreak genotype is unique to the region thus far, and molecularly distinct from closely related isolates from both California and South America.
While the homogeneous nature of the VGIIa/major isolates based on robust molecular typing validated previous models, an underlying diversity within this group was also discovered. First, we further validated that the isolate ICB107 (ST13), from Brazil, was indeed distinct from the ST1 VGIIa/major clade. This isolate differs at one MLST marker (LAC1), and three VNTR markers (VNTR3, VNTR15, VNTR34). Additionally, the high-resolution sequence analysis was able to discriminate other VGIIa isolates that were collected from California. These include isolate CBS7750 (ST3), collected from the environment in San Francisco in 1990 [48], and isolate CA1014 (ST2), which was isolated from a patient with HIV infection in southern California. Each of these two isolates differs from ST1 due to unique mutations within the VNTR7 and VNTR34 loci, respectively. This shows that similar VGIIa genotype isolates have been found elsewhere, but that none are identical to those circulating as part of the ongoing Vancouver Island outbreak. Whether these isolates are a result of drift from ST1, or if ST1 arose from one of these related genotypes is not known.
In addition to discriminating VGIIa isolates that were not from the outbreak region, we also found a novel ST, ST30, which is highly similar to ST1, but divergent at a unique region of VNTR34. Interestingly, all three of the ST30 isolates are exclusively from Oregon, including two human clinical cases and one marine mammal case (Figure 1, Figure 3, Table S3). These results are consistent with an expansion followed by genetic drift in the highly variable VNTR loci. Isolates of ST30 have not been detected on Vancouver Island, indicating that this divergence is recent, and likely occurred after the expansion of ST1 into the United States. Alternatively, both ST1 (VGIIa/major) and ST30 may have been present for a long period, with only ST1 having been transferred to Vancouver Island.
To gain insights into the potential origins of the VGIIc genotype, and to assess its position within the overall VGII clade, clustering analysis was applied. Analysis of the combined dataset including 41 sequence types generated from 115 C. gattii isolates shows that the VGIIc genotype is independent, but similar to VGIIa (Figure 3). The closest relationship determined from the analysis was to ST34, an isolate from Colombia, which is also of the opposite a mating type. Moving beyond the direct branch, it appears that the VGIIc genotype shares sequence similarities to global isolates from South America, Africa, and also European isolates with likely African origins based on collected clinical case histories. Additionally, the VGIIc group also shares the IGS1 allele with isolates from Australia, further obscuring the possible origins and necessitating a more thorough analysis (Figure 4A).
When the clustering analysis was expanded to include additional MLST loci (Figure 4A), both with and without the VNTR markers, the relationships of VGIIc to other global genotypes was further elucidated, with close relationships observed with global isolates from South America, Africa, Europe (Greece), and Australia (Figure 4B, Figure 4C, Table S4). These results increase the comprehensiveness of the analysis, and allow predictions of the relationship of this genotype to global isolates. Examination of alleles illustrates that, when the analysis is expanded, the VGIIc group appears to be more diverse from VGIIa and VGIIb. Each allele represented in green was initially denoted as an allele that was unique to the VGIIc genotype, with a total of seven such alleles (Figure 4A). To further elucidate the possible origins of these alleles, isolates selected based on their global diversity were sequenced at these loci (Figure 4A). Identical matches for four of the seven VGIIc-unique alleles were identified in isolates from Brazil, Australia, Europe, and European isolates with likely African origins, while three alleles (SXI1α, HOG1, and CRG1) remain unique to this novel genotype and only seen in Oregon thus far (Figure 4A).
To further characterize the genetic relationships among the global isolates in relation to the outbreak isolates, maximum likelihood (ML) analysis was applied. Initially, the isolates were characterized at 15 MLST loci, excluding the MAT locus so that both α and a isolates could be included. This analysis indicates that VGIIc may be more distantly related to the VGIIa/major genotype than initially observed. In addition, analysis of the 15 MLST loci shows a possible relation of VGIIc with isolates from South America, Africa, Europe, and Australia (Figure 4B). When this analysis was expanded to also include the four VNTR loci, similar results for the global comparisons of all genotypes and the relation of VGIIc to global isolates were observed (Figure 4C). For these reasons, additional sampling and analysis will be necessary to more precisely elucidate if this novel virulent genotype originated locally, or originated in an under-sampled region.
In addition to clustering analyses, TCS haplotype-mapping software was applied to establish the evolutionary histories of the MLST alleles examined during the analysis (Figure 5, Figure 6, Figure S3). From the sequence results, all of the VGIIc isolates were determined to be 100% identical, indicating that there was likely a recent emergence in which all of the isolates are clonally derived. To test this hypothesis, the TCS analysis allowed for the examination of individual loci to determine which alleles are likely ancestral, intermediate, or recently derived. Of the sixteen loci examined, eight were consistent with VGIIc possessing the ancestral allele, six of the alleles were distal nodes at the terminal end of the respective haplotype networks, and two loci were of intermediate allele positions.
Alleles with ancestral genotypes are less informative because these alleles may not have diversified over time in the VGIIc lineage for various reasons, including selection pressures and overall lack of diversity at the allele. When only non-ancestral alleles were examined, 75% lay at the distal ends of their haplotype maps. Intriguingly, the three VGIIc alleles unique to the genotype (SXI1α, HOG1, and CRG1) all have distal placements (Figure 5A–C). Additionally, the most recent ancestor to VGIIc in all three cases can be shown to derive from isolates that are from South America and Australia, indicating that VGIIc may have emerged out of one of these regions (Figure 5). While other regions including Europe and North America can be seen, no other regions are observed for all three of these alleles. These distal placements are consistent with a recent divergence of the unique VGIIc lineage. The haplotype analysis, in combination with the lack of any underlying diversity within the nine VGIIc isolates analyzed, indicates a recent emergence of this novel virulent genotype in Oregon.
To examine the role that recombination may have played in the population structure of the VGII molecular type, we conducted paired allele analysis for 25 representative global isolates (Figure 6, Figure S4). The discovery of all four possible allele combinations between two unlinked loci (AB, ab, Ab, aB) serves as evidence for likely recombination [49]. From this analysis, we show that isolates collected from South America, Africa, and Australia appear to be involved in recombination events. Representative VGIIa/major, VGIIb/minor, and VGIIc/novel isolates were found among groups of recombinant isolates. A group of ten isolates, all α, from South America and Africa (Figure S4) appeared most commonly as recombinant partners, although several a mating type isolates were also less frequently involved. In further support, when we examined the number of genotypes present by region and compared this data to the total number of genotypes represented (Figure S1), it is clear that South America and Africa populations are more diverse when compared with isolates from North America, which are more clonal. Additionally, while the observed diversity in Australia was lower than South America and Africa, this may be attributable to sampling bias of clonal regions as prior studies have shown that this continent is a region with high levels of recombination due to both same-sex and opposite-sex mating events [50]. In addition to the paired allele analysis, allele diagrams were constructed to observe possible recombination within individual MLST loci (Figure S5). The most parsimonious explanation for allelic diversity in 11 of the MLST loci analyzed is as a result of consecutive and/or independent mutations within the population. Within the four remaining loci, there exists at least one hybrid allele that may be the result of a recombination event between two hypothesized parental alleles in the global VGII population (Table 3, Figure S5). Phenotypic mating results were conducted and illustrate that the VGIIa/major (α), VGIIc/novel (α), VGII mating type a genotypes, as well as several of the proposed parental contributors from the allelic and genotypic recombination analysis show fertility with the production of spores when mated with fertile VGIII isolates (Table S5). Taken together, this suggests that both α-α and a-α mating events may be contributing to the formation of recombinant genotypes as well as the production of infectious spores. There were no examples of alleles introgressed into VGII from VGI, VGIII, or VGIV, in accord with findings that the four VG molecular types likely represent cryptic species [6], [29]. In summary, these results suggest that recombination events may be critical driving forces in the evolution of C. gattii VGII diversity, which may in part contribute to the generation of genotypes displaying increased virulence.
It has recently been shown that intracellular proliferation rate (IPR) values for cryptococcal cells within macrophages are positively correlated with virulence in the murine model for cryptococcosis [31]. To further elucidate the potential virulence of outbreak isolates collected from the United States, proliferation rates of selected isolates were tested and compared to other isolates for which proliferation data had been previously obtained. In total, IPR values for eight of the nine VGIIc isolates were measured (Figure 7A). In addition, the type strains for VGIIa/major (R265) and VGIIb/minor (R272) were included as controls, and previously published data for other VGIIa and VGIIb isolates were included for comparisons [31]. On the basis of individual strains, seven of the eight VGIIc/novel isolates showed high IPR levels, with only a single outlier (EJB52) that had a low IPR value (0.97). Taken together, the median IPR value for VGIIc is significantly closer to that of VGIIa/major than to VGIIb/minor (Figure 7A). These results indicate that the VGIIc genotype has a similar intracellular phenotype, and thus virulence profile to the VGIIa/major genotype. This is noteworthy because previous analysis showed that the VGIIa/major genotype isolates from the outbreak had unusually high IPR values, and the VGIIc isolates from the same outbreak are here shown to have similarly high IPR values.
Another unique feature of the outbreak VGIIa/major isolates is the ability to form highly tubular mitochondria after intracellular parasitism, a characteristic that correlates with both IPR and murine virulence [31]. To explore the morphology of VGIIc isolates, we examined selected isolates in DMEM media and after exposure to macrophages. This analysis included two VGII environmental isolates (CBS8684, CBS7750) and four of the VGIIc/novel isolates. As expected, the vast majority of the mitochondria for all six isolates were non-tubular after exposure to DMEM media alone (Figure 7B). However, after exposure to macrophages, three of the four VGIIc isolates tested showed significantly higher percentages of tubular morphology (Figure 7C). The lone VGIIc isolate that did not exhibit this morphology (EJB52) was the same isolate that also had a low IPR value, and is thus an overall outlier for the VGIIc genotype.
When the results of IPR versus percentage of cells exhibiting tubular morphology were plotted, the graph showed a statistically significant correlation of the two measures with an R2 value of 0.85 (Figure 7D). These results further indicate that the VGIIc genotype is phenotypically similar to the Vancouver Island VGIIa/major outbreak strains. Our results also support evidence for similar mechanisms regulating the increased virulence seen in the novel VGIIc genotype. The exact roles that the mitochondrial tubular morphology might play in virulence are not yet known. However, the distinct phenotype is clearly unique to the outbreak isolates and is correlated with an increased ability to grow and divide within host innate immune cells.
The VGIIc isolates were found to be highly virulent in the murine inhalation model of infection. Two studies were conducted to examine virulence. In the first murine experiment a total of six isolates (n = 5 animals/isolate), were examined including two VGIIc isolates (Figure 8A). The VGIIa/major isolate R265 served as a positive control for high virulence, based on prior studies [6], and the VGIIc isolates EJB15 and EJB18 showed similar virulence with this well characterized virulent isolate. Additionally, two VGIIa isolates that are not hypothesized to be from the current Vancouver Island outbreak, including NIH444, which is fully identical across 34 markers, and isolate CA1014, which differs from R265 at VNTR34, show a significant reduction in virulence compared to the high virulence isolates (P<0.05). Finally, in accordance with previous studies, the VGIIb/minor type strain R272 from Vancouver Island was avirulent in this model.
The analysis of virulence within the VGII genotype was extended in a second experiment, in which 12 isolates (n = 9–10 animals/isolate) were examined. This study included two VGIIa/major isolates from the outbreak zone, two VGIIb/minor isolates from the outbreak zone, five of the novel VGIIc isolates, two VGIIa-related isolates that are not part of the outbreak, and the C. neoformans var. grubii type strain, H99. The H99 isolate used (H99S) has been shown to be highly virulent in the murine model of infection [44], [51].
As expected, all five of the VGIIc isolates from Oregon as well as the VGIIa/major isolates from Vancouver Island and Oregon, and the highly virulent H99 isolate exhibited a high level of virulence (median survival = 20.6 days). The VGIIb/minor isolates tested were significantly decreased in virulence compared to the more virulent VGIIa and VGIIc genotypes (P<0.005). The VGIIb isolate R272 was avirulent whereas the VGIIb isolate EJB53 from Oregon exhibited significantly less virulence compared to the VGIIa/major and VGIIc isolates (P<0.005, median survival = 46 days). Similar to the first animal study, two VGIIa isolates that differ at one or more molecular markers from the major VGIIa outbreak genotypes were also tested. The environmental isolate CBS7750 and a clinical isolate from South America ICB107 were significantly attenuated (P<0.005) (Figure 8B). These results provide further evidence that these are related to but distinguishable from isolates that are specific to the Vancouver Island outbreak, and subsequent United States expansion, and are decreased in ability to mount fatal infections in a mouse intranasal instillation model of infection.
The cause of infection was further evaluated by histopathological analysis of lung sections recovered from two infected animals per isolate at sacrifice. Harvested organs were processed and sectioned for slides with H&E staining. The lungs from the virulent isolates showed significant inflammation and numerous cryptococcal cells dispersed throughout the alveoli, in accordance with severe pulmonary infection. Our findings show that there are no major clinical differences between pulmonary infections with the infectious genotypes VGIIa/major (Figure 8C), and the novel VGIIc genotype (Figure 8D). These results further support similar disease progression caused by these two highly virulent outbreak genotypes.
The findings presented here document that the outbreak of C. gattii in Western North America is continuing to expand throughout this temperate region, and that the outbreak isolates in the United States of both the VGIIa/major genotype and the novel VGIIc genotype are clonally derived and highly virulent in host models of infection. These conclusions are based on an extensive molecular analysis of isolates collected from the United States (Table 2) and a comprehensive global collection of VGII isolates of diverse geographic origin (Figure S1), examining both conserved and divergent regions of the genome. The virulence analysis is based on assays in both murine derived macrophages and mice. These findings demonstrate that this emerging and fatal outbreak is continuing to expand, and that the virulence of these isolates is unusually high when compared to isolates of closely related but distinguishable genotypes found in other non-outbreak regions.
The continued expansion of C. gattii in the United States is ongoing, and the diversity of hosts increasing. Cases have been observed in urban and rural areas, and have occurred in a range of mammals [16], [52]. On Vancouver Island and the mainland of British Columbia, cases have been documented in marine and terrestrial mammals including cats, dogs, porpoises, ferrets, and llamas [15], [52], [53]. This trend has continued in the United States, with several cases in agrarian, domestic, and wild terrestrial mammals, as well as marine mammals, adding elk, alpacas, and sheep to the aforementioned list (Table S1) [13], [14], [17]. The co-expansion of the outbreak among mammals and humans is significant for several reasons. Non-migratory mammals serve as sentinels for disease expansion, particularly given that isolation of C. gattii from the environment is difficult, and not yet successful at all in Oregon. Additionally, the threat to agricultural and domestic animals is significant and thus the need for cooperation among health officials is critical. Finally, the widespread spectrum of disease illustrates that the organism is likely to be pervasive in the environment, and that physicians and veterinarians should be well informed of symptoms to facilitate early diagnoses, and successful isolate collection and tracking.
A major question in the study of this outbreak is whether sexual recombination, either within or between mating types, is occurring or has occurred in the region. The possibility of meiosis is important for two reasons. The first is that sexual recombination is postulated to be a driving force for the increased virulence of the VGIIa/major genotype, supported by the discovery of a diploid VGIIa/major isolate, an intermediate in unisexual mating (all nine VGIIc/novel isolates are haploid) [6], [36]. C. gattii has also been shown to undergo opposite sex mating in the laboratory, although this has not yet been observed to occur between two isolates of the VGII molecular type [36], [54], [55]. Studies in C. neoformans have shown that this related pathogen completes a full a-α sexual cycle in association with plants [56]. Additionally, a recent study of environmentally sampled Australian VGI isolates demonstrated evidence for recombination via both opposite and same-sex mating [50]. Taken together, available evidence indicates that both opposite and same-sex mating are naturally occurring in populations. This evidence lends support to the hypothesis that meiosis might be a factor in the forces that are driving high virulence in the outbreak region.
The second major event that results from sexual processes in the pathogenic Cryptococcus species is the formation of spores. Small spores ranging from 1–2 µm in diameter have been observed to be produced in large numbers as the result of opposite sex mating in both C. neoformans and C. gattii [57], [58]. Studies by Lin and colleagues showed that sexual spores can be produced as the result of a meiotic process occurring between cells of the same mating type, a process referred to as unisexual or same-sex mating [59]. Several studies have shown spores to be pathogenic in animal models of infection. Two previous studies both showed evidence for virulence of Cryptococcus spores, and in one case provided evidence for enhanced virulence compared to yeast cells [60], [61]. More recently, studies have shown that Cryptococcus neoformans spores are indeed virulent in the murine intranasal instillation model of infection [44], [62], providing evidence that spores should be considered as infectious propagules in models examining infections, expansion, and emergence of both C. neoformans and C. gattii. Given that all of the Pacific NW isolates are α mating type, and particles small enough to be spores are present in the air [26], [63], the most parsimonious model is that if these are spores, they are produced via α-α unisexual reproduction.
Our findings further indicate that mitochondria may play a significant role in the increased virulence seen in the outbreak isolates [31]. Tubular morphology and the increased ability to proliferate within immune cells indicate that the ability to proliferate and survive within host cells is fundamental to virulence. The possible role of mitochondrial involvement is intriguing and also increasingly relevant based on studies that have shown mitochondrial inheritance and recombination may impact C. gattii evolution, with the inheritance of the mitochondrial genome from the a mating type parent in opposite-sex mating [64], [65]. Future studies in this area should address the roles that mitochondrial genes, or nuclear genes that regulate mitochondria may play in the hypervirulence observed in the outbreak isolates. Furthermore, it may be that cell-cell fusion events via mating and mitochondrial exchange without meiosis or nuclear genetic exchange have played roles in recombination and virulence acquisition in naturally occurring C. gattii populations [64], [65].
A central question in the field lies in the possible origins of the virulent genotypes. For the VGIIa and VGIIc lineages, it is clear that those are unique to the Pacific NW, and either arose there locally, or were transferred from an under-sampled region (Australia, South America, Africa). Isolates that are related to, but distinct at one or more molecular marker from VGIIa have been identified in San Francisco (CBS7750), southern California (CA1014), and South America (ICB107). However, in each of these cases, the isolates are not identical with the VGIIa/major isolates from the Pacific NW. Whether the outbreak isolates are derived from these isolates, or alternatively that these isolates are derived from the outbreak lineage is at present unclear. In the VGIIb/minor outbreak lineage, isolates from Australia are identical at all 30 MLST loci and four VNTRs analyzed, and the most parsimonious model is that the two are directly related. While it is conceivable that both the Australian and the Vancouver Island VGIIb/minor genotype isolates were dispersed independently from another geographic locale, until isolates are identified conclusively from another locale the most parsimonious model is transfer from Australia to the Pacific NW. We note that a single isolate with a related but distinct genotype (isolate 99/473) from the Caribbean has been identified; and other isolates have been reported to share the VGIIb genotype but have been analyzed at a limited number of MLST markers (n = 7) which is insufficient to establish how closely related these isolates are to the outbreak VGIIb/minor genotype strains [29]. The origins of VGIIc are unclear, with the genotype possibly arriving in the Pacific NW from South America, Africa, Europe, or Australia. Alternatively, this novel unique genotype may have arisen locally.
As for the geographic origins of VGII diversity, this also remains to be established and may involve populations in Australia, South America, and Africa. It is clear that there is considerable diversity among isolates from South America. As we originally proposed as an alternative model [6], and has been independently presented by other investigators (W. Meyer, T. Boekhout, JP Xu, pers. comm.), South America may represent a source of diversity and ongoing generation of novel isolates. Analysis of 8 MLST loci in this study indicates that in South America and the Caribbean there are 14 genotypes seen in 21 isolates, while in North America only 3 genotypes have been observed through the analysis of 64 isolates (Figure S1). Additionally, there is accumulating evidence that fertile isolates of both a and α mating type are present in South America [29], and thus ongoing a-α opposite sex mating may be occurring there. It is also clear that a unique set of VGII isolates are circulating in Australia, and there is evidence for ongoing recombination in α only and a-α populations, suggesting that mating contributes to the generation of diversity in Australia [36], [49], [54], [55], [66], [67]. Finally, the analysis of global VGII isolates reveals genetic diversity in Africa, and given the recent findings that C. neoformans likely originated in sub-Saharan Africa (A. Litvintseva and T. Mitchell, pers. Comm.), further analysis of African C. gattii isolates is clearly warranted.
It remains possible that South America, Africa, or both represent the ancestral populations of C. gattii, and that more recent dispersal events from other established populations (for example, from Australia to the Pacific Northwest) have occurred to contribute to the outbreak. As yet, all of the isolates found in the Pacific Northwest are α mating type. Thus, if sexual reproduction is occurring in the Pacific Northwest, it would appear to involve same-sex mating occurring under environmental conditions. Recent studies have documented that C. neoformans and C. gattii are stimulated to undergo opposite-sex mating in laboratory conditions that simulate environmental niches (pigeon guano medium, co-culture with plants) and thus similar conditions may be necessary in nature [56], [68]. Overall, both the VGIIa/major and the VGIIc/novel genotypes contain a number of MLST loci that are thus far restricted to these lineages, and their origins remain to be identified.
Independently of the variables leading up to and influencing this outbreak, the major concern is and continues to be the inexorable expansion throughout the region. From 1999 through 2003, the cases were largely restricted to Vancouver Island. Between 2003 and 2006, the outbreak expanded into neighboring mainland British Columbia and then into Washington and Oregon from 2005 to 2009. Based on this historical trajectory of expansion, the outbreak may continue to expand into the neighboring region of Northern California, and possibly further.
The rising incidence of cryptococcosis cases in humans and animals highlights the need for enhanced awareness in the region, and those regions that may potentially become involved. While rare, little is currently known about how or why specific humans and animals become infected. Increased vigilance may decrease the time from infection to diagnosis, and thus lead to more effective treatment and a reduction in mortality rates. The potential dangers of travel-associated risks should be noted, as a growing number of cases attributable to travel within the Pacific NW region have been documented [69], [70]. Northern California has similar temperate climates to endemic regions within Oregon, leading to the hypothesis that the emergence may expand there, while expansion eastward may be limited by winters with average temperatures often below freezing [17].
The expansion of the outbreak into California is plausible based on several studies documenting the presence of C. gattii throughout the state and in Mexico. C. gattii molecular type VGII was environmentally isolated in the San Francisco area in 1990 (isolate CBS7750) [48], and there have also been two confirmed and one travel-associated case of C. gattii molecular type VGI in California. Of the VGI cases, one occurred in a male Atlantic bottlenose dolphin in San Diego, one was isolated from a liver transplant recipient in San Francisco, and the other from an otherwise healthy patient in North Carolina with travel history to the San Francisco region [71], [72], [73]. In addition C. gattii has been reported in southern California among a cohort of HIV/AIDS patients [74]. Recently, studies of clinical isolates from Mexico revealed all four molecular types of C. gattii to be present [75]. Taken together, the hypothesis that the virulent isolates from the Pacific NW will expand into California must be considered by both physicians and public health officials.
During the coming years, monitoring and researching the outbreak expansion as a multidisciplinary effort will be critical. The ability to bring diverse groups of professionals interested in C. gattii expansion has been greatly facilitated through the formation of the Cryptococcus gattii working group of the Pacific Northwest [17]. From a research standpoint, further examination of the molecular mechanisms underlying the increased virulence in both VGIIa/major and VGIIc/novel will be useful for the development of aggressive treatments that may be needed. Furthermore, increased efforts to determine the ecology and population dynamics of C. gattii in the region, and elucidating the evolutionary history of the VGIIc genotype will be critical to gain further insights into the origins of this unprecedented and frequently fatal fungal outbreak.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.